1: %\documentclass[12pt,preprint]{aastex}
2: %\documentclass[preprint2]{aastex}
3: %\documentclass[manuscript]{aastex}
4: \documentclass[10pt,twocolumns,apj]{emulateapj}
5:
6: \usepackage{afterpage}
7: %\usepackage{float}
8: %\restylefloat{figure}
9:
10:
11: \newcommand{\D}{\displaystyle}
12: \newcommand\eex[1]{\mbox{$\times 10^{#1}$}} % #1 x 10^#2
13: \newcommand\eez[1]{\mbox{$10^{#1}$}} % just 10^#2
14: \newcommand\hi{\ion{H}{1}}
15: \newcommand\hii{\ion{H}{2}}
16: \newcommand\civ{\ion{C}{4}}
17: \newcommand\ovi{\ion{O}{6}}
18: \newcommand\halpha{\mbox{H$\alpha$}}
19: \newcommand\vdev{\mbox{$v_{\rm dev}$}}
20: \newcommand\vhel{\mbox{$v_{\rm helio}$}}
21: \newcommand\lsun{\hbox{L$_{\odot}$}}
22: \newcommand\msun{\hbox{$M_{\odot}$}}
23: \newcommand\mdot{\hbox{$\dot M$}}
24: \newcommand\zsun{\mbox{$Z_{\odot}$}}
25: \newcommand\R{\mbox{$R$}}%% R filter
26: \newcommand\U{\mbox{$U$}}%% U filter
27: \newcommand\B{\mbox{$B$}}%% B filter
28: \newcommand\V{\mbox{$V$}}%% V filter
29: \newcommand\rsloan{\hbox{$r^\prime$}}
30: \newcommand\kps{\mbox{${\rm km~s^{-1}}$}}
31: \newcommand\cm{\mbox{${\rm cm^{-2}}$}}
32: \newcommand\nh{\mbox{$N_{\rm HI}$}}
33: \newcommand\mhi{\mbox{$M_{\rm HI}$}}
34: \newcommand\mtot{\mbox{$M_{\rm tot}$}}
35: \newcommand{\pv}{\mbox{$p$-$v$}}
36: \newcommand{\bc}{\begin{center}}
37: \newcommand{\ec}{\end{center}}
38:
39: % settings rom http://mintaka.sdsu.edu/GF/bibliog/latex/floats.html
40: \renewcommand{\topfraction}{0.9} % max fraction of floats at top
41: \renewcommand{\bottomfraction}{0.8} % max fraction of floats at bottom
42: % Parameters for TEXT pages (not float pages):
43: \setcounter{topnumber}{2}
44: \setcounter{bottomnumber}{2}
45: \setcounter{totalnumber}{4} % 2 may work better
46: \setcounter{dbltopnumber}{2} % for 2-column pages
47: \renewcommand{\dbltopfraction}{0.9} % fit big float above 2-col. text
48: \renewcommand{\textfraction}{0.07} % allow minimal text w. figs
49: % Parameters for FLOAT pages (not text pages):
50: \renewcommand{\floatpagefraction}{0.7} % require fuller float pages
51: % N.B.: floatpagefraction MUST be less than topfraction !!
52: \renewcommand{\dblfloatpagefraction}{0.7} % require fuller float pages
53:
54:
55: \slugcomment{Accepted by the Astrophysical Journal}
56:
57: \shorttitle{High-Velocity Clouds in M 83}
58: \shortauthors{Miller et al.}
59:
60: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
61: \begin{document}
62:
63: \title{High-Velocity Clouds in the Nearby Spiral Galaxy M 83}
64:
65: \author{Eric D. Miller\altaffilmark{1},
66: Joel N. Bregman\altaffilmark{2},
67: and Bart P. Wakker\altaffilmark{3}}
68:
69: \altaffiltext{1}{Kavli Institute for Astrophysics and Space Research,
70: Massachusetts Institute of Technology, Cambridge, MA 02139;
71: milleric@mit.edu}
72: \altaffiltext{2}{Department of Astronomy, University of Michigan,
73: Ann Arbor, MI 48105}
74: \altaffiltext{3}{Department of Astronomy, University of Wisconsin,
75: Madison, WI 53706}
76:
77: \begin{abstract}
78:
79: We present deep \hi\ 21-cm and optical observations of the face-on spiral
80: galaxy M 83 obtained as part of a project to search for high-velocity
81: clouds (HVCs) in nearby galaxies. Anomalous-velocity neutral gas is
82: detected toward M 83, with 5.6\eex{7} \msun\ of \hi\ contained in a disk
83: rotating 40--50 \kps\ more slowly in projection than the bulk of the gas.
84: We interpret this as a vertically extended thick disk of neutral material,
85: containing 5.5\% of the total \hi\ within the central 8 kpc. Using an
86: automated source detection algorithm to search for small-scale \hi\
87: emission features, we find eight distinct, anomalous-velocity \hi\ clouds
88: with masses ranging from 7\eex{5} to 1.5\eex{7} \msun\ and velocities
89: differing by up to 200 \kps\ compared to the \hi\ disk. Large on-disk
90: structures are coincident with the optical spiral arms, while unresolved
91: off-disk clouds contain no diffuse optical emission down to a limit of 27
92: \rsloan\ mag per square arcsec. The diversity of the thick \hi\ disk and
93: larger clouds suggests the influence of multiple formation mechanisms, with
94: a galactic fountain responsible for the slowly-rotating disk and on-disk
95: discrete clouds, and tidal effects responsible for off-disk cloud
96: production. The mass and kinetic energy of the \hi\ clouds are consistent
97: with the mass exchange rate predicted by the galactic fountain model. If
98: the HVC population in M 83 is similar to that in our own Galaxy, then the
99: Galactic HVCs must be distributed within a radius of less than 25 kpc.
100:
101: \end{abstract}
102:
103: \keywords{galaxies: individual (M 83) --- galaxies: ISM --- galaxies: kinematics and dynamics --- radio lines: galaxies}
104:
105: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
106: % SECTION -- Introduction
107: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
108: \section{INTRODUCTION}
109:
110: Our understanding of the gaseous content of our Galaxy has advanced greatly
111: in recent years as instruments sensitive to a variety of wavebands have
112: become available. One component of the Galaxy's gaseous neighborhood
113: remains poorly understood: the ensemble of high-velocity clouds (HVCs).
114: Initially discovered in \citeyear{Mulleretal63} by
115: \citeauthor*{Mulleretal63}, HVCs are clouds of \hi\ gas that deviate in
116: velocity from the differentially rotating disk of gas by 70--300 \kps.
117: These structures have been studied extensively in \hi\ emission, \halpha\
118: emission, and absorption against background point sources, with several
119: published catalogs detailing their observed characteristics
120: \citep{WvW91,WvW97,Wakker01,Putmanetal2002}.
121:
122: A cursory inspection of the HVC catalogs reveals a tremendous variation in
123: size, morphology, and velocity; this variation has led to a number of
124: formation scenarios that predict different physical properties. One
125: favored model is the Galactic fountain, a cyclic phenomenon whereby hot ($T
126: \sim \eez{6}$ K) gas is ejected into the lower halo ($z \lesssim 10$ kpc)
127: by supernova explosions \citep{ShapiroField76,Bregman80}. Here the gas
128: cools and loses its buoyancy, free-falling back to the disk in sheets at
129: mainly negative velocities approaching 100 \kps. A second scenario
130: suggests accretion of close companion galaxies, either through tidal or
131: ram-pressure stripping as a companion moves through the hot, extended
132: Galactic halo. This is undoubtedly the origin of the Magellanic Stream
133: \citep{Mathewson74}, and there is recent evidence for gaseous remnants of
134: satellite galaxies \citep{Lockman2003,Putmanetal2004}. Other scenarios
135: propose alternate extragalactic origins, such as accreted intergalactic
136: medium (IGM) material \citep{Oort66,Oort70,Oort81} or discrete remnants of
137: galaxy formation \citep{Oort66,Blitz99,BraunBurton99}. This last model
138: attributes the most compact HVCs (CHVCs) to an ensemble of primordial,
139: dark-matter-dominated ``minihalos'' streaming into the Local Group along
140: filaments of moderate overdensity. Such clouds would have a typical
141: distance of $\sim$ 750 kpc, and total mass of $\sim \eez{8}$--$\eez{9}$
142: \msun\ per cloud (assuming $\mhi = 0.1 \mtot$). Modeling by
143: \citet{SternbergMcKeeWolfire02} and simulations of CDM halo evolution by
144: \citet{Klypinetal99} and \citet{Mooreetal99} support a more local
145: distribution of $\sim$ 150 kpc. The clouds need not be dominated by dark
146: matter, as uneven CDM halo cooling can result in fragmentation and produce
147: a population of moderately-ionized, pressure-confined clouds embedded in a
148: hot corona \citep{MallerBullock04}.
149:
150: \begin{deluxetable*}{lrrrrr}
151: \tabletypesize{\normalsize}
152: \tablewidth{0pt}
153: \tablecaption{Predictions of HVC Production Scenarios
154: \label{tab:scenarios}}
155: \tablehead{
156: \colhead{scenario} &
157: \colhead{distance} &
158: \colhead{$|\vdev|$} &
159: \colhead{size} &
160: \colhead{$M_{\rm cloud}$} &
161: \colhead{$Z$} \\
162: \colhead{} &
163: \colhead{(kpc)} &
164: \colhead{(\kps)} &
165: \colhead{(kpc)} &
166: \colhead{(\msun)} &
167: \colhead{(\zsun)}
168: }
169: \startdata
170: IGM infall & $< 3$ & $< 150$ & $< 1$ & $<$\eez{4}
171: & 0.1--0.3 \\
172: Galactic fountain & $< 10$ & $< 150$ & $< 1$ & \eez{4}--\eez{5}
173: & $\ge$ 1 \\
174: companion stripping & 5--100 & $< 300$ & $\sim 1$ & \eez{5}--\eez{6}
175: & 0.1--1 \\
176: circumgalactic warm clouds & $\sim 150 $ & $< 300$ & $\sim 1$ & \eez{6}-\eez{7}
177: & 0.1--0.3 \\
178: circumgalactic DM halos & $\sim 150 $ & $< 300$ & $\sim 1$ & \eez{6}-\eez{7}
179: & 0.1--0.3 \\
180: Local Group DM halos & $\sim 750 $ & $< 300$ & 1--10 & \eez{8}
181: & 0.1--0.3 \\
182: \enddata
183: \tablecomments{\footnotesize{Sizes and masses shown are typical values based on the
184: median observed parameters of the HVC ensemble. Masses for the first three
185: scenarios assume \hi\ is the primary constituent of the cloud, while the
186: others assume an neutral gas fraction of 0.1 (warm clouds) or \hi\ to dark
187: matter ratio of 0.1 (dark matter halos). Velocities for the Galactic
188: fountain and IGM infall models assume an adiabatic corona of fully ionized
189: gas with $T \sim \eez{6}$ K.}}
190: \end{deluxetable*}
191:
192: One key discriminator between these production mechanisms is distance, yet
193: this parameter is difficult to determine since HVCs contain no visible
194: stars. The most direct method of searching for HVC absorption against
195: background halo stars has met with limited success until very recently,
196: yielding upper
197: limits on the distances to five HVCs and eight intermediate-velocity clouds
198: \citep[IVCs;][]{Wakker01}. The upper bounds to the vertical $z$ heights of
199: these HVCs range from 0.2 kpc to 7 kpc, and the corresponding mass limits
200: range from about 1 \msun\ for a very small HVC to 2\eex{6} \msun\ for
201: Complex A.
202: Recent work has placed stronger constraints on an additional four HVCs,
203: placing them at distances of $\sim$ 5--15 kpc
204: \citep{Thometal2006,Wakkeretal2007,Wakkeretal2008}.
205: In particular, \citet{Wakkeretal2008} infer a distance of 3.7--11.2 kpc for
206: Complex C, suggesting an \hi\ mass of 3--14\eex{6} \msun.
207:
208: Indirect methods have also been used to determine HVC distances,
209: such as constraining the degree of ionization on the cloud surface. It is
210: difficult to model the strength of the UV radiation field leaking out of
211: the Milky Way, although there are indications that the field is sufficient
212: to ionize at least the surfaces of an HVC \citep{BregmanHarrington86,
213: Bland-HawthornMaloney97}. Several groups have identified ionized ``skins''
214: in a number of HVCs
215: \citep{WeinerWilliams96,Tufteetal98,Bland-Hawthornetal98}, and observations
216: by \citet{Putmanetal2003} of \halpha\ emission in 25 HVCs limit their
217: distances to $5 < z < 40$ kpc.
218:
219: While recent progress has been made, large gaps remain in our
220: understanding of Galactic HVCs, the nature of the gaseous Galactic halo,
221: and the relationship between such a halo and the thin, star-forming disk.
222: The various HVC production scenarios predict different consequences for
223: galaxy formation and evolution. In Table \ref{tab:scenarios}, we outline
224: the predictions of the scenarios described above in terms of observables
225: (size, velocity) and physical properties (mass, metallicity). Our vantage
226: point remains an impediment to discriminating between these scenarios, as
227: we are embedded within the very medium we attempt to disentangle. The
228: superposition of material in position and velocity complicates the
229: analysis, and chance locations of statistically anomalous sources (such as
230: the Magellanic Stream and large HVC complexes) bias the conclusions.
231:
232:
233: These problems are largely overcome by observing a large sample of external
234: galaxies, constraining the fundamental properties of their diffuse halos,
235: and extrapolating to the Milky Way. Until recently, extragalactic \hi\
236: studies have concentrated on the structure of the kinematically cold disk,
237: and the discovery of anomalous-velocity material has been serendipitous.
238: Early evidence of anomalous \hi\ appeared in the face-on galaxies M 101
239: \citep{vanderHulstSancisi88,Kamphuisetal91}, NGC 628
240: \citep{KamphuisBriggs92}, and NGC 6946 \citep{KamphuisSancisi93}, with
241: detections of a few clouds of mass $\mhi > \eez{7}$ \msun. Integrated \hi\
242: profiles from a sample of 14 face-on galaxies show evidence for
243: high-velocity gas in 10 of the systems \citep{Schulmanetal1994}. More
244: recent observations of inclined galaxies have uncovered anomalous \hi\ gas
245: in every target imaged with sufficient depth. Deep radio synthesis
246: observations of NGC 891
247: \citep{SwatersSancisivanderHulst97,Oosterlooetal2007}, NGC 2403
248: \citep{Fraternalietal02a}, NGC 4559 \citep{Barbierietal2005}, NGC 253
249: \citep{Boomsmaetal2005},
250: and NGC 6946 \citep{Boomsmaetal2008}
251: have revealed
252: vertically extended, slowly rotating ``thick'' disks of \hi, suggestive of
253: Galactic fountain activity. Modeling of a subset of these data confirm
254: this conclusion, although some amount of IGM infall is required to prevent
255: escape of the fountain gas and to produce the gas kinematics that are
256: observed \citep{FraternaliBinney2008}. A GBT study of M 31 has uncovered
257: for the first time a population of discrete HVC-like \hi\ clouds in an
258: external galaxy \citep{Thilkeretal04}. These objects are all within 50 kpc
259: of the galaxy. Surveys of galaxy groups have so far failed to detect HVC
260: analogs without optical counterparts \citep[e.g.,][and references
261: therein]{Zwaan01,Pisanoetal2007}. Extrapolation by \citet{Pisanoetal2007}
262: of results from six Local Group analogs suggest the Galactic HVCs all lie
263: within 90 kpc of the Galaxy.
264:
265: In this paper, we describe the first results of a search for HVCs in nearby
266: external spiral galaxies. We concentrate on face-on galaxies, since the
267: line-of-sight velocity of the rotating \hi\ disk is minimized and the
268: parameter space available for HVC searching is maximized. These targets
269: provide a necessary complement to the numerous inclined spirals which have
270: been recently observed, as detailed above. By combining the \hi\ data with
271: deep optical surface photometry, we can constrain the amount of starlight
272: in any anomalous \hi\ clouds we find.
273:
274: The next section summarizes the observations and reduction of both the \hi\
275: and broad-band optical data. In Section 3, we discuss features of the bulk
276: \hi\ and optical disks. In Section 4, the cold \hi\ disk is modeled and
277: subtracted, and extended anomalous \hi\ emission is analyzed. Section 5
278: discusses the detection of discrete anomalous \hi\ emission, introducing a
279: new detection algorithm (further detailed in the Appendix \ref{app:snrch})
280: and presenting simulations that constrain the effectiveness of this
281: technique. The discussion in Section 6 places our results in context of
282: previous work and predictions. Our conclusions are summarized in Section
283: 7.
284:
285:
286: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
287: % SECTION -- Data
288: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
289: \section{OBSERVATIONS \& DATA REDUCTION}
290:
291: \begin{deluxetable*}{lrclr}
292: \tabletypesize{\normalsize}
293: \tablewidth{0pt}
294: \tablecaption{M 83 Properties
295: \label{tab:m83params}}
296: \tablehead{\vspace{0pt}}
297: \startdata
298: NGC number & 5236 & ~~~ & Type\tablenotemark{c} & SBc(s)II \\
299: RA (J2000) & 13h37m00.8s & & Inclination\tablenotemark{d} & 24\arcdeg \\
300: Dec (J2000) & $-$29\arcdeg51\arcmin59\arcsec & & Pos.\ angle\tablenotemark{d} & 225\arcdeg \\
301: $l$ & 314.6\arcdeg & & $D_{25}$\tablenotemark{b} & 12\farcm9 \\
302: $b$ & 32.0\arcdeg & & & 17 kpc \\
303: Distance\tablenotemark{a} & 4.5 Mpc & & $R_{\rm Holm}$\tablenotemark{e} & 7\farcm 9 \\
304: $v_{\rm hel}$\tablenotemark{b} & 503--516 \kps & & & 10 kpc \\
305: $B$\tablenotemark{b} & 8.2 & & Total mass\tablenotemark{e} & 1--5$\eex{11}$ \msun \\
306: $M_B$ & $-$20.1 & & HI mass\tablenotemark{e} & 6.1$\eex{9}$ \msun \\
307: $A_V$ & 0.22 & & 1\arcmin & 1.3 kpc \\
308: \enddata
309: \tablenotetext{a}{\footnotesize{\cite{Karachentsevetal02}}}
310: \tablenotetext{b}{\footnotesize{\cite{RC3}}}
311: \tablenotetext{c}{\footnotesize{\cite{RSA}}}
312: \tablenotetext{d}{\footnotesize{\cite{TalbotJensenDufour79}}}
313: \tablenotetext{e}{\footnotesize{\cite{HuchtBohn81}}}
314: \end{deluxetable*}
315:
316: M 83 is a nearby \citep[4.5 Mpc; ][]{Karachentsevetal02}, face-on grand-design
317: spiral galaxy located in the Centaurus A group. While the optical,
318: star-forming disk is small and well defined
319: \citep{TalbotJensenDufour79}, the \hi\ disk is very
320: extended, reaching to 6.5 times the optical Holmberg radius with 80\% of
321: the total \hi\ mass found outside this radius \citep{HuchtBohn81}. The
322: \hi\ disk is resolved into distinct rings and arms in this region,
323: exhibiting a high degree of warping \citep{TilanusAllenM83}. The basic
324: properties of M 83 are presented in Table \ref{tab:m83params}.
