0706.1586/ms.tex
1: \documentclass[apj]{emulateapj}
2: \usepackage{graphicx}
3: \usepackage{txfonts}
4: \begin{document}
5: 
6: \title{21-cm synthesis observations of VIRGOHI 21 - a possible dark 
7: galaxy in the Virgo Cluster.}
8: \author{
9: Robert Minchin\altaffilmark{1} \altaffilmark{2}, 
10: Jonathan Davies\altaffilmark{2}, 
11: Michael Disney\altaffilmark{2},
12: Marco Grossi\altaffilmark{2} \altaffilmark{3},
13: Sabina Sabatini\altaffilmark{4}, 
14: Peter Boyce\altaffilmark{5}, 
15: Diego Garcia\altaffilmark{2} \altaffilmark{6},
16: Chris Impey\altaffilmark{7}, 
17: Christine Jordan\altaffilmark{8}, 
18: Robert Lang\altaffilmark{2},
19: Andrew Marble\altaffilmark{7}, 
20: Sarah Roberts\altaffilmark{2}, 
21: \and 
22: Wim van Driel\altaffilmark{9}}
23: \affil{}
24: 
25: \altaffiltext{1}{Arecibo Observatory, National Astronomy and Ionosphere Center,
26: Arecibo, PR 00612, US; rminchin@naic.edu}
27: \altaffiltext{2}{School of Physics and Astronomy, Cardiff University, Cardiff,
28: CF24 3YB, UK; Jonathan.Davies@astro.cf.ac.uk,
29: Mike.Disney@astro.cf.ac.uk, LangRH@cardifff.ac.uk,
30: Sarah.Roberts@astro.cf.ac.uk}
31: \altaffiltext{3}{Istituto di Fisica dello Spazio Interplanetario, via del Fosso
32: del Cavaliere 100, 00133, Rome, Italy; Marco.Grossi@roma1.infn.it}
33: \altaffiltext{4}{Osservatorio Astronomico di Roma, via Frascati 33, I-00040,
34: Monte Porzio, Italy; sabatini@mporzio.astro.it}
35: \altaffiltext{5}{Registry, Cardiff University, Cardiff, 
36: CF10 3UA, UK; BoyceP@cardiff.ac.uk}
37: \altaffiltext{6}{University of Bonn, Germany}
38: \altaffiltext{7}{Steward Observatory, University of Arizona, 933 N. Cherry 
39: Ave., Tucson, AZ 85721-0065, US}
40: \altaffiltext{8}{Jodrell Bank Observatory, University of Manchester, 
41: Macclesfield, Cheshire, SK11 9DL, UK; caj@jb.man.ac.uk}
42: \altaffiltext{9}{Observatoire de Paris, GEPI, CNRS UMR 8111 and Universit\'e 
43: Paris 7, 5 place Jules Janssen, F-92195 Meudon Cedex, France; 
44: Wim.vanDriel@