325:
326: \subsection{\hi\ 21-cm Data \label{sect:vlaobs}}
327:
328: The radio data for M 83 were obtained in three separate observations with
329: the VLA\footnote{The VLA is operated by the National Radio Astronomy
330: Observatory, a facility of the National Science Foundation operated under
331: cooperative agreement by Associated Universities, Inc.}
332: on 24 and 28 February and 1 March 1999. To match the beam size with the
333: scale of the expected emitting regions (1 kpc = 46\arcsec\ at 4.5 Mpc), we
334: chose to observe in the DnC configuration, in which the northern arm of the
335: array is in the more extended C configuration. For sources at southern
336: declinations such as M 83 (meridian altitude = 26\fdg1 at the VLA's
337: latitude of $+$34\fdg1), this configuration produces a more symmetric
338: $uv$-plane coverage pattern and a synthesized beam of about 35\arcsec\
339: FWHM (0.76 kpc at a distance of 4.5 Mpc). Each observing run consisted of a
340: 5 min observation of the primary flux calibrator 3C 286, with a scan
341: integration time of 20 s. This was followed with a series of $\sim$ 50 min
342: exposures of M 83 with integration times of 60 s, intermixed with 5 min
343: exposures of the secondary calibrator 3C 283. The total time spent on
344: source was 12.2 h.
345:
346: The VLA spectral line correlator was configured to observe in the 1420 MHz
347: frequency band, or ``L-band''. The correlator configuration and resulting
348: data parameters are summarized in Table \ref{tab:m83hiparams}. To allow
349: detection of \hi\ emission at highly anomalous velocities, we used the largest
350: bandwidth offered by the correlator, 3.125 MHz. This corresponds to a
351: velocity width of 660 \kps, of which only the inner 540 \kps\ were retained
352: because of edge noise. With the bandwidth centered on the systemic
353: velocity of M 83, this enabled detection of features deviating by $\pm 270$
354: \kps\ from systemic. We expected the features to have linewidths of
355: $\sim$~\nolinebreak20--30 \kps\ FWHM, therefore a velocity resolution of
356: 5--10 \kps\ would have been sufficient for this project. To make the
357: dataset generally usable to the community, we chose to observe at the full
358: spectral resolution, with no on-line Hanning smoothing applied, and using
359: the single IF correlator mode (sensitive only to right-handed circular
360: polarization). This resulted in 256 channels with a channel separation of
361: 12.2 kHz (2.58 \kps) and a channel width of 14.6 kHz (3.10 \kps).
362:
363: The $uv$ calibrator data were edited in a standard iterative way, with the
364: inner 75\% of the bandpass combined and inspected for obviously bad
365: samples. Once all anomalous samples were excised from the calibrator data,
366: the frequency-independent calibration was applied to the object data, which
367: were then edited. The calibration solutions had a maximum closure error of
368: 1\% in amplitude and 1\arcdeg\ in phase. The (frequency-dependent) bandpass
369: correction was applied using the secondary calibrator 3C 283 as reference,
370: again following an iterative editing/calibrating method. Application of
371: the bandpass correction produced closure errors of 0.5\% in amplitude and
372: 0.5\arcdeg\ in phase, however because of the limited integration time of
373: the calibrator, the noise in the final data cube was increased by a factor
374: of 1.8. With the editing and calibration completed, the $uv$ data were
375: Hanning smoothed in velocity space. Besides improving the signal-to-noise
376: in individual channels, smoothing with a Hanning kernel reduced the number
377: of channels from 256 to 128 and made them independent, with channel width
378: and separation equal at a value of 24.4 kHz (5.17 \kps).
379:
380: \begin{deluxetable*}{lrr}
381: \tabletypesize{\scriptsize}
382: \tablewidth{0pt}
383: \tablecaption{VLA Configuration and Data Parameters
384: \label{tab:m83hiparams}}
385: \tablehead{\vspace{0pt}}
386: \startdata
387: Dates of observation & \multicolumn{2}{r}{24 Feb, 28 Feb, 1 Mar 1999} \\
388: Time on source & \multicolumn{2}{r}{12.2 hr} \\
389: No.~antennas & \multicolumn{2}{r}{26} \\
390: Central freq. & \multicolumn{2}{r}{1.418 GHz} \\
391: Primary beam HPBW & \multicolumn{2}{r}{31\farcm2} \\
392: Synth.~beam HPBW & \multicolumn{2}{r}{$37\farcs2 \times 33\farcs1$}\\
393: Synth.~beam PA & \multicolumn{2}{r}{$-72.5\arcdeg$} \\
394: 1-$\sigma$ noise (per 5.17 \kps\ chan.) & \multicolumn{2}{r}{0.94 mJy/beam} \\
395: 1-$\sigma$ HI col. density (per 5.17 \kps\ chan.) & \multicolumn{2}{r}{4.6\eex{18} \cm} \\
396: 1 mJy/beam & \multicolumn{2}{r}{0.49 K} \\
397: \tableline
398: \multicolumn{3}{l}{\vspace{-5pt}} \\
399: & Raw data & Processed/cleaned \\
400: \multicolumn{3}{l}{\vspace{-5pt}} \\
401: \tableline
402: Bandwidth & 3.125 MHz & 2.56 MHz \\
403: & 660 \kps & 540 \kps \\
404: Velocity range & 183--843 \kps & 230--770 \kps \\
405: No.~channels & 256 & 105 \\
406: Channel separation & 2.58 \kps & 5.17 \kps \\
407: Channel width & 3.10 \kps & 5.17 \kps \\
408: \enddata
409: \end{deluxetable*}
410:
411: The ultimate success of any interferometry observation hinges on whether a
412: reliable map of source intensity in real space can be produced from
413: incomplete sampling of the coherence function in the $uv$ plane.
414: At the imaging stage, weighting visibilities
415: based on their local population density or distance from the tracking
416: center affects the strength of sidelobe structure. We applied two
417: weighting schemes to improve the quality of the principal solution: first,
418: the ``robust'' scheme developed by \citet{Briggs95} to produce a mixture of
419: natural and uniform weighting with a single parameter; and second, a
420: Gaussian taper function to downweight visibilities in the sparsely-sampled
421: outer $uv$ plane. We used the {\tt AIPS IMAGR}\ task \citep{AIPSCookbook}
422: and applied a variety of values for these two weighting schemes,
423: empirically determining the best compromise between sidelobe structure,
424: resolution and sensitivity. The best compromise among all the
425: considerations was a ROBUST value of 0 and a Gaussian $uv$-taper with
426: $r_{.3} = 6$ k$\lambda$, where $r_{.3}$ is the half-width at 30\% of
427: maximum. This weighting scheme increased the noise in the final image by a
428: factor of 1.2 compared to pure natural weighting and no $uv$-taper.
429:
430: A dirty image cube was produced from the full $uv$ dataset, and channels
431: free of \hi\ line emission were identified. The continuum was subtracted
432: by fitting a first-order polynomial to 10 line-free channels on each edge
433: of the band and subtracting this from all channels. A continuum-free dirty
434: image cube was produced from this $uv$ dataset, using a cell size of
435: 7\arcsec\ per pixel and only including the central 105 channels to
436: eliminate edge noise.
437:
438: To correct for incomplete sampling in the $uv$ plane and reduce the level
439: of the sidelobes, one typically uses the dirty image to iteratively
440: construct a model of the true brightness distribution. This is the
441: strategy behind the CLEAN algorithm \citep{Hogbom74,Clark80}, under
442: which the brightness distribution is modeled iteratively as a set of
443: $\delta$-function ``clean components''. CLEAN is very
444: effective at removing sidelobes from point sources, since the clean
445: components themselves are modeled as points. Spatially-extended \hi\
446: emission, however, is poorly modeled by a set of point sources.
447: Furthermore, while the strength of extended emission is usually relatively
448: weak compared to a continuum point source, and thus the sidelobes
449: associated with a single grid point often fall below the noise in a dirty
450: map, the summation of sidelobes due to complex extended structure can be
451: significant and difficult to disentangle.
452:
453: To combat this problem, another method of image deconvolution was developed
454: by \citet{WakkerSchwarz88}. Their Multi-Resolution CLEAN (MRC) runs the
455: CLEAN algorithm, but operates on the point-like and extended sources
456: separately. First, the dirty map and beam are smoothed then
457: subtracted from the original map and beam to obtain a
458: full-resolution difference map and beam. Clean beams are constructed for
459: the smooth and difference beams, and CLEAN is performed on the
460: smooth and difference maps separately. By performing CLEAN on a smoothed
461: map, two major improvements are made in reproducing extended emission:
462: the clean beam is closer in size to the extended sources one is
463: attempting to model; and the signal-to-noise is improved, allowing
464: detection of fainter sources. The original resolution is retained (and
465: point sources are deconvolved) by cleaning the difference map separately
466: and then combining the two resulting cleaned maps along with the
467: full-resolution residuals. A detailed explanation of the method and the
468: exact scaling parameters used are presented by \citet{WakkerSchwarz88}.
469:
470: After trying standard CLEAN and MRC and comparing the results, it was
471: determined that MRC was more effective in removing sidelobes. Our
472: procedure was to clean the full region of each channel containing
473: emission above the 3-$\sigma$ level, then examine the map
474: for remaining strong sidelobes, which were present in most of
475: the channels. The output cube was spatially smoothed to twice the beam
476: size and binned up by 5 channels, and regions that contained significant
477: emission were manually delineated with polygons. The cleaning was repeated
478: with MRC, using the original dirty cube and allowing clean components only
479: within the specified region of each channel. Successively deeper cleaning
480: was performed, down to a threshold of 1-$\sigma$.
481:
482: The 1-$\sigma$ noise level, calculated from the emission-free regions of
483: each image plane, and other parameters for the cleaned data cube are
484: summarized in Table \ref{tab:m83hiparams}. The noise of 0.94 mJy/beam is
485: larger than the theoretical value of 0.7 mJy/beam for this instrumental
486: setup and image weighting parameters, and this is likely a result of the
487: bandpass calibration. The noise is similar from channel to channel,
488: varying no more than 5\%, as is shown in Figure \ref{fig:m83channoise}.
489:
490: \begin{figure}
491: \plotone{f1.eps}
492: \caption{The 1-$\sigma$ noise (crosses) and the peak \hi\ flux (circles) in
493: individual channels of the cleaned M 83 data cube. The noise levels have
494: been multiplied by a factor of 100 and placed on the same scale as the peak
495: flux to facilitate comparison. The noise is close to constant over the
496: entire cube, with an RMS of 1.7\% and maximum variations less than 5\% of
497: the mean.}
498: \label{fig:m83channoise}
499: \end{figure}
500:
501: To reduce image size and ease further computational analysis, the image
502: cube was binned $2 \times 2$ in space. With a final pixel size of
503: 14\arcsec, the $\sim 35$\arcsec\ (FWHM) synthesized beam is still
504: well-sampled. Noise was not reduced, however, since neighboring pixels are
505: correlated. The channel maps for the cleaned data cube are shown in Figure
506: \ref{fig:m83chanmaps}.
507:
508: \subsection{Optical Data}
509:
510: The optical data for M 83 were taken with the 0.6m Curtis Schmidt telescope
511: located at Cerro Tololo Inter-American Observatory\footnote{CTIO is part of
512: the National Optical Astronomy Observatory, which is operated by the
513: Association of Universities for Research in Astronomy (AURA), Inc., under
514: cooperative agreement with the National Science Foundation.}
515: (CTIO) in March 1999. Use of a thinned SITe 2K CCD chip resulted in a
516: field of view of nearly 70\arcmin\ with 2.3\arcsec\ square pixels. To
517: reduce flat-fielding errors that could be produced by non-repeatable filter
518: wheel positioning and dislodged dust particles, we used a single Sloan
519: \rsloan\ filter, keeping the filter wheel unchanged throughout the run. A
520: standard observing scheme was followed, with twenty bias frames and six to
521: eight twilight flats taken each evening, and standard star fields imaged at
522: a range of airmass throughout the night. Approximately half of the
523: telescope time was spent imaging galaxies and half imaging blank sky flats,
524: with exposure times of 15 minutes each. The total exposure times were 10
525: hours on source and 8.75 hours on blank sky. Dark exposures taken during
526: one cloudy night indicated no measurable CCD dark current.
527:
528: The initial reduction procedure followed a standard formula. The bias and
529: overscan were removed from all twilight, sky and object frames. A
530: first-order flatfield was created for each night from the twilights flats.
531: This flatfield was applied to the sky flats, which were co-added and
532: smoothed with a $128\times128$ boxcar to remove stars. This sky correction
533: image was normalized and multiplied by the first-order twilight flat field
534: to produce a master flat field for each night, which was applied to
535: the M 83 data. The individual galaxy frames were examined by eye, and
536: eight of the 40 images contained undesirable features such as strong
537: scattered light and were rejected. The remaining frames were
538: sky-subtracted, aligned, and co-added. For proper noise estimation in
539: future analysis, an average sky value of 4423 ADU was added back in to the
540: co-added frame. Finally, the poorly-sampled outer regions were clipped,
541: leaving a final 72\arcmin$\times$69\arcmin\ image with a total integration
542: time of 8.0 hours.
543:
544: Photometric calibration was accomplished with observations of standard
545: stars. The number of standard stars for the Sloan filter system was
546: limited at the time of observations, but using a star list provided
547: by J.A.~Smith (private communication), we identified a number of
548: standard stars within \citet{Landolt92} fields. Nightly imaging of these
549: fields yielded 50 observations of seven standard stars covering a range of
550: airmasses ($1.07 < \sec{z} < 2.0$) and colors ($0.0 < B-V < 1.0$). For a
551: pixel size of 5.31 square arcsec and an effective exposure time of 900 s,
552: the surface brightness corresponding to 1 ADU pix$^{-1}$ in the final
553: averaged galaxy frame was $30.11 \pm 0.05$ \rsloan\ magnitudes per square
554: arcsec. The average sky value of 4423 ADU corresponds to a surface
555: brightness of 21.04 \rsloan\ magnitudes per square arcsec, and the
556: pixel-to-pixel noise in the final image (6.5 ADU) corresponds to 28.1
557: \rsloan\ magnitudes per square arcsec. Our 3-$\sigma$ limiting surface
558: brightness is 26.9 \rsloan\ magnitudes per square arcsec, not accounting
559: for systematic errors caused by scattered light and large-scale
560: flatfielding effects.
561:
562: Foreground stars were removed by modeling the stellar point spread function
563: (PSF) and masking each star. Using the DAOPHOT \citep{Stetson87} package
564: in IRAF\footnote{IRAF is distributed by the National Optical Astronomy
565: Observatories, which are operated by the Association of Universities for
566: Research in Astronomy, Inc., under cooperative agreement with the National
567: Science Foundation.}, we estimated the brightness of all stars in the frame
568: by searching for 3-$\sigma$ peaks in the image and performing simple
569: aperture photometry on them. This procedure produced 38,721 stellar
570: objects; of these, 11 bright, isolated stars were chosen to construct the
571: PSF. The PSF stars had an average (standardized) brightness of \rsloan =
572: 16.3, with one moderately saturated star included to improve the fit in the
573: wings. The PSF was constructed from the unsaturated regions of the PSF
574: stars, using an elliptical Moffat function with $\beta = 2.5$ as the
575: analytic PSF component plus a single look-up table for the residual PSF
576: component. The PSF was scaled to each PSF star and subtracted, leaving a
577: residual image and revealing previously-unidentified stars in the PSF
578: stars' wings. These were fit with the PSF and removed from the original
579: frame, and a new PSF was constructed. This procedure was iterated several
580: times until the companion stars were completely removed and the PSF did not
581: improve between iterations.
582:
583: Once the PSF was modeled, it was used to identify and measure all stars in
584: the field. Of the stars originally identified by aperture photometry, a
585: subset of 33,950 were found to have profiles similar to the PSF. This
586: amounts to an average of 6.8 field stars per square arcmin, down to a
587: detection limit of \rsloan = 23.8. The core of a star becomes saturated at
588: $\rsloan \sim 14.5$, but by fitting to the unsaturated wings, stars
589: down to $\rsloan \sim 11$ could be reliably measured.
590:
591: Stars were removed by masking each one out to the radius at which its flux
592: fell below a specified threshold. This method is a compromise between
593: elimination of polluting starlight and retention of usable data. To
594: construct the mask, the model PSF was averaged radially to a 1-d function,
595: scaled to each star according to its magnitude, and used to determine the
596: radius at which the stellar flux fell below threshold. A threshold of 2
597: ADU, well below the 7 ADU RMS noise in the co-added frame, was a
598: satisfactory compromise between starlight removal and pixel retention.
599: Typical PSF reference stars were masked out to a radius of 7 pixels
600: (16\arcsec), and the brightest unsaturated stars were masked out to a
601: radius of 9 pixels (21\arcsec), about 26 times larger than the average
602: seeing HWHM of about 0.8\arcsec.
603:
604: \begin{figure*}[p]
605: \plotone{f2a.eps}
606: \caption{M 83 \hi\ channel maps. Heliocentric velocities are marked in the
607: upper right of each channel map. The grayscale ranges from $-3$ to 15 mJy
608: to show the fainter structure in each channel, therefore most of the bright
609: features appear saturated. Only channels with measurable \hi\ emission are
610: displayed.}
611: \label{fig:m83chanmaps}
612: \end{figure*}
613:
614: \begin{figure*}[p]
615: \plotone{f2b.eps}
616: \\
617: Figure~\ref{fig:m83chanmaps} (continued)
618: \end{figure*}
619:
620: \begin{figure*}[t]
621: \plotone{f2c.eps}
622: \\
623: Figure~\ref{fig:m83chanmaps} (continued)
624: \end{figure*}
625:
626: %\clearpage
627:
628: \begin{figure*}[htp]
629: \plotone{f3.eps}
630: \caption{\hi\ column density map for M 83. The inner circle shows the FWHM
631: of the primary beam, while the outer circle shows the full width at 10\% of
632: peak sensitivity. Contours show the mean \hi\ velocity and range from 350
633: to 650 \kps\ in steps of 10 \kps\ (twice the velocity resolution). Column
634: density is in units of \eez{20} \cm. The marked features are discussed in
635: Section \ref{subsect:m83:hidisk:disk}.}
636: \label{fig:m83mom0}
637: \end{figure*}
638:
639: Some contaminating features did not conform to the shape of the PSF and
640: were ignored by the above procedure. These included background galaxies,
641: very bright saturated stars, diffraction spikes, and charge transfer
642: (``bleeding'') features, and they were masked by hand.
643:
644:
645: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
646: % SECTION -- The Normal Galactic Disk of M 83
647: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
648: \section{The Normal Galactic Disk of M 83}
649: \label{subsect:m83:hidisk}
650:
651: \subsection{\hi\ Moment Map Construction}
652: \label{subsect:m83:hidisk:mom}
653:
654: The bulk distribution and kinematics of emission-line gas are most easily
655: summarized by moment maps, including the total \hi\ surface brightness (0th
656: moment) and the intensity-weighted mean velocity field (1st moment). To
657: construct the moment maps, the image cube was first smoothed spatially to a
658: round beam of 52\arcsec\ FWHM, roughly 1.5 times the size of the original
659: synthesized beam. This smoothing is done to improve the S/N and ensure
660: that the faint extended components of emission peaks are included in the
661: summation. Pixels with flux greater than 4-$\sigma$ were included,
662: as were immediate neighbor pixels in both space and velocity.
663:
664: The moment maps are shown in Figures \ref{fig:m83mom0} and
665: \ref{fig:m83mom1}. The primary beam attenuation correction was applied to
666: only the moment 0 map (which is shown in column density units) after
667: integrating in velocity. Throughout this work, all \hi\ intensity maps
668: displaying either integrated \hi\ emission or \hi\ column density have been
669: corrected for the primary beam sensitivity.
670:
671: \begin{figure*}[htp]
672: \plotone{f4.eps}
673: \caption{Intensity-weighted mean velocity map for M 83. Notations and
674: contours are the same as in Figure \ref{fig:m83mom0}, and velocities are
675: heliocentric.}
676: \label{fig:m83mom1}
677: \end{figure*}
678:
679: The total \hi\ mass determined from the \hi\ surface brightness map is
680: 4.7\eex{9} \msun. This is lower than the distance-corrected value of
681: 6.1\eex{9} obtained by \citet{HuchtBohn81}, who mapped the emission out to
682: about twice the radius as the present study. Their single-dish scans were
683: impervious to zero-spacing issues of interferometry and recovered all of
684: the emission in the beam. Our result is therefore in reasonable agreement
685: with their value.
686:
687: \subsection{\hi\ Disk Features}
688: \label{subsect:m83:hidisk:disk}
689:
690: The total \hi\ map (Figure \ref{fig:m83mom0}) is a good tracer of spatial
691: morphology, and it shows a number of notable features. The inner disk of M
692: 83 contains a minimum in the \hi\ approximately 1\arcmin\ (1.3 kpc) in
693: diameter, with a measured column density of 4\eex{20} \cm. This feature is
694: surrounded by the inner gaseous disk, which extends to a radius of
695: 6\arcmin\ (8 kpc) and ranges in column density from a minimum of
696: 7\eex{20} \cm\ to a (probably optically thick) maximum of 1.7\eex{21} \cm.
697: The inner disk shows a hint of spiral structure, with an irregular pattern
698: of bright spots, and an aspect ratio close to unity, as one would expect
699: for a nearly face-on galaxy. The small-scale structure is unresolved in
700: our map, but it agrees qualitatively with the results of
701: \citet{TilanusAllenM83}, who studied M 83 at higher spatial resolution.
702:
703: Beyond 5\arcmin, the \hi\ brightness falls off abruptly, with $\nh \lesssim
704: 4\eex{20}$ \cm\ in the region immediately around the inner disk. In the
705: regions NE and SW of the inner disk, the column density becomes very low,
706: approaching 5\eex{19} \cm. Further out, there is evidence of a bright ring
707: of \hi\ at a radius of 10\arcmin\ (13 kpc). The ring has a peak column
708: density of 1\eex{21} \cm, similar to the inner disk, and the projected
709: thickness varies between 60\arcsec\ and 80\arcsec. With the smoothed beam
710: size (52\arcsec) this scale is unresolved, but the width can be no more
711: than 1\arcmin\ (1.3 kpc) after removal of the instrumental profile. The
712: emission in the ring appears clumpy, with the brightest region in an arc on
713: the west side of the galaxy. In addition, the aspect ratio of the ring is
714: quite different from that of the inner disk, with semi-major and semi-minor
715: axes of 10\arcmin\ and 5\arcmin\, respectively. The position angle is
716: close to that of the inner disk (see below).
717:
718: Outside of the ring, the \hi\ spatial distribution becomes irregular, with
719: arms and streams at various projected orientations. One arm
720: appears just beyond the ring on the NW side, and spans about 25\arcmin\ (33
721: kpc) to a point 16\arcmin\ (21 kpc) NE of the disk center. There is a
722: corresponding arm on the SE to S side of the galaxy, at a similar
723: projected distance. Two notable features are the arms further west of
724: this, which extend from a common point 14\arcmin\ (18 kpc) west of the
725: center and extend toward the north. The longer of these arms reaches a
726: projected galactic radius of 26\arcmin\ (34 kpc) and is clearly detached
727: from the rest of the disk. Both NW arms appear to be mirrored in the SE,
728: and similar mirroring can be seen between other features and clumps in the
729: extended \hi. Since the primary antenna sensitivity falls to 10\% of the
730: central value for the most extended arms, these features could be even more
731: extended than observed. It is clear that the arms are distinct from the
732: main \hi\ disk, down to our detection limit. The symmetrical nature and
733: varying orientations of the outer arms indicates either a warp that varies
734: with radius or a collection of kinematically distinct orbiting streams.
735:
736: Analysis of the kinematic morphology can improve our understanding of the
737: complex outer disk features, and this is most thoroughly done by combining
738: the velocity field map, the individual channel maps, and the data cube as a
739: whole. The velocity map is presented in both Figure \ref{fig:m83mom0} (as
740: contours) and Figure \ref{fig:m83mom1} (as an image map and contours).
741: In both figures, the inner disk shows a typical differential rotation
742: pattern (a so-called ``spider diagram''), with a central systemic value of
743: $\sim 500$ \kps. To estimate the systemic velocity and orientation, we fit
744: the central 5\arcmin\ with a Brandt rotation curve using the {\tt GAL} task
745: in {\tt AIPS} \citep{AIPSCookbook}. We find a systemic velocity of
746: $v_{\rm hel} = 513 \pm 2$ \kps, where the error is estimated from the
747: spread of best-fit values at different radii, and using the optical
748: definition
749: $v_{\rm hel,opt} = c (\lambda_{\rm obs}-\lambda_{\rm ref}) / \lambda_{\rm
750: ref}$. The best-fit position angle is $225\arcdeg \pm 2\arcdeg$ (measured
751: to the receding side from north to east), and the best-fit inclination is
752: $i = 23\arcdeg \pm 2\arcdeg$.
753:
754: Our estimates of the inner \hi\ disk parameters are consistent with
755: previous results. Previous values for the systemic velocity range from 505
756: \kps\ determined by \citet{Comte81} through \halpha\ interferometry to 516
757: \kps\ quoted by \citet{RC3} and determined by \hi\ observations. The
758: HIPASS Bright Galaxy Catalog reports a value of $513 \pm 2$ \kps\
759: \citep{HIPASS04}. We adopt 513 \kps\ as the systemic velocity for the
760: remainder of this work. Our estimates of the orientation and inclination
761: are consistent with the those quoted by \citet{TilanusAllenM83}, who have
762: data with higher spatial resolution. Therefore we adopt $i = 22\arcdeg$
763: for our inclination, and note that the maximum rotation rate of the
764: inner disk is $v_{\rm rot,max} = (v_{\rm hel,max}-v_{\rm sys}) / \sin{i} =
765: 200$ \kps.
766:
767: Outside of the inner disk, a discontinuity develops in the rotational
768: velocity. This is most clearly seen in Figure \ref{fig:m83mom0}, where the
769: velocity contours bend abruptly into a warp, and in the channel maps for
770: 440 \kps $< \vhel <$ 540 \kps\ (see Figure \ref{fig:m83chanmaps}). On the
771: W side of the galaxy, the warp correlates well with the \hi\ ring in
772: projection, but on the E side the warp peaks inside of the ring. The
773: isovelocity contours are shifted by 2\arcmin\ counter-clockwise from where
774: they would be under ordinary differential galactic rotation. Beyond the
775: ring, the isovelocity contours switch direction again, and they no longer
776: reflect a flat rotation curve. In addition, the aspect ratio of the disk
777: increases in this region, indicating that the axis of rotation is becoming
778: more inclined to our line of sight as the radius increases. The highest
779: radial velocities are seen near the very edge of the detected \hi\
780: distribution, at a distance of 8\arcmin\ (11 kpc) N and S of the galactic
781: center.
782:
783: It is apparent from this analysis that the inner disk, the ring, and the
784: outer arms are kinematically distinct features. The ring and outer arms
785: could be considered anomalous features themselves, however we wish to
786: search for anomalous-velocity emission superposed on the bulk \hi\
787: emission, and for this purpose we assume the outer arms are part of an
788: extended disk of varying inclination.
789:
790: \subsection{Large-Scale Optical Features}
791:
792: The final optical image for M 83 shows many of the typical features of a
793: face-on, grand-design spiral galaxy (see Figure \ref{fig:m83opt}).
794: The spiral arms are seen radiating from a central bar-like structure, and
795: these arms contain a number of star clusters and dust lanes. The integrated
796: \hi\ intensity traces the spiral arms closely. The spiral structure is
797: apparent only out to a radius of about 4.5\arcmin\ (6 kpc) in the north and
798: 5.5\arcmin\ (7 kpc) in the south, at which point the surface brightness has
799: dropped to 22.5 \rsloan\ magnitudes per square arcsec. This is the same
800: radius at which the inner \hi\ disk falls off abruptly and at which the
801: warp in the velocity field appears.
802:
803: \begin{figure*}
804: \plotone{f5.eps}
805: \caption{Optical image for M 83, displayed in linear units of $\rsloan$
806: magnitude per square arcsec. The contours show the column density of \hi,
807: starting at 5\eex{19} \cm\ and increasing by factors of 2. Also labeled is
808: KK208, a faint companion galaxy or stellar stream to the northeast.}
809: \label{fig:m83opt}
810: \end{figure*}
811:
812: Beyond the inner optical disk, there is an extended optical envelope devoid
813: of any structure. This envelope reaches a surface brightness of 26
814: \rsloan\ mag per square arcsec at a radius of about 10\arcmin\ (13 kpc),
815: falling to undetectable levels at the location of the \hi\ ring. In
816: addition, the aspect ratio of the extended light, appearing stretched
817: northeast to southwest, is different from that of the inner disk, which is
818: nearly circular.
819:
820: An arc-like feature can be seen projected 20\arcmin\ (26 kpc) northwest of
821: center, extending 12\arcmin\ (16 kpc) northeast to southwest. This feature
822: is barely visible in the Palomar Sky Survey plate covering this region, and
823: it has been cataloged as a dwarf galaxy by
824: \citet{KarachentsevaKarachentsev98}, who designate it KK208. The surface
825: brightness across the feature varies from 25--26 \rsloan\ mag per square
826: arcsec, although there are three bright stars along the arc that complicate
827: masking and flux measurements. There is no measurable \hi\ anywhere along
828: the feature down to a 3-$\sigma$ column density limit of 1.5\eex{19}
829: \cm. The arc lies between two outer \hi\ arms, and it has a shape similar
830: to those arms.
831:
832:
833:
834: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
835: % SECTION -- Anomalous-Velocity HI: Extended Emission
836: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
837: \section{Anomalous-Velocity \hi: Extended Emission}
838: \label{sect:m83:hiavext}
839:
840: \subsection{\hi\ Disk Modeling}
841: \label{subsect:m83:hiavext:model}
842:
843: To search for anomalous-velocity material, one must first establish the
844: criteria that make such material ``anomalous''. The difficulty of this can
845: be appreciated by considering the ensemble of Galactic HVCs. A
846: commonly-used definition limits membership to complexes with velocities
847: deviating by greater than 100 \kps\ (or sometimes 90 \kps) relative to the
848: Local Standard of Rest (LSR). This definition was intended to separate the
849: HVCs from the so-called Intermediate Velocity Clouds (IVCs) and gas in the
850: plane. However, such a definition does not take into account differential
851: galactic rotation, an effect which can vary the observed velocity of a
852: cloud by as much as 100 \kps, depending on the line of sight direction
853: \citep{Wakker91a}. To account for this effect, some authors have used the
854: velocity relative to the so-called Galactic Standard of Rest (GSR), which
855: is a point at the location of the Sun that does not participate in galactic
856: rotation. This definition works well for gas at high $|z|$. However, it
857: over-corrects the velocity of any cloud that has a rotational component of
858: motion, such as may be found closer to the plane. \citet{Wakker91a}
859: introduced the concept of ``deviation velocity,'' which is the difference
860: between the observed LSR velocity of the cloud and the maximum velocity
861: expected along that line of sight due to differential rotation, assuming a
862: simple model for the gaseous Galactic disk. At low latitudes in the Milky
863: Way this varies greatly, but at latitudes above $\sim 40\arcdeg$ the
864: velocity limit is around 60 \kps.
865:
866: For external spiral galaxies, and especially face-on galaxies, the
867: situation is simplified in that the bulk disk velocity is easily measured
868: along any line of sight. Material at velocities different from this,
869: assuming some velocity threshold or kinematic model for the disk, would be
870: kinematically distinct from the cold \hi\ disk and therefore anomalous.
871: Hereafter we define anomalous-velocity material to be any \hi\ emission
872: that does not participate in the normal differential rotation of the bulk
873: of \hi\ along that line of sight. Implicit in the definition is the
874: inclusion of spatially isolated \hi\ clumps, since the disk rotation is
875: undefined in regions where the disk does not exist. Therefore an
876: anomalous-velocity clump (hereafter AVC or AV gas) can be anomalous in
877: velocity or position.
878:
879: As low-velocity AV gas could overlap and merge with the disk velocity, it
880: is useful to remove any \hi\ disk emission. Additionally, it is important
881: to ignore this emission in a non-interactive source detection scheme. We
882: modeled the disk of M 83 by fitting a Gaussian to the spectrum along every
883: pixel of the cleaned image cube. The fitting was performed only along
884: lines of sight where the peak exceeded 3.5-$\sigma$ (3.3 mJy/beam), and
885: only included channels whose flux exceeded 1.7-$\sigma$ (1.6 mJy/beam).
886: These criteria were applied to ensure that faint, velocity-broadened
887: emission would not contaminate the thin disk fit.
888:
889: Disk fitting and removal failed at warp discontinuities, where the \hi\
890: disk has multiple velocity components across a scale of about 1\arcmin. It
891: is not clear whether this spatial overlap of kinematically distinct disk
892: components is a real feature or an effect of beam smearing, but in either
893: case it poses a difficulty for effective disk removal. It is clear that
894: the warp features are not AV material in our definition, since they define
895: the disk velocity. We therefore attempted to remove this emission by
896: refitting the original data cube with multiple Gaussian components. For
897: lines-of-sight where the residual flux matched the above single-Gaussian
898: criteria (i.e., peak above 3.5-$\sigma$, 3 or more adjacent pixels with
899: flux above 1.7-$\sigma$), a two-component Gaussian was fit. The fitting
900: threshold was kept high so that true AV gas would escape the fit.
901:
902: To estimate the quality of the disk model, we examined the reduced $\chi^2$
903: and covariance matrices of all the fits performed. The fits that included
904: only a few data points were obviously poorly constrained, but inspection of
905: the residual data cube showed that even these fits were reasonable. There
906: were several lines-of-sight that had noticeably bad fits, but these were
907: generally isolated pixels whose neighbors had similar parameters to
908: each other. To remove the outliers, we performed a median box filtering on
909: the fitting parameter images, using a kernel width of 3 pixels.
910: The disk model and residual cube were reproduced from the adjusted fitting
911: parameters, and inspection showed that the poor fits were removed and the
912: disk subtraction was improved.
913:
914: The fitted line-of-sight velocity dispersion varies across the galaxy,
915: as can be seen in Figure \ref{fig:m83gparsig}. It reaches a maximum of 20
916: \kps\ in the center, averages about 15 \kps\ in the central \hi\ disk and
917: ring, and falls to 10 \kps\ or less in the extended disk and arms.
918:
919: \subsection{Extended, Disk-Like AV Gas}
920: \label{subsect:m83:hiavext:beard}
921:
922: Removal of the \hi\ disk results in a residual cube with significant \hi\
923: emission remaining. For ease of display, we created position-velocity
924: (\pv) plots, or slices through velocity along a line of specified location,
925: length, and position angle. The slices were taken parallel to the major
926: axis at a spacing of 70\arcsec\ (twice the beamwidth). The value of a
927: given pixel in the \pv\ image equals the distance-weighted average of the
928: nearest four pixels in the channel. A sample of slices is shown in Figure
929: \ref{fig:m83slices} for both the full data cube and the residual.
930:
931: \begin{deluxetable}{ccccccc}
932: \tabletypesize{\normalsize}
933: \tablewidth{0pt}
934: \tablecaption{Extended AV Gas: Dynamical Properties
935: \label{tab:m83beard}}
936: \tablehead{
937: \colhead{Component} &
938: \colhead{$F$} &
939: \colhead{$\mhi$} &
940: \colhead{KE} \\
941: \colhead{} &
942: \colhead{(Jy \kps)} &
943: \colhead{(\eez{7} \msun)} &
944: \colhead{(\eez{53} erg)}
945: }
946: \startdata
947: Full residual cube & 203.0 & 97.0 & 21.0 \\
948: Inner disk & \phn17.6 & \phn8.4 & 16.0 \\
949: Inner disk, low-$v$ & \phn11.8 & \phn5.6 & \phn8.7 \\
950: \enddata
951: \end{deluxetable}
952:
953: Comparison of the slices with and without the disk emission clearly shows
954: an emission component not well modeled by differential galactic rotation.
955: This is especially evident along the major axis. The anomalous component
956: is spatially extended and clumpy, with deviation velocities generally
957: $\pm40$--50 \kps\ near the emission peak along any line of sight. The
958: emission generally appears brighter on one velocity wing of the disk \hi,
959: toward the systemic galactic velocity\footnote{Hereafter, velocities on the
960: systemic side of the disk velocity will be referred to as ``low relative
961: velocity'', while velocities on the opposite side of the disk velocity will
962: be referred to as ``high relative velocity''. This designation places the
963: velocities in the reference frame of the rotating galaxy.}. Some of the
964: anomalous emission occurs in the ``forbidden'' velocity region, which is
965: the \pv\ quadrant on the opposite side of the systemic velocity from the
966: disk. One interpretation is that material at velocities between the disk
967: and galactic systemic would be rotating more slowly than the bulk of
968: material in the galaxy. Material in the forbidden \pv\ region would be
969: moving in the opposite sense of the bulk rotation along that line of sight,
970: i.e., counter-rotating. These interpretations assume the disk and
971: anomalous material have similar spatial distribution and rotation axis
972: orientation. In a face-on galaxy like M 83, material in the forbidden
973: region is likely to be in vertical motion.
974:
975: \begin{figure*}[p]
976: \plotone{f6.eps}
977: \caption{Line-of-sight \hi\ velocity dispersion for M 83, obtained by
978: fitting a Gaussian model to the disk emission. The fit was performed along
979: every line of sight (pixel), but the values shown have been median box
980: filtered to remove poor fits. The velocity dispersion is 20 \kps\ in the
981: center, falling to less than 10 \kps\ in the outer regions.}
982: \label{fig:m83gparsig}
983: \end{figure*}
984:
985: \begin{figure*}[p]
986: \plotone{f7a.eps}
987: \caption{Position-velocity (\pv) diagrams for M 83, taken parallel to the
988: major axis at separations of 70\arcsec, as plotted on the \hi\ emission map
989: (upper right) with the slice direction indicated by the arrow. Each slice,
990: boxcar smoothed in velocity by 3 channels, is shown with the disk emission
991: included (top) and subtracted (bottom). The rotation curve is traced by
992: the white and black points. Grayscale ranges from 0 to 6 mJy/beam. Black
993: and white contours begin at $+2$-$\sigma$ and increase by factors of two.
994: Red contours begin at $-2$-$\sigma$ and decrease by factors of two.
995: Residual emission can be seen in all the slices, especially between the
996: systemic and rotational velocity along the major axis.}
997: \label{fig:m83slices}
998: \end{figure*}
999:
1000: \begin{figure*}[t]
1001: \plotone{f7b.eps}
1002: \\
1003: Figure~\ref{fig:m83slices} (continued)
1004: \end{figure*}
1005:
1006: To quantify the amount of AV gas in M 83, we constructed column density and
1007: mean velocity maps from the residual cube (see Figure \ref{fig:m83beard1}).
1008: The dynamical properties of the AV \hi\ were determined from these maps and
1009: are summarized in Table \ref{tab:m83beard}. The mass of AV \hi\ in the
1010: residual cube is 9.7\eex{8} \msun, corresponding to 20\% of the total \hi.
1011: It was apparent that poor subtraction of the \hi\ ring/warp and outer \hi\
1012: arms contaminated the residual cube, therefore we manually defined a region
1013: containing only anomalous emission projected on the central disk. The
1014: region used was an ellipse centered on the galaxy with a semi-major axis of
1015: 6.0\arcmin (radius of 8 kpc), a semi-minor axis of 4.5\arcmin, and a
1016: position angle of 45\arcdeg\, as shown in Figures \ref{fig:m83beard1}.
1017: This area contains the most dynamically homogeneous disk gas and produced
1018: reasonable model parameters. The total \hi\ mass in this region, including
1019: the cold disk, is 1.02\eex{9} \msun, of which 8.4\eex{7} \msun\ or 8.2\% is
1020: present in the anomalous component (``inner disk'' in Table
1021: \ref{tab:m83beard}). The low-relative-velocity emission contributes
1022: 5.6\eex{7} \msun\ or 5.5\% of the total inner disk \hi.
1023:
1024: The velocity field shows two high-velocity spots to the east and west of
1025: the galaxy center, identified by rectangles in Figure \ref{fig:m83beard1}.
1026: These features are also seen on the high relative velocity side of the
1027: \pv\ slices, including one structure $+1$\arcmin\ along the $+140$\arcsec\
1028: \pv\ diagram in Figure \ref{fig:m83slices}, at velocities between $+$50
1029: and $+$100 \kps. In a simple model of disk rotation, these clumps would
1030: be rotating more quickly than the bulk of the \hi. They are discrete in
1031: projection compared to the low relative velocity component described
1032: above, and will be discussed in detail in Section \ref{sect:m83:hiavcs}
1033: (as AVCs 1 and 2).
1034:
1035: Aside from these features, the motion and morphology of the residual
1036: emission mimic that of the galactic disk, with the southwest side
1037: receding and the northeast side approaching. The line-of-sight velocity
1038: dispersion ranges from 10--15 \kps, similar to the thin disk, although this
1039: may be an underestimate if the wings of the anomalous emission are excluded
1040: in the thin disk mask. A map of deviation velocity was produced by
1041: subtracting the velocity map of the modeled disk from that of the residual
1042: cube (see Figure \ref{fig:m83beard2}), and it shows a rotation rate 40--50
1043: \kps\ slower in projection. If the inclination of this disk of AV emission
1044: is also about 20\arcdeg, then its rotation rate is about 100--150 \kps,
1045: compared to 200 \kps\ for the cold disk.
1046:
1047: We used the deviation velocities to construct a map of the kinetic energy
1048: with respect to the cold disk (see Figure \ref{fig:m83beard2}). Unlike
1049: the previous analysis, this interpretation assumes all motion is
1050: perpendicular to the galaxy plane\footnote{Lacking any knowledge of
1051: transverse motion, we actually assumed motion fully along the line of
1052: sight, which is offset by 24\arcdeg\ from the galactic pole direction.
1053: Thus velocities and kinetic energies are 10\% and 20\% lower, respectively,
1054: then what would be calculated for gas with only $z$-direction motion.}, and
1055: it provides an upper limit to the energy apportioned to vertical bulk
1056: motion. The total kinetic energy so derived is 2.1\eex{54} erg, with
1057: 1.6\eex{54} erg from gas within the inner disk and 8.7\eex{53} erg of that
1058: from the low-relative-velocity material. The last figure excludes discrete
1059: clumps, such as the one projected 2\arcmin\ northwest of center; these
1060: features possess nearly half of the kinetic energy of the inner AV gas, and
1061: are fully discussed in Section \ref{sect:m83:hiavcs}. The 1.6\eex{54} erg
1062: from the inner 8 kpc is equivalent to the energy of $~ 8$ supernovae
1063: kpc$^{-2}$, although the $z$ component of velocity is not known for this
1064: material, so this is an upper limit to the kinetic energy.
1065:
1066: \begin{figure*}[p]
1067: \plottwo{f8a.eps}{f8b.eps}
1068: \caption{Maps of the AV emission in M 83, showing ({\it left\/}) column
1069: density (in units of \eez{20} \cm) and ({\it right\/}) mean velocity. The
1070: ellipse indicates the inner, well-modeled portion of the cold \hi\ disk,
1071: and the rectangles identify discreet high-relative-velocity features, as
1072: described in the text. The anomalous gas appears disk-like, rotating about
1073: 40--50 \kps\ more slowly in projection than the cold \hi\ disk.}
1074: \label{fig:m83beard1}
1075: \end{figure*}
1076:
1077: \begin{figure*}[p]
1078: \plottwo{f9a.eps}{f9b.eps}
1079: \caption{Maps of the AV emission in M 83, showing ({\it left\/}) deviation
1080: velocity and ({\it right\/}) kinetic energy per $14\arcsec \times
1081: 14\arcsec$ pixel (in units of \eez{50} erg). The contours show the mean
1082: \hi\ disk velocity in steps of 10 \kps\ for reference, and the other
1083: notations are as described in Figure \ref{fig:m83beard1}. The anomalous
1084: gas appears disk-like, rotating about 40--50 \kps\ more slowly in
1085: projection than the cold \hi\ disk, but with about the same rotation axis.
1086: Most of the kinetic energy is contained in the feature 2\arcmin\ northwest
1087: of center.}
1088: \label{fig:m83beard2}
1089: \end{figure*}
1090:
1091: While the distribution of extended gas is well-characterized by moment maps
1092: and \pv\ slices, faint, compact emission sources are easily overlooked.
1093: Detection of such systems requires a quantitative, objective searching
1094: technique that utilizes the entire set of data in a statistical way. The
1095: next section addresses our search for faint, discrete \hi\ emission
1096: sources.
1097:
1098: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1099: % SECTION -- Anomalous-Velocity HI: Discrete Emission
1100: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1101: \section{Anomalous-Velocity \hi: Discrete Emission}
1102: \label{sect:m83:hiavcs}
1103:
1104: \subsection{Searching for AVCs}
1105: \label{subsect:m83:hiavcs:clumpfind}
1106:
1107: Compared to other observational subfields of astronomy, radio synthesis
1108: imaging suffers from a lack of robust, objective source detection software.
1109: Identifying sources and measuring their fluxes and other parameters is
1110: typically done by hand from moment maps and similar data products. For
1111: bright, point-like sources, this is not a problem, but difficulties arise
1112: when one wishes to search statistically for emission of unknown scale and
1113: location that may be projected on much stronger sources.
1114:
1115: We have developed a suite of software that searches a 3-d data cube for
1116: signal in a statistical way. The software is described in detail in
1117: Appendix \ref{app:snrch} and summarized here. The data cube is smoothed
1118: with a tunable velocity filter to emphasize sources of a particular
1119: velocity width, following the method of \citet{Uson91}. Peaks in this
1120: smoothed cube are identified and used to define 3-d ``islands'' of emission
1121: by way of contouring, using a variation of the CLUMPFIND algorithm
1122: developed by \citet{WilliamsdeGeusBlitz94}. The velocity smoothing allows
1123: detection of kinematically broad but statistically significant features
1124: that might otherwise be overlooked. In addition, this algorithm allows a
1125: quantitative analysis of errors with the use of simulated datasets.
1126:
1127: To search for discrete sources, we masked the residual \hi\ cube within
1128: $\pm 20$ \kps\ of the \hi\ disk velocity to remove poorly-subtracted
1129: emission in the center of the disk profile. We ran the software with a
1130: peak threshold of 4-$\sigma$, a clump merging threshold of 2.5-$\sigma$, and
1131: a clump extending threshold of 1-$\sigma$. The width of the Gaussian
1132: filtering kernel was varied from 3--9 channels (15--45 \kps). For sources
1133: that were detected at multiple filter widths, we used the results from the
1134: filtering that produced the highest S/N (i.e., the kernel that was closest
1135: in width to the unsmoothed feature). The spatial and kinematic morphology
1136: of the candidate clouds was analyzed interactively using a variety of
1137: tools, including \pv\ plots, channel maps and 3-d visual rendering tools.
1138: A two-component Gaussian fit along the velocity axis at the clump centroid
1139: recovered the velocity and line width of the clump, as well as the line
1140: width of any \hi\ disk emission along that line of sight. Finally, the
1141: flux of each clump was verified by hand using the standard radio method of
1142: aperture photometry. Values were consistent with the results from our
1143: automated software.
1144:
1145: \subsection{Detected Systems}
1146: \label{subsect:m83:hiavcs:avcs}
1147:
1148: We discovered 14 discrete clouds of \hi\ emission at anomalous velocities
1149: or positions in the M 83 data cube. These anomalous-velocity clumps
1150: (AVCs) are distinct from the bulk of the \hi\ disk emission in that their
1151: contours of emission detach from the disk at levels approaching the map
1152: noise. The AVC properties are given in Tables \ref{tab:m83avcs1} and
1153: \ref{tab:m83avcs2}. The locations of the AVCs are shown superposed on the
1154: \hi\ column density map in Figure \ref{fig:m83mom0avcs} and superposed on
1155: the velocity map in Figure \ref{fig:m83mom1avcs}.
1156:
1157: \begin{deluxetable*}{ccccccccc}
1158: \tabletypesize{\scriptsize}
1159: \tablewidth{0pt}
1160: \tablecaption{Discrete AVCs: General Properties
1161: \label{tab:m83avcs1}}
1162: \tablehead{
1163: \colhead{AVC\tablenotemark{a}} &
1164: \colhead{RA} &
1165: \colhead{Dec} &
1166: \colhead{$v_{\rm AVC}$\tablenotemark{b}} &
1167: \colhead{$\Omega$\tablenotemark{c}} &
1168: \multicolumn{2}{c}{$l$\tablenotemark{d}} &
1169: \multicolumn{2}{c}{$r_{proj}$\tablenotemark{e}} \\
1170: \colhead{} &
1171: \colhead{(J2000)} &
1172: \colhead{(J2000)} &
1173: \colhead{(km/s)} &
1174: \colhead{(sq.~arcmin)} &
1175: \colhead{(\arcmin)} &
1176: \colhead{(kpc)} &
1177: \colhead{(\arcmin)} &
1178: \colhead{(kpc)}
1179: }
1180: \startdata
1181: 1 & 13 36 49.2 & -29 51 13 & 595$\pm\phn2$ & 2.63 & 5.3 & 6.9 & \phn1.7 & \phn2.2 \\
1182: 2 & 13 37 19.3 & -29 51 13 & 398$\pm\phn2$ & 2.75 & 5.5 & 7.2 & \phn3.7 & \phn4.8 \\
1183: 3 & 13 36 42.7 & -29 57 31 & 667$\pm\phn2$ & 0.77 & 2.8 & 3.6 & \phn7.1 & \phn9.3 \\
1184: 4 & 13 37 18.3 & -29 49 35 & 547$\pm\phn2$ & 0.58 & 2.5 & 3.3 & \phn4.4 & \phn5.7 \\
1185: 5 & 13 37 12.9 & -30 04 59 & 512$\pm\phn2$ & 0.39 & 1.7 & 2.2 & 13.4 & 17.4 \\
1186: 6 & 13 37 08.6 & -30 06 23 & 498$\pm\phn3$ & 0.39 & 1.8 & 2.3 & 14.6 & 19.0 \\
1187: 7 & 13 37 06.4 & -30 02 11 & 411$\pm\phn2$ & 0.39 & 1.5 & 2.0 & 10.0 & 13.0 \\
1188: 8 & 13 36 56.8 & -29 43 31 & 361$\pm\phn2$ & 0.39 & 1.9 & 2.5 & \phn8.4 & 11.0 \\
1189: \hline
1190: 9 & 13 37 33.4 & -30 10 35 & 690$\pm\phn8$ & 0.39 & 1.3 & 1.7 & 19.9 & 25.8 \\
1191: 10 & 13 38 11.8 & -29 27 38 & 593$\pm\phn5$ & 0.39 & 1.2 & 1.6 & 28.6 & 37.2 \\
1192: 11 & 13 37 21.6 & -30 15 15 & 407$\pm12$ & 0.39 & 1.6 & 2.1 & 23.9 & 31.1 \\
1193: 12 & 13 35 45.7 & -29 50 58 & 591$\pm\phn5$ & 0.39 & 1.0 & 1.3 & 16.5 & 21.5 \\
1194: 13 & 13 37 25.9 & -30 17 35 & 405$\pm\phn5$ & 0.39 & 1.4 & 1.8 & 25.9 & 33.7 \\
1195: 14 & 13 35 19.0 & -29 37 39 & 633$\pm32$ & 0.39 & 1.2 & 1.6 & 26.3 & 34.2 \\
1196: \enddata
1197: \tablenotetext{a}{AVCs 1--8 are considered real detections, AVCs 9-14 are
1198: likely spurious.}
1199: \tablenotetext{b}{Heliocentric radial velocity of the peak of the AVC
1200: emission.}
1201: \tablenotetext{c}{Solid angle subtended by the AVC after collapsing its
1202: HI emission along the velocity axis. A value of 0.39 sq.~arcmin indicates
1203: the feature is unresolved.}
1204: \tablenotetext{d}{Greatest linear extent of the AVC, uncorrected for
1205: broadening by the synthesized beamwidth.}
1206: \tablenotetext{e}{Projected distance from the galactic center.}
1207: \end{deluxetable*}
1208:
1209: \begin{deluxetable*}{cccccccc}
1210: \tabletypesize{\scriptsize}
1211: \tablewidth{0pt}
1212: \tablecaption{Discrete AVCs: Dynamical Properties
1213: \label{tab:m83avcs2}}
1214: \tablehead{
1215: \colhead{AVC\tablenotemark{a}} &
1216: \colhead{$F_{\rm meas}$\tablenotemark{b}} &
1217: \colhead{$F_{\rm corr}$\tablenotemark{c}} &
1218: \colhead{$M_{\rm HI}$\tablenotemark{d}} &
1219: \colhead{$N_{\rm HI}$\tablenotemark{e}} &
1220: \colhead{$\vdev$\tablenotemark{f}} &
1221: \colhead{$\Delta v$\tablenotemark{g}} &
1222: \colhead{$KE$\tablenotemark{h}} \\
1223: \colhead{} &
1224: \colhead{(Jy km/s)} &
1225: \colhead{(Jy km/s)} &
1226: \colhead{(\eez{6} \msun)} &
1227: \colhead{(\eez{20} \cm)} &
1228: \colhead{(km/s)} &
1229: \colhead{(km/s)} &
1230: \colhead{(\eez{53} erg)}
1231: }
1232: \startdata
1233: 1 & 3.08 $\pm$0.07 & 3.12 $\pm$0.08 & 14.9 $\pm$0.4 & \phn2.9 & \phn$+$51 & \phn29$\pm2$ & 3.9 $\pm$0.1 \\[3pt]
1234: 2 & 1.42 $\pm$0.06 & 1.48 $\pm$0.06 & \phn7.1 $\pm$0.3 & \phn1.4 & \phn$-$41 & \phn28$\pm2$ & 1.2 $\pm$0.0 \\[3pt]
1235: 3 & 0.99 $\pm$0.05 & 1.12 $\pm$0.05 & \phn5.4 $\pm$0.3 & \phn2.6 & \phn$+$74 & \phn56$\pm4$ & 2.9 $\pm$0.1 \\[3pt]
1236: 4 & 0.23 $\pm$0.03 & 0.24 $\pm$0.03 & \phn1.2 $\pm$0.1 & \phn0.6 & $+$113 & \phn32$\pm2$ & 1.5 $\pm$0.2 \\[3pt]
1237: 5 & 0.11 $\pm$0.02 & 0.18 $\pm$0.04 & \phn0.8 $\pm$0.2 & \phn1.0 & \phn$-$77 & \phn15$\pm2$ & 0.5 $\pm$0.1 \\[3pt]
1238: 6 & 0.11 $\pm$0.02 & 0.21 $\pm$0.04 & \phn1.0 $\pm$0.2 & \phn1.1 & $-$103 & \phn15$\pm2$ & 1.0 $\pm$0.2 \\[3pt]
1239: 7 & 0.13 $\pm$0.02 & 0.17 $\pm$0.03 & \phn0.8 $\pm$0.2 & \phn1.2 & $-$166 & \phn44$\pm3$ & 2.3 $\pm$0.4 \\[3pt]
1240: 8 & 0.11 $\pm$0.02 & 0.13 $\pm$0.03 & \phn0.6 $\pm$0.1 & \phn0.7 & \phn$-$90 & \phn24$\pm2$ & 0.5 $\pm$0.1 \\[3pt]
1241: \hline
1242: 9 & 0.08 $\pm$0.02 & 0.24 $\pm$0.06 & \phn1.2 $\pm$0.3 & \phn2.3 & \phn\nodata & \phn30$\pm2$ & \nodata \\[3pt]
1243: 10 & 0.14 $\pm$0.03 & 2.77 $\pm$0.52 & 13.2 $\pm$2.5 & 29.1 & \phn\nodata & \phn24$\pm2$ & \nodata \\[3pt]
1244: 11 & 0.19 $\pm$0.03 & 1.09 $\pm$0.17 & \phn5.2 $\pm$0.8 & \phn9.2 & \phn\nodata & \phn82$\pm5$ & \nodata \\[3pt]
1245: 12 & 0.06 $\pm$0.02 & 0.14 $\pm$0.04 & \phn0.7 $\pm$0.2 & \phn1.3 & $+$135 & \phn29$\pm2$ & 1.2 $\pm$0.3 \\[3pt]
1246: 13 & 0.16 $\pm$0.03 & 1.51 $\pm$0.27 & \phn7.2 $\pm$1.3 & 13.5 & \phn\nodata & \phn20$\pm2$ & \nodata \\[3pt]
1247: 14 & 0.15 $\pm$0.03 & 1.68 $\pm$0.31 & \phn8.0 $\pm$1.5 & 17.4 & \phn\nodata & 164$\pm12$ & \nodata \\[3pt]
1248: \enddata
1249: \tablenotetext{a}{AVCs 1--8 are considered real detections, AVCs 9-14 are
1250: likely spurious.}
1251: \tablenotetext{b}{Total flux in the AVC, uncorrected for the primary beam
1252: attenuation. Errors are 1-$\sigma$.}
1253: \tablenotetext{c}{Total flux in the AVC, corrected for the primary beam
1254: attenuation.}
1255: \tablenotetext{d}{Mass of HI contained in the AVC.}
1256: \tablenotetext{e}{Peak column density of HI, integrated across the
1257: velocity width of the emission.}
1258: \tablenotetext{f}{Deviation velocity of the AVC, defined as the
1259: difference between the radial velocity of the peak AVC emission and the
1260: velocity fitted to the HI disk (if detectable) at that location, as
1261: described in Section \ref{subsect:m83:hiavcs:clumpfind}.}
1262: \tablenotetext{g}{Full width at half maximum of a Gaussian fit to the
1263: velocity profile.}
1264: \tablenotetext{h}{Kinetic energy of the AVC found from the deviation
1265: velocity and mass. This is a lower limit to the total kinetic energy as
1266: only radial velocities are used.}
1267: \end{deluxetable*}
1268:
1269: \begin{figure*}
1270: \plotone{f10.eps}
1271: \caption{The \hi\ column density map for M 83, showing the locations of the
1272: detected AVCs.
1273: The shapes indicate the projected spatial extent of each AVC
1274: out to 1-$\sigma$ in \hi\ column density (5\eex{18} \cm).
1275: The large gray circles indicate the primary beam
1276: sensitivity. The column density is in units of \eez{20} \cm.}
1277: \label{fig:m83mom0avcs}
1278: \end{figure*}
1279:
1280: \begin{figure*}
1281: \plotone{f11.eps}
1282: \caption[M 83 AVCs plotted on the \hi\ velocity map]
1283: {The \hi\ intensity-weighted mean velocity map for M 83, showing the
1284: locations of the detected AVCs. Each square is centered at the centroid of
1285: the AVC, with the color indicating its heliocentric radial velocity.}
1286: \label{fig:m83mom1avcs}
1287: \end{figure*}
1288:
1289: \begin{figure*}
1290: \plotone{f12a.eps}
1291: \caption{Position-velocity slices taken along the E-W direction and
1292: velocity profiles are shown for each AVC, which is centered at the
1293: intersection of the solid lines in the \pv\ plot. Grayscale ranges from 0
1294: to 6 mJy/beam. Black and white contours begin at 2-$\sigma$
1295: ($\sigma = 0.94$ mJy/beam) and increase by factors of two. Red contours
1296: begin at $-2$-$\sigma$ and decrease by factors of two. The profile plot,
1297: taken at the centroid of the AVC total \hi\ emission, shows the Gaussian
1298: velocity fit to the AVC and \hi\ disk (if present). AVCs which show up in
1299: multiple slices are identified by small numerals.}
1300: \label{fig:m83avcs}
1301: \end{figure*}
1302:
1303: \begin{figure*}
1304: \plotone{f12b.eps}
1305: \\
1306: Figure~\ref{fig:m83avcs} (continued)
1307: \end{figure*}
1308:
1309: \begin{figure}
1310: %\centering{\includegraphics[width=5in]{f13.eps}}
1311: \plotone{f13.eps}
1312: \caption{Integrated \hi\ intensity maps for the four spatially-extended
1313: AVCs. AVCs 1 and 2 have bright cores and low-level extended features,
1314: while AVCs 3 and 4 are more compact and only mildly resolved. Grayscale
1315: ranges from 0--0.25 Jy beam$^{-1}$ \kps, and the contours are spaced at
1316: 0.025 Jy beam$^{-1}$ \kps. The ellipse in the AVC 3 plot shows the HPBW of
1317: the synthesized beam. Coordinates are with respect to the center of M 83,
1318: with north up and east to the left.}
1319: \label{fig:m83avcmoms}
1320: \end{figure}
1321:
1322: The AVCs are found across the \hi\ disk of M 83, but they do exhibit a
1323: small degree of clustering, and several are located in regions where the
1324: kinematics of the disk are most complicated. Three clumps (1, 2 and 4) are
1325: located within the kinematically ``normal'' inner disk. One clump (3) is
1326: found at the inner disk/warp interface, two additional sources (7 and
1327: 8) are projected just beyond the ring, and four clumps (5, 6, 9, and 12)
1328: are found projected near outer \hi\ arms. The remaining four AVCs (10, 11,
1329: 13, 14) are located well away from the center of the galaxy, at projected
1330: radii where the primary beam sensitivity falls to 10\% or lower.
1331:
1332: Gaussian fits and \pv\ plots were produced as described in the previous
1333: section, and these are shown for each AVC in Figure \ref{fig:m83avcs}. The
1334: integrated \hi\ intensity maps for the spatially-extended AVCs are shown in
1335: Figure \ref{fig:m83avcmoms}. To eliminate the confusion of
1336: spatially-varying noise, the \pv\ slices and intensity maps have not been
1337: corrected for the primary beam attenuation. The fluxes, masses, and column
1338: densities reported in Table \ref{tab:m83avcs2} have been corrected for this
1339: effect.
1340:
1341: A number of the AVCs have fluxes or velocity structure in the \pv\ plots
1342: that suggest they are spurious. From a ``by-eye'' analysis, we estimate
1343: that eight of these features (AVCs 1--8) are real detections and the
1344: remainder are not. Below we describe specific properties of individual
1345: AVCs, grouping them according to similar distribution and attributes.
1346: For completeness, we include features thought to be spurious, although
1347: these are excluded in our discussion in Section \ref{sect:m83:disc}.
1348:
1349: \noindent
1350: {\bf AVC 1}
1351:
1352: \noindent
1353: This source is the brightest and most massive AVC detected, with
1354: $M = 1.5\eex{7}$ \msun. Centered 2 kpc from the dynamical center of M 83,
1355: AVC 1 is also the only discrete clump projected over the inner \hi\ disk.
1356: The intensity map and \pv\ plots show that it is spatially resolved and
1357: extended in velocity, with a deviation velocity $\vdev = 51$ \kps\ and line
1358: width of 29 \kps\ FWHM. The measured line width is likely only accurate
1359: near the emission peak, as the \pv\ slices display a connection in velocity
1360: to the underlying disk emission. The projected intensity map supports this
1361: idea, with a secondary maximum projecting to the south, where the cold disk
1362: approaches the velocity of the cloud. The characteristics of this AVC are
1363: consistent with material flowing into or out of the disk, although the lack
1364: of spatial information along the line of sight complicates conclusions
1365: about the source's $z$ height.
1366:
1367: The moment and dynamical maps described in Section
1368: \ref{subsect:m83:hiavext:beard} clearly show emission from this feature
1369: (see Figures \ref{fig:m83beard1} and \ref{fig:m83beard2}). The peak of the
1370: emission can be seen 1.5\arcmin\ west and 1.5\arcmin\ north of center in
1371: Figure \ref{fig:m83beard1}a.
1372:
1373: \noindent
1374: {\bf AVCs 2, 4}
1375:
1376: \noindent
1377: These two clumps are close together in projection at a radius of about 5
1378: kpc (near the edge of the inner disk), yet they differ greatly in their
1379: other characteristics. AVC 2 is bright, massive ($M = 7\eex{6}$ \msun) and
1380: extended ($\Omega = 2.8$ sq.~arcmin, $l = 7$ kpc), while AVC 4 is less
1381: massive ($M = 1\eex{6}$ \msun) and barely resolved (largest linear extent
1382: $l = 3.3$ kpc). In addition, the deviation velocities of these clumps are
1383: quite different, with AVC 2 at high relative velocity ($\vdev = -41$ \kps)
1384: and AVC 4 at low relative velocity ($\vdev = 113$ \kps). The line widths
1385: are similar,
1386: but the emission of AVC 2 appears to blend with the disk velocity.
1387: Emission from AVC 4 also blends kinematically with the disk, but the peak
1388: is more clearly separated, as can be seen in the velocity profiles in
1389: Figure \ref{fig:m83avcs}.
1390: The heliocentric velocity
1391: of AVC 4 (522 \kps) places it in the ``forbidden'' region of
1392: counter-rotating material in the \pv\ plane. The spatial coincidence of
1393: these clumps hints at a correlation between them, possibly in the form of
1394: an expanding bubble \citep[e.g.,][]{Kamphuisetal91}, but it is difficult
1395: to reconcile their differing morphological and kinematic structure. As
1396: with AVC 1, these two clumps appear in the AV gas maps shown in Figures
1397: \ref{fig:m83beard1} and \ref{fig:m83beard2}.
1398:
1399: \noindent
1400: {\bf AVC 3}
1401:
1402: \noindent
1403: At a projected distance of 9 kpc from the galactic center, this AVC appears
1404: in a region where the \hi\ ring is disturbed and has a relatively low
1405: column density of $\nh = 3\eex{20}$ \cm. The clump is extended in space,
1406: covering two synthesized beams in area and spanning 3.6 kpc. It is also
1407: extended in velocity, with a line width of 56 \kps\ FWHM, and its emission
1408: appears to merge with the disk \hi\ at lower velocities. The source is
1409: massive ($M = 5.4\eex{6}$ \msun) and has a large kinetic energy of
1410: 3\eex{53} erg.
1411:
1412: Although the clump is projected on a region of changing inclination between
1413: the inner disk and outer arms, it does not appear to result from
1414: beam-smearing of a warp discontinuity. First, it is a compact structure,
1415: whereas a warp feature would be extended like the \hi\ ring or outer arms.
1416: Second, it is much broader in velocity than the kinematically cold \hi\
1417: elsewhere in the galaxy. Finally, the kinematics of the surrounding disk
1418: material are regular in comparison to that seen at warp interfaces in other
1419: regions of the galaxy. This cloud is the only such feature seen along the
1420: \hi\ ring.
1421:
1422: This AVC appears at a location of low disk \nh. The \pv\ plots and Figure
1423: \ref{fig:m83mom0avcs} show a lack of emission in the disk a few arcmin to
1424: the east of this location, although \hi\ disk emission is present at the
1425: projected location of AVC 3.
1426:
1427: \noindent
1428: {\bf AVCs 5, 6}
1429:
1430: \noindent
1431: These two AVCs have similar mass, size, line width, and deviation velocity.
1432: At a projected radius of about 14\arcmin\ (18 kpc) to the south of the
1433: galactic center, AVCs 5 and 6 are separated by 2\arcmin\ (2.6 kpc) in
1434: projected space and 20 \kps\ in velocity. AVC 6 is the more massive and
1435: kinematically anomalous clump, with a mass of 1\eex{6} \msun\ and deviation
1436: velocity of $-103$ \kps. AVC 5 has a mass of 8\eex{5} \msun\ and deviation
1437: velocity of $-77$ \kps. The line width of both clumps is 15 \kps,
1438: producing a velocity dispersion of 6 \kps. Unlike the previously discussed
1439: AVCs, the emission from these \hi\ clumps does not merge with that of the
1440: disk in position or velocity.
1441:
1442: In addition to similar kinematics, these two clouds have similar sizes and
1443: are spatially unresolved by the VLA synthesized beam. The FWHM of a
1444: Gaussian fit across either feature is 45\arcsec; when deconvolved from the
1445: $\sim 35\arcsec$ synthesized beam, this implies a maximum source size of
1446: 30\arcsec\ ($\sim 0.7$ kpc). Their similar characteristics and location
1447: hint at a possible relation, and they might be bright spots in a single
1448: \hi\ feature, although this speculation is made less likely by the 20 \kps\
1449: difference in deviation velocity and the small radial velocity dispersions.
1450:
1451: \noindent
1452: {\bf AVCs 7, 8}
1453:
1454: \noindent
1455: These emission sources are similar in that they appear between the \hi\
1456: ring and an outer arm, although they are projected on opposite sides of
1457: center. Both have highly negative deviation velocities (\vdev\ = $-166$
1458: and $-90$ \kps), are spatially unresolved, and have similar fluxes and
1459: masses ($M = 8$ and 6\eex{5} \msun). Clump 7 has a large fraction of the
1460: kinetic energy of the AVC ensemble, due primarily to its large deviation
1461: velocity. From the integrated \hi\ maps, which have the same colormap
1462: scale, one can see that AVC 7 is more sharply peaked than AVC 8.
1463: Consideration of the \pv\ plots and velocity profiles shows that the line
1464: of sight to AVC 7 contains emission at a variety of other velocities (e.g.,
1465: $-100$, $-40$, +40 \kps). The clumps producing this emission fall below
1466: our threshold for clump identification, and it is unclear whether they
1467: result from real velocity structure in the AVC or from a systematic problem
1468: such as poor continuum subtraction or bandpass calibration.
1469:
1470: \noindent
1471: {\bf AVCs 9, 12}
1472:
1473: \noindent
1474: Two clumps lie superposed on or near \hi\ arms outside the half power width
1475: of the primary beam. Both are spatially unresolved and have similar line
1476: widths of 30 \kps\ FWHM. At these radii, the primary beam attenuation
1477: correction factor is about 3, so although both have small measured fluxes,
1478: the corrected masses are significant (1.1 and 0.6\eex{6} \msun). Both AVCs
1479: have high deviation velocities, with $\vdev = +135$ \kps\ for clump 12, and
1480: a velocity difference of $+100$ \kps\ for clump 9 measured from the \hi\
1481: arm projected nearby. These detections are the least statistically
1482: significant of the set (4-$\sigma$ and 3-$\sigma$ for AVC 9 and 12,
1483: respectively), and they are hereafter treated as spurious.
1484:
1485: \noindent
1486: {\bf AVCs 10, 11, 13, 14}
1487:
1488: \noindent
1489: All four of these clumps are in low-sensitivity regions, where the
1490: attenuation correction factor approaches 10. AVCs 11 and 13 are projected
1491: close together in space (2\arcmin\ apart) and velocity (2 \kps\ apart). All
1492: four have strongly peaked emission maps, and their lines of sight contain
1493: emission at velocities throughout the data cube. In addition, AVCs 11 and
1494: 14 have very large line widths compared to the other AVCs. It is likely
1495: that these AVCs are not real, but are the result of systematic effects such
1496: as calibration errors. If they are real, the beam correction produces
1497: large masses and kinetic energies for these clumps.
1498:
1499: \subsection{Optical Counterparts to the AVCs}
1500:
1501: Searching for optical counterparts to the \hi\ AVCs was carried out by
1502: hand. Each AVC was overlaid on the unmasked optical image, which was
1503: inspected for optical emission (see Figure \ref{fig:m83optavcs}). To
1504: search for small-scale emission structure superposed on the extended galaxy
1505: profile, the masked optical image was median filtered with a $5\times5$
1506: pixel box. The location of each AVC was fit with a plane, excluding the
1507: four AVCs within the inner disk, which is difficult to fit with a simple
1508: model. The fit residuals were inspected for signal above the noise, which
1509: was not found in any case. The noise in the residuals was used to
1510: estimate surface brightness upper limits for optical counterparts. The
1511: results are shown in Table \ref{tab:m83optavcs}.
1512:
1513: \begin{figure*}
1514: \plotone{f14.eps}
1515: \caption{The optical M 83 image overlaid with the locations of the \hi\
1516: AVCs.}
1517: \label{fig:m83optavcs}
1518: \end{figure*}
1519:
1520: \begin{deluxetable}{cccc}
1521: \tabletypesize{\scriptsize}
1522: \tablewidth{0pt}
1523: \tablecaption{Optical Counterparts to HI AVCs
1524: \label{tab:m83optavcs}}
1525: \tablehead{
1526: \colhead{AVC\tablenotemark{a}} &
1527: \colhead{$\mu_{\rsloan,0}$\tablenotemark{b}} &
1528: \colhead{$\mu_{\rsloan,{\rm lim}}$\tablenotemark{c}} &
1529: \colhead{Notes\tablenotemark{d}}
1530: }
1531: \startdata
1532: 1 & 20.4 & \nodata & inner disk, spiral arm \\
1533: 2 & 20.7 & \nodata & inner disk, spiral arm \\
1534: 3 & 24.2 & \nodata & inner disk \\
1535: 4 & 22.2 & \nodata & inner disk \\
1536: 5 & 27.8 & 27.2 & stars \\
1537: 6 & 27.9 & 27.4 & stars \\
1538: 7 & 25.4 & 27.1 & stars \\
1539: 8 & 24.4 & 26.4 & outer disk, stars \\
1540: \hline
1541: 9 & 26.5 & 27.0 & outer disk, stars \\
1542: 10 & 26.8 & 27.3 & stars, galaxy 1.5\arcmin\ SE \\
1543: 11 & 27.2 & 26.8 & stars \\
1544: 12 & 27.5 & 26.8 & bright coincident stars \\
1545: 13 & 28.1 & 27.4 & stars \\
1546: 14 & 27.7 & 26.9 & stars, galaxy \\
1547: \enddata
1548: \tablenotetext{a}{AVCs 1--8 are considered real detections, AVCs 9-14 are
1549: likely spurious.}
1550: \tablenotetext{b}{The median surface brightness, in \rsloan\ magnitudes per
1551: square arcsec, at the location of the AVC. For extended AVCs, the median
1552: was taken within the HPBW centered on the peak HI emission.}
1553: \tablenotetext{c}{The 3-$\sigma$ surface brightness limit on any optical
1554: features coincident with the AVC. This was determined by fitting and
1555: subtracting a plane from the optical image at that point. See that text
1556: for details.}
1557: \tablenotetext{d}{The notes identify optical features coincident with the
1558: AVC, including foreground stars, extended emission from M 83 (designated by
1559: location in the disk), and background galaxies.}
1560: \end{deluxetable}
1561:
1562: Below we describe possible counterparts and limits to the optical surface
1563: brightness for the individual AVCs. We discuss these limits in the context
1564: of other deep optical studies in Section \ref{sect:m83:disc:accret}.
1565:
1566: \noindent
1567: {\bf AVCs 1, 2, 3, 4, 7, 8}
1568:
1569: \noindent
1570: These AVCS are all projected over the bright optical disk. The bright cores
1571: of AVCs 1 and 2 are projected on spiral arms, appearing to trace the shape
1572: of the arms, although this is possibly due to the method used to detect and
1573: delineate the \hi\ clumps. The optical light coincident with AVC 1 peaks
1574: at 20.4 \rsloan\ mag per square arcsec, excluding the bright star clusters
1575: in the spiral arms, and falls off to 22.1 \rsloan\ mag per square arcsec in
1576: the outer part of the clump. Likewise, the core of AVC 2 is projected on a
1577: bright spiral arm, with $\mu_{\rsloan}$ = 20.7 \rsloan\ mag per
1578: square arcsec. The optical light falls to 24.5 \rsloan\ mag per square
1579: arcsec in the outer regions of this extended \hi\ clump. While these AVCs
1580: align with spiral arms that contain star clusters, the spatial resolution
1581: of the \hi\ data is insufficient to draw direct comparisons between AVC and
1582: star cluster morphology.
1583:
1584: The other two AVCs within the inner disk, AVCs 3 and 4, have no obvious
1585: small-scale optical counterparts aside from stars. The inner edge of AVC 4
1586: coincides with the outer edge of a spiral arm, although this is again
1587: likely due to the searching and masking technique we have employed. AVC 3
1588: is projected on a disruption in the \hi\ disk, but no evidence of such a
1589: feature is seen in the optical data.
1590:
1591: Clumps 7 and 8 are projected further out, over the south and north edge of
1592: the extended stellar envelope. The underlying surface brightnesses in
1593: these regions is 25.4 and 24.4 \rsloan\ mag per square arcsec,
1594: respectively. There are no small-scale optical features, aside from stars,
1595: down to the detection limit of about 27 \rsloan\ mag per square arcsec.
1596:
1597: \noindent
1598: {\bf AVCs 5, 6, 9, 11, 13}
1599:
1600: \noindent
1601: The majority of the large-radius AVCs contain only stars within the lowest
1602: \hi\ contour. Clumps 5 and 6 are isolated from bright stars, but AVCs 9,
1603: 11 and 13 lie in the wings of very bright stars. As a result, much of the
1604: light in these regions is masked out to determine the surface brightness, and
1605: the amount of flux below a limiting magnitude of about 26 \rsloan\ mag per
1606: square arcsec is poorly constrained. There is no evidence for emission
1607: down to this level, however.
1608:
1609: \noindent
1610: {\bf AVCs 10, 12, 14}
1611:
1612: \noindent
1613: These three AVCs are positioned on or near optical features that could be
1614: related. AVC 10 is projected 1.5\arcmin\ northwest of an object with
1615: a galactic light profile. The contours of AVC 14 contain a similar object.
1616: Both objects are smaller than the extent of the AVCs, although
1617: the \hi\ clumps are unresolved. Given their apparent size, it is likely
1618: that these optical sources are background galaxies, although it is
1619: impossible to verify without spectroscopic data.
1620:
1621: Clump 12 is centered on the location of two bright stars, making the
1622: limiting magnitude at this location difficult to determine. There is no
1623: signal at this position in the 1.4 GHz continuum map, therefore it is
1624: unlikely that the detected \hi\ results from poor continuum subtraction.
1625: It is more likely that this is a chance superposition between bright stars
1626: and a detected \hi\ source (real or spurious).
1627:
1628: \subsection{AVC Detection Simulations}
1629: \label{subsect:m83:hiavcs:sim}
1630:
1631: To analyze our results quantitatively, it is necessary to determine
1632: significance of the detections in terms of the completeness and false
1633: detection rate. Interferometry maps differ from images made with
1634: non-interferometric devices in that neighboring spatial pixels are not
1635: independent. The transformation of limited samples from the $uv$ plane to
1636: the image plane results in the signal from every location being spatially
1637: smeared. Resolution elements therefore overlap each other, and more
1638: importantly, noise features are extended and often similar in size to the
1639: objects being searched for. Prior to primary beam correction, the noise is
1640: nearly constant across the field. It is upon this background of
1641: spatially-dependent fluctuations that one searches for discrete emission
1642: features.
1643:
1644: To investigate the sensitivity of our data, we simulated 2000 observations
1645: of Gaussian noise in the $uv$ plane, using the same $uv$ distribution and
1646: noise characteristics as our filtered M 83 observations. Each dataset was
1647: transformed to the image plane using the same technique described in
1648: Section \ref{sect:vlaobs}, so that the resultant image cube had dimensions,
1649: spatial coherence (i.e., synthesized beamwidth) and $\sigma$ identical to
1650: that of the cleaned, binned M 83 data cube. Each simulated cube was
1651: processed with our detection software, using identical parameters as
1652: described in Section \ref{subsect:m83:hiavcs:clumpfind}, except for a
1653: single filtering kernel of 5 channels (25 \kps) FWHM, the average linewidth
1654: of the AVCs. The histogram of clump flux values, scaled to the parameter
1655: space volume of a single masked M 83 data cube, is shown in Figure
1656: \ref{fig:m83simfalse}. The low-flux side of the histogram shows a rapid
1657: fall-off due to decreasing detection efficiency. The histogram peaks near
1658: 0.045 Jy \kps, which is about 4.5 times the 1-$\sigma$ flux for a single
1659: beam in the smoothed cube (see Appendix \ref{app:snrch}). This is
1660: consistent with our detection threshold of 4-$\sigma$.
1661:
1662: We can use the simulated distribution to calculate the likelihood of
1663: obtaining at least one spurious detection above a given flux threshold in
1664: the data cube. If we assume the false detection rate is determined by
1665: Poisson statistics, we find that the likelihood of obtaining at least one
1666: detection above flux $F$ is given by
1667: $p_F = 1 - e^{-\mu_F} \mu_F^{\nu}/\nu! = 1 - e^{-\mu_F}$ for $\nu = 0$.
1668: The mean expected number of detections $\mu$ is given by the integral of
1669: the empirical distribution between $F$ and $+\infty$, and it is a
1670: straightforward matter to find $F$ for various values of $p_F$. The flux
1671: limits for $p_F$ = 0.32, 0.05, and 0.01 are plotted in Figure
1672: \ref{fig:m83simfalse}. Several of our AVCs have fluxes (uncorrected for
1673: attenuation) within the region of non-zero false detection rate. If we
1674: exclude the three significant clumps with uncorrected fluxes near or above
1675: 1.0 Jy \kps\ (AVCs 1--3), we are left with 11 clumps in the 0.06--0.23 Jy
1676: \kps\ flux range. These are indicated in Figure \ref{fig:m83simfalse}.
1677:
1678: \begin{figure}
1679: \plotone{f15.eps}
1680: \caption {Flux distribution of false detections in the simulated pure noise
1681: data cubes. The binned values have been scaled to equal the mean number of
1682: false detections per masked M 83 data cube per 0.01 Jy \kps\ flux bin, so
1683: that the integral of this distribution equals the mean expected number of
1684: false detections in our data cube, a value of 4.1. The dotted lines note
1685: the likelihood $p_F$ of obtaining at least one false detection above the
1686: given flux value in a single data cube (see text for further explanation).
1687: Open circles indicate the fluxes of detected clumps in the M 83 data, with
1688: the three brightest clumps (AVCs 1--3) excluded. For clump flux $>$ 0.1 Jy
1689: \kps, there is an excess over purely spurious detections.}
1690: \label{fig:m83simfalse}
1691: \end{figure}
1692:
1693: \begin{figure}
1694: \plotone{f16.eps}
1695: \caption{Detection efficiency of the simulated clumps. The efficiency is
1696: 90\% above 0.15 Jy \kps.}
1697: \label{fig:m83simeffic}
1698: \end{figure}
1699:
1700: Integrating the full empirical distribution leads to an expected mean of
1701: 4.1 false detections per masked data cube. With an algorithm flux
1702: detection threshold of 4-$\sigma$, and for 177,500 independent samples in
1703: our masked data cube, we expect 5.3 detections of greater than 4-$\sigma$
1704: significance from purely statistical considerations. That the simulations
1705: average fewer false detections indicate limitations of our detection
1706: algorithm. Both false detection rates are consistent with our ``by-eye''
1707: analysis indicating that AVCs 9--14 are spurious. The chance of obtaining
1708: at least 6 spurious detections given the simulation mean of 4.1 is 22\%;
1709: given the statistical mean of 5.3, the chance is 44\%.
1710:
1711: We note that the simulations do not account for systematic errors which may
1712: occur from bandpass calibration or continuum subtraction and which may
1713: introduce artifacts similar to clumps of emission. Some of our
1714: ``detections'' show indications of this effect (e.g., the multiple velocity
1715: components of AVCs 11 and 14). Excluding these, the number of false
1716: detections is still consistent with the expected rate. Modeling of
1717: systematic calibration errors will be included in future versions of the
1718: simulation software.
1719:
1720: The completeness of our sample was determined from the detection
1721: efficiency, which in turn was found from additional simulations. Twenty
1722: data cubes with noise characteristics similar to the data were created in
1723: the manner described above. To each of them were added 200 3-d Gaussian
1724: sources with spatial FWHM = 40\arcsec\ (approximately the synthesized
1725: beamwidth), velocity FWHM = 5 channels (25 \kps), and peak brightness
1726: varying between 1.0 and 5.0 mJy/beam in steps of 0.5 mJy/beam. The
1727: integrated fluxes of the sources varied between 0.01 and 0.27 Jy \kps. The
1728: positions of the sources were produced semi-randomly so that the centers
1729: were not necessarily at integer pixels but the sources did not overlap
1730: within 5-$\sigma$ in any direction. The resulting cubes were processed
1731: with our detection software in an identical fashion as the source data.
1732:
1733: At least 90\% of the simulated sources were detected to a flux of 0.16 Jy
1734: \kps, as is shown in Figure \ref{fig:m83simeffic}. Below this, the
1735: detection efficiency falls off rapidly, reaching 50\% at a flux of about
1736: 0.1 Jy \kps. The shape of the efficiency curve may explain the lack of
1737: low-flux detections in our sample. Also from these simulations, we note a
1738: systematic flux overestimate of $\sim$ 10\% for a simulated 0.2 Jy \kps\
1739: clump and $\sim$ 30\% for 0.1 Jy \kps. The measured AVC fluxes have not
1740: been corrected for this effect.
1741:
1742: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1743: % SECTION -- Discussion
1744: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1745: \section{DISCUSSION}
1746: \label{sect:m83:disc}
1747:
1748: Different HVC production scenarios predict different characteristics for
1749: the HVCs. The galactic fountain predicts that material will be projected
1750: on the disk, while tidal stripping and accretion predict HVCs across the
1751: field of view. We can divide the detected AVCs into two categories: those
1752: projected on the disk (AVCs 1, 2 and 4 and the low relative velocity
1753: extended emission), and those projected away from the disk. Separate
1754: treatment of each group places constraints on the importance of the various
1755: HVC formation schemes.
1756:
1757: \subsection{Galactic Fountain: ``The Beard'' and HVCs}
1758: \label{sect:m83:disc:beard}
1759:
1760: Under the galactic fountain model, hot gas is ejected from the disk by
1761: multiple supernova explosions in star-forming regions. This material rises
1762: at roughly the adiabatic sound speed, moves outward due to a decrease in
1763: the radial gravitational potential, and rotates more slowly due to
1764: conservation of angular momentum. After the gas cools, it loses its
1765: buoyancy and spirals back to the disk, either as small clouds or sheets
1766: \citep{ShapiroField76,Bregman80}.
1767:
1768: The observational implications of this model depend in detail on the nature
1769: of the hot corona and magnetic field, which are ill-constrained, but
1770: several general features should be apparent in observations of external
1771: galaxies. First of all, discrete clouds of scale $\sim$ 1 kpc should be
1772: observable at velocities between systemic and systemic plus the coronal
1773: sound speed ($\lesssim 150$ \kps), on both wings of the disk velocity
1774: profile. Second, \hi\ emission could appear at velocities between systemic
1775: and the disk, i.e., on the low-relative velocity wing of the disk profile.
1776: This emission could be spatially extended (spanning a few kpc) if
1777: small-scale condensations are suppressed
1778: \citep{Field65,MathewsBregman78,ChevalierOegerle79,Bregman80}. Finally,
1779: anomalous velocity \hi\ should coincide with the spiral arms and \halpha\
1780: disk emission, especially the discrete high-velocity clumps, which would be
1781: infalling near their point of origin.
1782:
1783: Our results are consistent with these predictions. The extended anomalous
1784: disk of \hi\ appears to be rotating more slowly than the main \hi\ disk,
1785: which would be the case for emission from a galactic fountain.
1786: This material must be spatially separated from the bulk \hi, and we
1787: conclude that it lies in a vertically extended disk.
1788: A similar
1789: phenomenon has been observed in a number of inclined spiral galaxies,
1790: including NGC 891 \citep{SwatersSancisivanderHulst97,Oosterlooetal2007},
1791: NGC 2403 \citep{Fraternalietal02a}, NGC 4559 \citep{Barbierietal2005},
1792: NGC 253 \citep{Boomsmaetal2005},
1793: and NGC 6946 \citep{Boomsmaetal2008}.
1794: In particular, for the the $i = 60\arcdeg$ spiral NGC 2403,
1795: \citet{Fraternalietal02a} argue that
1796: this emission, termed the ``beard'' due to its appearance in \pv\ diagrams,
1797: arises from a vertically extended \hi\ component that is rotating 20--50
1798: \kps\ more slowly than the cold disk. They model the anomalous disk and
1799: characterize it as having a radial inflow at velocity $\lesssim$ 20 \kps,
1800: implying an inflowing mass rate of $\mdot~\sim~0.3$--0.6~\msun~yr$^{-1}$
1801: for an infall timescale of $\sim \eez{8}$ yr. They also detect
1802: \eez{6}--\eez{7}~\msun\ ``streams'' and ``spurs'' of \hi\ emission, ranging
1803: in size from 5--10 kpc and deviation velocity from 50--60 \kps. They
1804: conclude that this material arises from a galactic fountain, with a mass
1805: exchange rate close to the predicted value of $\sim$ 1 \msun~yr$^{-1}$.
1806: Further evidence is provided by {\it Chandra\/} X-ray observations showing
1807: an extended corona of $T \sim 5\eex{6}$ K gas with a radiative cooling rate
1808: of 0.1--0.2~\msun~yr$^{-1}$ \citep{Fraternalietal02b}. Additional modeling
1809: by \citet{FraternaliBinney2006} supports this conclusion, although it
1810: suggests that some amount of IGM accretion must be present to temper the
1811: fountain and produce the low angular momentum material.
1812:
1813: The face-on inclination of M 83 makes it difficult to model the rotation of
1814: the beard component in the same way as \citet{Fraternalietal02a},
1815: \citet{FraternaliBinney2006}, and others. It is also not clear whether the
1816: anomalous disk is radially infalling as is the case of NGC 2403, nor is it
1817: clear how vertically extended the material is. If the beard is the result
1818: of a galactic fountain, it is likely that it represents material near
1819: maximum $z$ because of its low velocity, and it probably is gas that has
1820: recently condensed and is beginning to fall back onto the disk.
1821: The discrete clouds projected on the disk at higher deviation velocities
1822: would be in freefall at lower $z$. If the column density of a cloud
1823: exceeds that of the disk, the cloud would punch through and disrupt the
1824: \hi\ disk in that location. AVC 3 is projected over a region of low disk
1825: \nh, however it is projected near the warp and not the spiral arms, so it
1826: may have a different origin than the other disk-coincident AVCs.
1827:
1828: The dynamical properties of the AV gas are consistent with what would be
1829: expected in a galactic fountain.
1830: The beard mass of 5.6\eex{7}~\msun\ (8.4\eex{7}~\msun\ including the AVCs)
1831: and an estimated cooling/freefall time of 5\eex{7} yr leads to a mass
1832: exchange rate of 1~\msun~yr$^{-1}$, similar to the predictions of the
1833: model \citep{Bregman80}.
1834: The star formation rates of M 83 and the Milky Way are similar, so one
1835: would expect similar mass exchange rates. \citet{Talbot80} estimates a
1836: star formation rate (SFR) in the Milky Way between 1.7~\msun~yr$^{-1}$ (from
1837: converting CO measurements to H$_2$) and 2.0~\msun~yr$^{-1}$ (from \halpha).
1838: \citet{BellKennicutt01} constrain the the SFR of M 83 to be between
1839: 1.1--2.4~\msun~yr$^{-1}$ from FUV and \halpha\ data, using the relations of
1840: \citet{Kennicutt98}. The mass infall rates in both galaxies are estimated
1841: to be $\sim$ 1~\msun~yr$^{-1}$, consistent with the similar SFR.
1842: The kinetic energy of the discrete AVCs projected
1843: over the disk total about 7\eex{53}~erg, equaling the total output of
1844: about 1000 supernovae. This is a lower limit to the kinetic energy
1845: initially imparted to the gas by the collection of supernovae, since the
1846: velocity is measured in only one direction and the $z$ height of the AVCs
1847: is unknown. A supernova rate of 0.01~yr$^{-1}$ is consistent
1848: with this if only 1\% of the energy is converted to kinetic energy.
1849:
1850: \subsection{Galactic Disruption and Accretion: External HVCs}
1851: \label{sect:m83:disc:accret}
1852:
1853: It is difficult to reconcile the presence of AVCs projected outside the
1854: stellar disk with the galactic fountain model, thus there is likely another
1855: HVC production scenario at work. AVCs 3, 7 and 8 are projected on the \hi\
1856: warp/ring, and are possibly associated with that. The remainder of the AVCs
1857: are compact, and all except AVCs 5 and 6 are likely spurious detections.
1858:
1859: The optical results rule out emission from a population of dwarf galaxies
1860: similar to those in the Local Group. The 3-$\sigma$ detection limits for
1861: the AVCs projected outside the \hi\ disk are at least 26.8 \rsloan\
1862: magnitudes per square arcsec. Assuming an integrated color \vr\ $\sim$
1863: 0.4--0.6, typical of Local Group dwarfs \citep{Mateo98}, and using the SDSS
1864: filter transformation from \citet{Smithetal02}, this corresponds to about
1865: 27.2 \V\ magnitudes per square arcsec. This limit is faint enough to
1866: detect the center of any of the dwarfs tabulated by \citet{Mateo98}.
1867: Likewise, the inferred ratios of $\mhi/L_V > 2 \msun/\lsun$ are larger than
1868: those of the majority of Local Group dwarfs, indicating a dearth of
1869: starlight compared to these systems.
1870:
1871: Our optical depth is similar to that of other HVC studies.
1872: \citet{Willmanetal2002b} use SDSS data to search for stellar components in
1873: 13 Galactic HVCs, detecting none to limits of 26.7--30.1 \V\ magnitudes per
1874: square arcsec. \citet{Simonetal2006} obtain an upper limit of 25.25--26.25
1875: \V\ magnitudes per square arcsec for HVC Complex H, depending on the
1876: assumed stellar population. Fainter companions and stellar streams are
1877: observed in the Milky Way and M 31, but they do not have evidence of \hi\
1878: emission \citep[e.g.,][]{Zuckeretal2004,Willmanetal2005}.
1879:
1880: If they are real, these \hi\ clouds are best explained by a combination of
1881: tidal disruption and accretion. The \hi\ warp, \hi\ ring, and elongated,
1882: gas-free companion KK208 are all evidence supporting a tidal interaction at
1883: some point in the galaxy's past. In addition, an unpublished wide-field
1884: \hi\ map shows evidence of an \hi\ arm to the north of the galaxy,
1885: extending east from the tip of the longest NW arm visible in
1886: Figure~\ref{fig:m83mom0} \citep{Parketal01}. Our field of view is
1887: insufficient to include this region, but it is further evidence of a tidal
1888: interaction that can strip diffuse material and produce anomalous \hi\
1889: emission.
1890:
1891: \subsection{Limits on the Mass Distribution of Galactic HVCs}
1892: \label{sect:m83:disc:mwcomp}
1893:
1894: Our results are sensitive enough to place constraints on the mass
1895: distribution of the HVC ensemble in M 83 and, by extension, the Milky Way.
1896: Because of the variable mass sensitivity across the field, we restrict our
1897: analysis to AVCs detected within the HPBW of the primary beam. We assume
1898: that the full distribution of M 83 HVCs falls within this region, which is
1899: a reasonable assumption if the vertical distance $z$ is less than about 50
1900: kpc (and $z \sin{i} \lesssim 17$ kpc or 15\arcmin). Only the eight confirmed
1901: detections lie within the primary HPBW, and we scale our estimates of the
1902: false detection rate to the reduced number of independent sightlines we are
1903: now sampling. Five of the eight detected clumps have masses in a region
1904: where the false detection rate is significant (greater than 0.1 detection
1905: expected per mass bin), while the remaining AVCs are much more massive and
1906: unlikely to be spurious. Assuming a Poisson distribution for false
1907: detections, and using the integrated false detection rate as the mean
1908: ``background'' expected for a mass bin, we determine the
1909: background-corrected mass distribution with 90\% confidence limits. This
1910: is shown in Figure \ref{fig:m83mwmasscomp}, using the tabulated data of
1911: \citet{Gehrels86} to calculate confidence limits for low number counts. We
1912: obtain upper limits above an HVC mass of $1.2\eex{6} \msun$ and below a
1913: mass of $6\eex{5} \msun$; below $4\eex{5} \msun$, the completeness falls
1914: below 50\%, so we have cut off the distribution at this point.
1915:
1916: \begin{figure}
1917: \plotone{f17.eps}
1918: \caption{Mass distribution of the Milky Way HVCs, assuming uniform
1919: spherical distributions of varying mean distance and using the HVC catalog
1920: of \citet{WvW91}. The points show the distribution of our lowest-mass
1921: detections in M 83, with 95\% confidence errorbars and upper limits.
1922: Three detections lie off the horizontal scale at $\mhi > 5\eex{6}$ \msun.
1923: This distribution has been corrected for the false discovery rate (the
1924: ``background'') and the source detection efficiency. If the underlying HVC
1925: mass distributions of the two galaxies are identical, the Milky Way HVCs
1926: must be closer than about 25 kpc or we would have detected more HVCs in M
1927: 83.}
1928: \label{fig:m83mwmasscomp}
1929: \end{figure}
1930:
1931: To draw conclusions about the Galactic HVC population, two simplifying
1932: assumptions must be made. First, we assume that the underlying HVC mass
1933: distributions are identical for the Milky Way and M 83. Second, we assume
1934: the Galactic HVCs are contained in a uniform spherical distribution about
1935: the Galactic center. While this is likely incorrect, it is a sensible
1936: choice given our lack of knowledge of their distances. The collection of
1937: Galactic HVCs across the whole sky has been cataloged by \citet{WvW91}, and
1938: excluding the Magellanic Stream and outer Galactic arm, we randomly assign
1939: distances drawn from the assumed distribution to the cataloged HVCs. The
1940: assigned distances and cataloged \hi\ fluxes produce an estimate of the
1941: Galactic HVC mass distribution. This procedure was repeated
1942: 500 times to fill in the high-mass regions, and the resulting distribution
1943: was scaled to the number of HVCs in the catalog (560).
1944:
1945: The derived HVC mass distribution is plotted in Figure
1946: \ref{fig:m83mwmasscomp} for four different mean Galactic HVC distances,
1947: along with the HVC mass distribution for M 83. While the confidence limits
1948: are large, it is clear that an ensemble distance of less than about 25 kpc
1949: is consistent with the M 83 distribution, while a larger average distance
1950: results in a larger number of massive clouds. Such clouds would have been
1951: easily visible in our M 83 observations. It is possible that some HVCs in
1952: M 83 could fall outside the HPBW and therefore be missed if the distance is
1953: much greater than 50 kpc, however even at 50 kpc we detect an order of
1954: magnitude fewer HVCs than should be present in M 83. In addition, the
1955: heliocentric distances are upper limits to the more fundamental $z$ height
1956: of the HVCs, therefore the distances of the HVCs from the plane are
1957: consistent with being less than about 10 kpc. Finally, if the HVCs
1958: cataloged by \citet{WvW91} are not distinct but rather clumps within larger
1959: complexes, then the mass distribution derived here will underestimate the
1960: number of high-mass HVCs, and the constraints imposed by our results would
1961: be stronger.
1962:
1963: These constraints are consistent with previous results. \citet{Wakker01} has
1964: published a catalog of 18 HVCs and 8 IVCs with distances and metallicities
1965: constrained by absorption against background stars or extragalactic
1966: sources. The metallicities vary from solar to a few percent of solar. The
1967: upper bounds to the vertical $z$ heights range from 0.1 kpc to 7 kpc, and
1968: the corresponding mass limits range from about \eez{4} to 2\eex{6} \msun.
1969: More recent results indicate distances of 5--15 kpc to a handful of
1970: additional Galactic HVCs
1971: \citep{Thometal2006,Wakkeretal2007,Wakkeretal2008}.
1972: Similarly, \citet{Putmanetal2003} have detected \halpha\ emission in 25 HVCs
1973: and use this to limit the distance of these objects from the ionizing
1974: Galactic radiation field and thus constrain their height above the Galactic
1975: plane. The values range from $5 < z < 40$ kpc.
1976: Limits from external galaxies are similar.
1977: \citet{Thilkeretal04} detected an ensemble of 20 discrete \hi\ clouds
1978: within 50 kpc of M 31. These sources have \hi\ masses in the range
1979: \eez{5}--\eez{7} \msun. A handful of the objects identified by these
1980: authors would be visible if present in M 83, and our results are consistent
1981: with this. \citet{Pisanoetal2007} survey six analogs of the Local
1982: Group and detect no extragalactic \hi\ emission. Using a similar analysis
1983: to ours, they conclude the Galactic HVCs must be distributed within 90 kpc
1984: and have average \hi\ mass less than about $4\eex{5} \msun$. The handful
1985: of discrete HVCs we detect in M 83 could represent the high-mass tail of a
1986: much more abundant low-mass, nearby population.
1987:
1988:
1989: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1990: % SECTION -- Conclusions
1991: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1992: \section{SUMMARY AND CONCLUSIONS}
1993:
1994: We have performed deep, wide-field imaging of the \hi\ in M 83, mapping the
1995: outer features for the first time. The \hi\ ring and warp seen by
1996: \citet{TilanusAllenM83} are confirmed, and the outer arms extend to a
1997: radius of 34 kpc and show reflective symmetry, hinting at a tidal formation
1998: scenario. At least 80\% of the \hi\ mass is located outside the optical
1999: extent of the galaxy, with the possibility that more \hi\ lies undetected
2000: outside the primary beam HPBW. The optical companion KK208
2001: \citep[][see Figure \ref{fig:m83opt}]{KarachentsevaKarachentsev98}
2002: contains no \hi\ to a limiting column density of 1\eex{19} \cm.
2003:
2004: We discovered a spatially-extended component of anomalous-velocity gas
2005: deviating 40--50 \kps\ from the bulk \hi\ disk and coincident with it in
2006: projection, with a line-of-sight velocity dispersion of 10--15 \kps. We
2007: interpreted this as a vertically extended disk rotating in the same sense
2008: but about 100 \kps\ more slowly than the kinematically cold, thin disk.
2009: The 5.6\eex{7} \msun\ of \hi\ it contains is 5.5\% of the total \hi\
2010: within the stellar disk of the galaxy.
2011:
2012: We have introduced a new technique of searching for faint extended radio
2013: emission, combining several existing tools and including statistically
2014: robust data modeling. Other recent studies have used similar techniques of
2015: contouring and extended source extraction
2016: \citep{Putmanetal2002,deHeijBraunBurton02}. These developments are driven
2017: by the need for lower detection limits in projects constraining the missing
2018: satellites problem and obtaining a full census of the Galactic HVCs.
2019:
2020: Using this technique, our observations are sensitive to unresolved sources
2021: ($d \lesssim 1$ kpc) with masses greater than 5\eex{5} \msun. We have
2022: discovered 14 discrete anomalous-velocity emission sources,
2023: of which we consider 8 to be real detections,
2024: ranging in \hi\
2025: mass from 6\eex{5} to 1.5\eex{7} \msun\ and projected on and off the disk.
2026: Three of these sources are high-significance detections and appear to be
2027: distinct \hi\ clouds with masses in excess of 5\eex{6} \msun. They are
2028: spatially extended and coincide in projection with the optical spiral arms.
2029: The flux distribution of the low-mass sources is inconsistent with purely
2030: spurious detections, therefore we conclude that most of these are real HVCs
2031: in the M 83 system. They are generally unresolved, off-disk features and
2032: are free of diffuse optical light to a limiting surface brightness of 27
2033: \rsloan\ mag per square arcsec.
2034:
2035: We conclude that a combination of a galactic fountain and tidal stripping
2036: are responsible for the anomalous \hi\ features that we observe in M 83.
2037: The handful of HVCs we have detected are likely to represent the high-mass
2038: end of a substantial HVC population in this galaxy. We expect future deep
2039: \hi\ observations of nearby spirals to add to the growing number of HVC
2040: analogs detected in external galaxies, and to thereby shed light on the
2041: ubiquity of HVC activity and the nature of our own Galaxy's gaseous
2042: neighborhood.
2043:
2044:
2045: \acknowledgments
2046:
2047: We thank the staff of the VLA and CTIO for their assistance with the
2048: planning and execution of these observations. We especially thank Jonathan
2049: Williams and Juan Uson for providing access to and help with their source
2050: detection software code, and the anonmyous referee for constructive
2051: comments that improved the manuscript. EDM would like to thank Mario
2052: Mateo, Hugh Aller, and Tim McKay for their helpful suggestions toward the
2053: improvement of this work.
2054:
2055: \appendix
2056:
2057: \section{A NEW SOURCE DETECTION METHOD FOR SPECTRAL SYNTHESIS IMAGING}
2058: \label{app:snrch}
2059:
2060: We have developed a suite of software that searches a spatially-coherent,
2061: 3-d data cube for signal in a statistical way. This software borrows
2062: heavily from the methods and code developed by \citet{Uson91} and
2063: \citet{WilliamsdeGeusBlitz94}, but as we have made variations to the
2064: methods and translated much of the code to a new programming language, we
2065: here describe the new software in some detail.
2066:
2067: % program is SNRCH, written largely in PDL with calls to IDL for the
2068: % CLUMPFIND stuff
2069:
2070: The program begins by smoothing the input data cube in velocity with
2071: Gaussian kernels of the form
2072:
2073: \begin{equation}
2074: H(v) = e^{-(v^2/2\sigma_{v}^2)},
2075: \label{eq:snrchkernel}
2076: \end{equation}
2077:
2078: \noindent
2079: where $\sigma_{v}$ is the width of the particular Gaussian. The
2080: convolution is performed by a fast Fourier transform. If one assumes a
2081: Gaussian velocity profile for a clump of emission, the signal due to this
2082: feature as a function of channel or frequency $v$ would be
2083:
2084: \begin{equation}
2085: S(v) = A(v_0) e^{-(v-v_0)^2/2\sigma_{v,0}^2},
2086: \label{eq:snrchflux}
2087: \end{equation}
2088:
2089: \noindent
2090: and the smoothing function acts as a matched filter, enhancing features of
2091: similar width and de-emphasizing narrower or wider features. In a given
2092: channel, the normalized amplitude of the smoothed feature is given by
2093:
2094: \begin{equation}
2095: \langle A(v_0) \rangle = \frac{\D \int S(v)H(v)dv }
2096: {\D \int H(v)^2 dv},
2097: \label{eq:snrchamp}
2098: \end{equation}
2099:
2100: \noindent
2101: where the integral is taken over the entire feature. This value is
2102: maximized when $\sigma_{v} = \sigma_{v,0}$. The normalization factor
2103: $1/{\D \int H(v)^2 dv}$ is such that the amplitude of the smoothed feature
2104: will equal the amplitude of the original feature under this same condition;
2105: otherwise, the smoothed amplitude will be depressed compared to the
2106: unsmoothed value.
2107:
2108: One can calculate the significance level of such a detection by recognizing
2109: that the smoothed noise is
2110:
2111: \begin{equation}
2112: \langle \sigma(v_0)^2 \rangle =
2113: \frac{\D \int \sigma_i(v)^2 H(v-v_0)^2 dv }
2114: {\D \left[ \int H(v)^2 dv \right] ^2}\;\;\;,
2115: \label{eq:snrchnoise}
2116: \end{equation}
2117:
2118: \noindent
2119: if we assume that neighboring channels are independent and the noise (given
2120: as $\sigma_i$ for channel $i$) sums in quadrature. The noise in each
2121: channel is determined by reflecting the negative portion of the data
2122: histogram across the mean and performing an iterative 4-$\sigma$ rejection
2123: until the measured $\sigma$ converges.
2124: The significance level (S/N) is calculated as $\langle A(v_0) \rangle /
2125: \langle \sigma(v_0)^2 \rangle^{1/2}$ and written to disk as a data cube.
2126:
2127: Peaks in S/N are identified down to some detection threshold, typically
2128: around 4-$\sigma$, with neighboring above-threshold pixels merged into
2129: groups. These 3-d groups or ``islands'' serve as kernels for an iterative
2130: contouring scheme, with successively lower thresholds applied to the S/N
2131: cube. Each group is extended by adding new pixels that are above threshold
2132: and share at least two corners with any member pixel. That is, only
2133: neighbors that share at least one dimensional plane with the member are
2134: added, so that neighbors at a diagonal in all three dimensions are
2135: excluded. The algorithm for this procedure is based on the CLUMPFIND
2136: software developed by \citet{WilliamsdeGeusBlitz94}. The contours are
2137: produced in intervals (which only affect the processing speed) down to a
2138: specified level, typically 3-$\sigma$, iterating at each level until no new
2139: neighbors are added to any groups and merging groups which meet the
2140: neighbor criterion. Finally, any immediate neighbors (i.e., sharing four
2141: corners or two parameter planes) above 1.5-$\sigma$ are added to each group
2142: to improve flux measurements. No merging or iterating is done at this
2143: point. Groups are eliminated if they contain fewer pixels than a
2144: completely unresolved source (10 pixels in our case), as these features are
2145: unphysical.
2146:
2147: The group IDs are written to a data cube, which is applied to the original
2148: data cube as a mask for each group in turn. The total flux, centroid, and
2149: other parameters of the group are determined from the original data cube,
2150: using only the pixels that are members of the group. The primary beam
2151: correction is applied at this point, and the corrected flux and mass are
2152: recorded along with the uncorrected parameters.
2153:
2154:
2155:
2156: \begin{thebibliography}{68}
2157: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
2158:
2159: \bibitem[{{Barbieri} {et~al.}(2005){Barbieri}, {Fraternali}, {Oosterloo},
2160: {Bertin}, {Boomsma}, \& {Sancisi}}]{Barbierietal2005}
2161: {Barbieri}, C.~V., {Fraternali}, F., {Oosterloo}, T., {Bertin}, G., {Boomsma},
2162: R., \& {Sancisi}, R. 2005, \aap, 439, 947
2163:
2164: \bibitem[{{Bell} \& {Kennicutt}(2001)}]{BellKennicutt01}
2165: {Bell}, E.~F. \& {Kennicutt}, R.~C. 2001, \apj, 548, 681
2166:
2167: \bibitem[{{Bland-Hawthorn} \& {Maloney}(1997)}]{Bland-HawthornMaloney97}
2168: {Bland-Hawthorn}, J. \& {Maloney}, P.~R. 1997, Proc.~Astron.~Soc.~Aust., 14, 59
2169:
2170: \bibitem[{{Bland-Hawthorn} {et~al.}(1998){Bland-Hawthorn}, {Veilleux}, {Cecil},
2171: {Putman}, {Gibson}, \& {Maloney}}]{Bland-Hawthornetal98}
2172: {Bland-Hawthorn}, J., {Veilleux}, S., {Cecil}, G.~N., {Putman}, M.~E.,
2173: {Gibson}, B.~K., \& {Maloney}, P.~R. 1998, \mnras, 299, 611
2174:
2175: \bibitem[{{Blitz} {et~al.}(1999){Blitz}, {Spergel}, {Teuben}, {Hartmann}, \&
2176: {Burton}}]{Blitz99}
2177: {Blitz}, L., {Spergel}, D.~N., {Teuben}, P.~J., {Hartmann}, D., \& {Burton},
2178: W.~B. 1999, \apj, 514, 818
2179:
2180: \bibitem[{{Boomsma} {et~al.}(2005){Boomsma}, {Oosterloo}, {Fraternali}, {van
2181: der Hulst}, \& {Sancisi}}]{Boomsmaetal2005}
2182: {Boomsma}, R., {Oosterloo}, T.~A., {Fraternali}, F., {van der Hulst}, J.~M., \&
2183: {Sancisi}, R. 2005, \aap, 431, 65
2184:
2185: \bibitem[{{Boomsma} {et~al.}(2008){Boomsma}, {Oosterloo}, {Fraternali}, {van
2186: der Hulst}, \& {Sancisi}}]{Boomsmaetal2008}
2187: ---. 2008, \aap, 490, 555
2188:
2189: \bibitem[{{Braun} \& {Burton}(1999)}]{BraunBurton99}
2190: {Braun}, R. \& {Burton}, W.~B. 1999, \aap, 341, 437
2191:
2192: \bibitem[{{Bregman}(1980)}]{Bregman80}
2193: {Bregman}, J.~N. 1980, \apj, 236, 577
2194:
2195: \bibitem[{{Bregman} \& {Harrington}(1986)}]{BregmanHarrington86}
2196: {Bregman}, J.~N. \& {Harrington}, J.~P. 1986, \apj, 309, 833
2197:
2198: \bibitem[{{Briggs}(1995)}]{Briggs95}
2199: {Briggs}, D.~S. 1995, PhD thesis, New Mexico Institute of Mining and
2200: Technology, Socorro, New Mexico
2201:
2202: \bibitem[{{Chevalier} \& {Oegerle}(1979)}]{ChevalierOegerle79}
2203: {Chevalier}, R.~A. \& {Oegerle}, W.~R. 1979, \apj, 227, 398
2204:
2205: \bibitem[{{Clark}(1980)}]{Clark80}
2206: {Clark}, B.~G. 1980, \aap, 89, 377
2207:
2208: \bibitem[{{Comte}(1981)}]{Comte81}
2209: {Comte}, G. 1981, \aaps, 44, 441
2210:
2211: \bibitem[{{de Heij} {et~al.}(2002){de Heij}, {Braun}, \&
2212: {Burton}}]{deHeijBraunBurton02}
2213: {de Heij}, V., {Braun}, R., \& {Burton}, W.~B. 2002, \aap, 391, 159
2214:
2215: \bibitem[{{de Vaucouleurs} {et~al.}(1991){de Vaucouleurs}, {de Vaucouleurs},
2216: {Corwin}, {Buta}, {Paturel}, \& {Fouque}}]{RC3}
2217: {de Vaucouleurs}, G., {de Vaucouleurs}, A., {Corwin}, H.~G., {Buta}, R.~J.,
2218: {Paturel}, G., \& {Fouque}, P. 1991, {Third Reference Catalogue of Bright
2219: Galaxies} (New York:Springer-Verlag)
2220:
2221: \bibitem[{{Field}(1965)}]{Field65}
2222: {Field}, G.~B. 1965, \apj, 142, 531
2223:
2224: \bibitem[{{Fraternali} \& {Binney}(2006)}]{FraternaliBinney2006}
2225: {Fraternali}, F. \& {Binney}, J.~J. 2006, \mnras, 366, 449
2226:
2227: \bibitem[{{Fraternali} \& {Binney}(2008)}]{FraternaliBinney2008}
2228: ---. 2008, \mnras, 386, 935
2229:
2230: \bibitem[{{Fraternali} {et~al.}(2002{\natexlab{a}}){Fraternali}, {Cappi},
2231: {Sancisi}, \& {Oosterloo}}]{Fraternalietal02b}
2232: {Fraternali}, F., {Cappi}, M., {Sancisi}, R., \& {Oosterloo}, T.
2233: 2002{\natexlab{a}}, \apj, 578, 109
2234:
2235: \bibitem[{{Fraternali} {et~al.}(2002{\natexlab{b}}){Fraternali}, {van Moorsel},
2236: {Sancisi}, \& {Oosterloo}}]{Fraternalietal02a}
2237: {Fraternali}, F., {van Moorsel}, G., {Sancisi}, R., \& {Oosterloo}, T.
2238: 2002{\natexlab{b}}, \aj, 123, 3124
2239:
2240: \bibitem[{{Gehrels}(1986)}]{Gehrels86}
2241: {Gehrels}, N. 1986, \apj, 303, 336
2242:
2243: \bibitem[{{Greisen}(1998)}]{AIPSCookbook}
2244: {Greisen}, E. 1998, {The AIPS Cookbook}, National Radio Astronomy Observatory
2245: (http://www.cv.nrao.edu/aips/cook.html)
2246:
2247: \bibitem[{{H{\"o}gbom}(1974)}]{Hogbom74}
2248: {H{\"o}gbom}, J.~A. 1974, \aaps, 15, 417
2249:
2250: \bibitem[{{Huchtmeier} \& {Bohnenstengel}(1981)}]{HuchtBohn81}
2251: {Huchtmeier}, W.~K. \& {Bohnenstengel}, H.-D. 1981, \aap, 100, 72
2252:
2253: \bibitem[{{Kamphuis} \& {Briggs}(1992)}]{KamphuisBriggs92}
2254: {Kamphuis}, J. \& {Briggs}, F. 1992, \aap, 253, 335
2255:
2256: \bibitem[{{Kamphuis} \& {Sancisi}(1993)}]{KamphuisSancisi93}
2257: {Kamphuis}, J. \& {Sancisi}, R. 1993, \aap, 273, L31
2258:
2259: \bibitem[{{Kamphuis} {et~al.}(1991){Kamphuis}, {Sancisi}, \& {van der
2260: Hulst}}]{Kamphuisetal91}
2261: {Kamphuis}, J., {Sancisi}, R., \& {van der Hulst}, T. 1991, \aap, 244, L29
2262:
2263: \bibitem[{{Karachentsev} {et~al.}(2002){Karachentsev}, {Sharina}, {Dolphin},
2264: {Grebel}, {Geisler}, {Guhathakurta}, {Hodge}, {Karachentseva}, {Sarajedini},
2265: \& {Seitzer}}]{Karachentsevetal02}
2266: {Karachentsev}, I.~D., et al. 2002, \aap, 385, 21
2267:
2268: \bibitem[{{Karachentseva} \&
2269: {Karachentsev}(1998)}]{KarachentsevaKarachentsev98}
2270: {Karachentseva}, V.~E. \& {Karachentsev}, I.~D. 1998, \aaps, 127, 409
2271:
2272: \bibitem[{{Kennicutt}(1998)}]{Kennicutt98}
2273: {Kennicutt}, R.~C. 1998, \apj, 498, 541
2274:
2275: \bibitem[{{Klypin} {et~al.}(1999){Klypin}, {Kravtsov}, {Valenzuela}, \&
2276: {Prada}}]{Klypinetal99}
2277: {Klypin}, A., {Kravtsov}, A.~V., {Valenzuela}, O., \& {Prada}, F. 1999, \apj,
2278: 522, 82
2279:
2280: \bibitem[{{Koribalski} {et~al.}(2004){Koribalski}, {Staveley-Smith}, {Kilborn},
2281: {Ryder}, {Kraan-Korteweg}, {Ryan-Weber}, {Ekers}, {Jerjen}, {Henning},
2282: {Putman}, {Zwaan}, {de Blok}, {Calabretta}, {Disney}, {Minchin}, {Bhathal},
2283: {Boyce}, {Drinkwater}, {Freeman}, {Gibson}, {Green}, {Haynes}, {Juraszek},
2284: {Kesteven}, {Knezek}, {Mader}, {Marquarding}, {Meyer}, {Mould}, {Oosterloo},
2285: {O'Brien}, {Price}, {Sadler}, {Schr{\"o}der}, {Stewart}, {Stootman}, {Waugh},
2286: {Warren}, {Webster}, \& {Wright}}]{HIPASS04}
2287: {Koribalski}, B.~S., et al. 2004, \aj, 128, 16
2288:
2289: \bibitem[{{Landolt}(1992)}]{Landolt92}
2290: {Landolt}, A.~U. 1992, \aj, 104, 340
2291:
2292: \bibitem[{{Lockman}(2003)}]{Lockman2003}
2293: {Lockman}, F.~J. 2003, \apjl, 591, L33
2294:
2295: \bibitem[{{Maller} \& {Bullock}(2004)}]{MallerBullock04}
2296: {Maller}, A.~H. \& {Bullock}, J.~S. 2004, \mnras, 355, 694
2297:
2298: \bibitem[{{Mateo}(1998)}]{Mateo98}
2299: {Mateo}, M.~L. 1998, \araa, 36, 435
2300:
2301: \bibitem[{{Mathews} \& {Bregman}(1978)}]{MathewsBregman78}
2302: {Mathews}, W.~G. \& {Bregman}, J.~N. 1978, \apj, 224, 308
2303:
2304: \bibitem[{{Mathewson} {et~al.}(1974){Mathewson}, {Cleary}, \&
2305: {Murray}}]{Mathewson74}
2306: {Mathewson}, D.~S., {Cleary}, M.~N., \& {Murray}, J.~D. 1974, \apj, 190, 291
2307:
2308: \bibitem[{{Moore} {et~al.}(1999){Moore}, {Ghigna}, {Governato}, {Lake},
2309: {Quinn}, {Stadel}, \& {Tozzi}}]{Mooreetal99}
2310: {Moore}, B., {Ghigna}, S., {Governato}, F., {Lake}, G., {Quinn}, T., {Stadel},
2311: J., \& {Tozzi}, P. 1999, \apjl, 524, L19
2312:
2313: \bibitem[{{Muller} {et~al.}(1963){Muller}, {Oort}, \& {Raimond}}]{Mulleretal63}
2314: {Muller}, C.~A., {Oort}, J.~H., \& {Raimond}, E. 1963, C.~R.~Acad.~Sci.~Paris,
2315: 257, 1661
2316:
2317: \bibitem[{{Oort}(1966)}]{Oort66}
2318: {Oort}, J.~H. 1966, \bain, 18, 421
2319:
2320: \bibitem[{{Oort}(1970)}]{Oort70}
2321: ---. 1970, \aap, 7, 381
2322:
2323: \bibitem[{{Oort}(1981)}]{Oort81}
2324: ---. 1981, \aap, 94, 359
2325:
2326: \bibitem[{{Oosterloo} {et~al.}(2007){Oosterloo}, {Fraternali}, \&
2327: {Sancisi}}]{Oosterlooetal2007}
2328: {Oosterloo}, T., {Fraternali}, F., \& {Sancisi}, R. 2007, \aj, 134, 1019
2329:
2330: \bibitem[{{Park} {et~al.}(2001){Park}, {Kalnajs}, {Freeman}, {Koribalski},
2331: {Staveley-Smith}, \& {Malin}}]{Parketal01}
2332: {Park}, O.-K., {Kalnajs}, A., {Freeman}, K.~C., {Koribalski}, B.,
2333: {Staveley-Smith}, L., \& {Malin}, D.~F. 2001, in ASP Conf. Ser. 230: Galaxy
2334: Disks and Disk Galaxies, 109--110
2335:
2336: \bibitem[{{Pisano} {et~al.}(2007){Pisano}, {Barnes}, {Gibson},
2337: {Staveley-Smith}, {Freeman}, \& {Kilborn}}]{Pisanoetal2007}
2338: {Pisano}, D.~J., {Barnes}, D.~G., {Gibson}, B.~K., {Staveley-Smith}, L.,
2339: {Freeman}, K.~C., \& {Kilborn}, V.~A. 2007, \apj, 662, 959
2340:
2341: \bibitem[{{Putman} {et~al.}(2002){Putman}, {de Heij}, {Staveley-Smith},
2342: {Braun}, {Freeman}, {Gibson}, {Burton}, {Barnes}, {Banks}, {Bhathal}, {de
2343: Blok}, {Boyce}, {Disney}, {Drinkwater}, {Ekers}, {Henning}, {Jerjen},
2344: {Kilborn}, {Knezek}, {Koribalski}, {Malin}, {Marquarding}, {Minchin},
2345: {Mould}, {Oosterloo}, {Price}, {Ryder}, {Sadler}, {Stewart}, {Stootman},
2346: {Webster}, \& {Wright}}]{Putmanetal2002}
2347: {Putman}, M.~E., et al. 2002, \aj, 123, 873
2348:
2349: \bibitem[{{Putman} {et~al.}(2003){Putman}, {Bland-Hawthorn}, {Veilleux},
2350: {Gibson}, {Freeman}, \& {Maloney}}]{Putmanetal2003}
2351: {Putman}, M.~E., {Bland-Hawthorn}, J., {Veilleux}, S., {Gibson}, B.~K.,
2352: {Freeman}, K.~C., \& {Maloney}, P.~R. 2003, \apj, 597, 948
2353:
2354: \bibitem[{{Putman} {et~al.}(2004){Putman}, {Thom}, {Gibson}, \&
2355: {Staveley-Smith}}]{Putmanetal2004}
2356: {Putman}, M.~E., {Thom}, C., {Gibson}, B.~K., \& {Staveley-Smith}, L. 2004,
2357: \apjl, 603, L77
2358:
2359: \bibitem[{{Sandage} \& {Tammann}(1987)}]{RSA}
2360: {Sandage}, A. \& {Tammann}, G.~A. 1987, {A revised Shapley-Ames Catalog of
2361: bright galaxies} (2nd ed.; Washinton, DC:Carnegie Institution of
2362: Washington)
2363:
2364: \bibitem[{{Schulman} {et~al.}(1994){Schulman}, {Bregman}, \&
2365: {Roberts}}]{Schulmanetal1994}
2366: {Schulman}, E., {Bregman}, J.~N., \& {Roberts}, M.~S. 1994, \apj, 423, 180
2367:
2368: \bibitem[{{Shapiro} \& {Field}(1976)}]{ShapiroField76}
2369: {Shapiro}, P.~R. \& {Field}, G.~B. 1976, \apj, 205, 762
2370:
2371: \bibitem[{{Simon} {et~al.}(2006){Simon}, {Blitz}, {Cole}, {Weinberg}, \&
2372: {Cohen}}]{Simonetal2006}
2373: {Simon}, J.~D., {Blitz}, L., {Cole}, A.~A., {Weinberg}, M.~D., \& {Cohen}, M.
2374: 2006, \apj, 640, 270
2375:
2376: \bibitem[{{Smith} {et~al.}(2002){Smith}, {Tucker}, {Kent}, {Richmond},
2377: {Fukugita}, {Ichikawa}, {Ichikawa}, {Jorgensen}, {Uomoto}, {Gunn}, {Hamabe},
2378: {Watanabe}, {Tolea}, {Henden}, {Annis}, {Pier}, {McKay}, {Brinkmann}, {Chen},
2379: {Holtzman}, {Shimasaku}, \& {York}}]{Smithetal02}
2380: {Smith}, J.~A., et al. 2002, \aj, 123, 2121
2381:
2382: \bibitem[{{Sternberg} {et~al.}(2002){Sternberg}, {McKee}, \&
2383: {Wolfire}}]{SternbergMcKeeWolfire02}
2384: {Sternberg}, A., {McKee}, C.~F., \& {Wolfire}, M.~G. 2002, \apjs, 143, 419
2385:
2386: \bibitem[{{Stetson}(1987)}]{Stetson87}
2387: {Stetson}, P.~B. 1987, \pasp, 99, 191
2388:
2389: \bibitem[{{Swaters} {et~al.}(1997){Swaters}, {Sancisi}, \& {van der
2390: Hulst}}]{SwatersSancisivanderHulst97}
2391: {Swaters}, R.~A., {Sancisi}, R., \& {van der Hulst}, J.~M. 1997, \apj, 491, 140
2392:
2393: \bibitem[{{Talbot}(1980)}]{Talbot80}
2394: {Talbot}, R.~J. 1980, \apj, 235, 821
2395:
2396: \bibitem[{{Talbot} {et~al.}(1979){Talbot}, {Jensen}, \&
2397: {Dufour}}]{TalbotJensenDufour79}
2398: {Talbot}, R.~J., {Jensen}, E.~B., \& {Dufour}, R.~J. 1979, \apj, 229, 91
2399:
2400: \bibitem[{{Thilker} {et~al.}(2004){Thilker}, {Braun}, {Walterbos}, {Corbelli},
2401: {Lockman}, {Murphy}, \& {Maddalena}}]{Thilkeretal04}
2402: {Thilker}, D.~A., {Braun}, R., {Walterbos}, R.~A.~M., {Corbelli}, E.,
2403: {Lockman}, F.~J., {Murphy}, E., \& {Maddalena}, R. 2004, \apjl, 601, L39
2404:
2405: \bibitem[{{Thom} {et~al.}(2006){Thom}, {Putman}, {Gibson}, {Christlieb},
2406: {Flynn}, {Beers}, {Wilhelm}, \& {Lee}}]{Thometal2006}
2407: {Thom}, C., {Putman}, M.~E., {Gibson}, B.~K., {Christlieb}, N., {Flynn}, C.,
2408: {Beers}, T.~C., {Wilhelm}, R., \& {Lee}, Y.~S. 2006, \apjl, 638, L97
2409:
2410: \bibitem[{{Tilanus} \& {Allen}(1993)}]{TilanusAllenM83}
2411: {Tilanus}, R.~P.~J. \& {Allen}, R.~J. 1993, \aap, 274, 707
2412:
2413: \bibitem[{{Tufte} {et~al.}(1998){Tufte}, {Reynolds}, \&
2414: {Haffner}}]{Tufteetal98}
2415: {Tufte}, S.~L., {Reynolds}, R.~J., \& {Haffner}, L.~M. 1998, \apj, 504, 773
2416:
2417: \bibitem[{{Uson} {et~al.}(1991){Uson}, {Bagri}, \& {Cornwell}}]{Uson91}
2418: {Uson}, J.~M., {Bagri}, D.~S., \& {Cornwell}, T.~J. 1991, \apjl, 377, L65
2419:
2420: \bibitem[{{van der Hulst} \& {Sancisi}(1988)}]{vanderHulstSancisi88}
2421: {van der Hulst}, T. \& {Sancisi}, R. 1988, \aj, 95, 1354
2422:
2423: \bibitem[{{Wakker}(1991)}]{Wakker91a}
2424: {Wakker}, B.~P. 1991, \aap, 250, 499
2425:
2426: \bibitem[{{Wakker}(2001)}]{Wakker01}
2427: ---. 2001, \apjs, 136, 463
2428:
2429: \bibitem[{{Wakker} \& {Schwarz}(1988)}]{WakkerSchwarz88}
2430: {Wakker}, B.~P. \& {Schwarz}, U.~J. 1988, \aap, 200, 312
2431:
2432: \bibitem[{{Wakker} \& {van Woerden}(1991)}]{WvW91}
2433: {Wakker}, B.~P. \& {van Woerden}, H. 1991, \aap, 250, 509
2434:
2435: \bibitem[{{Wakker} \& {van Woerden}(1997)}]{WvW97}
2436: ---. 1997, \araa, 35, 217
2437:
2438: \bibitem[{{Wakker} {et~al.}(2007){Wakker}, {York}, {Howk}, {Barentine},
2439: {Wilhelm}, {Peletier}, {van Woerden}, {Beers}, {Ivezi{\'c}}, {Richter}, \&
2440: {Schwarz}}]{Wakkeretal2007}
2441: {Wakker}, B.~P., et al. 2007, \apjl, 670, L113
2442:
2443: \bibitem[{{Wakker} {et~al.}(2008){Wakker}, {York}, {Wilhelm}, {Barentine},
2444: {Richter}, {Beers}, {Ivezi{\'c}}, \& {Howk}}]{Wakkeretal2008}
2445: ---. 2008, \apj, 672, 298
2446:
2447: \bibitem[{{Weiner} \& {Williams}(1996)}]{WeinerWilliams96}
2448: {Weiner}, B.~J. \& {Williams}, T.~B. 1996, \aj, 111, 1156
2449:
2450: \bibitem[{{Williams} {et~al.}(1994){Williams}, {de Geus}, \&
2451: {Blitz}}]{WilliamsdeGeusBlitz94}
2452: {Williams}, J.~P., {de Geus}, E.~J., \& {Blitz}, L. 1994, \apj, 428, 693
2453:
2454: \bibitem[{{Willman} {et~al.}(2002){Willman}, {Dalcanton}, {Ivezi{\'c}},
2455: {Schneider}, \& {York}}]{Willmanetal2002b}
2456: {Willman}, B., {Dalcanton}, J., {Ivezi{\'c}}, {\v Z}., {Schneider}, D.~P., \&
2457: {York}, D.~G. 2002, \aj, 124, 2600
2458:
2459: \bibitem[{{Willman} {et~al.}(2005){Willman}, {Dalcanton}, {Martinez-Delgado},
2460: {West}, {Blanton}, {Hogg}, {Barentine}, {Brewington}, {Harvanek}, {Kleinman},
2461: {Krzesinski}, {Long}, {Neilsen}, {Nitta}, \& {Snedden}}]{Willmanetal2005}
2462: {Willman}, B., {Dalcanton}, J.~J., {Martinez-Delgado}, D., {West}, A.~A.,
2463: {Blanton}, M.~R., {Hogg}, D.~W., {Barentine}, J.~C., {Brewington}, H.~J.,
2464: {Harvanek}, M., {Kleinman}, S.~J., {Krzesinski}, J., {Long}, D., {Neilsen},
2465: Jr., E.~H., {Nitta}, A., \& {Snedden}, S.~A. 2005, \apjl, 626, L85
2466:
2467: \bibitem[{{Zucker} {et~al.}(2004){Zucker}, {Kniazev}, {Bell},
2468: {Mart{\'{\i}}nez-Delgado}, {Grebel}, {Rix}, {Rockosi}, {Holtzman},
2469: {Walterbos}, {Annis}, {York}, {Ivezi{\'c}}, {Brinkmann}, {Brewington},
2470: {Harvanek}, {Hennessy}, {Kleinman}, {Krzesinski}, {Long}, {Newman}, {Nitta},
2471: \& {Snedden}}]{Zuckeretal2004}
2472: {Zucker}, D.~B., {Kniazev}, A.~Y., {Bell}, E.~F., {Mart{\'{\i}}nez-Delgado},
2473: D., {Grebel}, E.~K., {Rix}, H.-W., {Rockosi}, C.~M., {Holtzman}, J.~A.,
2474: {Walterbos}, R.~A.~M., {Annis}, J., {York}, D.~G., {Ivezi{\'c}}, {\v Z}.,
2475: {Brinkmann}, J., {Brewington}, H., {Harvanek}, M., {Hennessy}, G.,
2476: {Kleinman}, S.~J., {Krzesinski}, J., {Long}, D., {Newman}, P.~R., {Nitta},
2477: A., \& {Snedden}, S.~A. 2004, \apjl, 612, L121
2478:
2479: \bibitem[{{Zwaan}(2001)}]{Zwaan01}
2480: {Zwaan}, M.~A. 2001, \mnras, 325, 1142
2481:
2482: \end{thebibliography}
2483:
2484:
2485:
2486:
2487: \end{document}
2488: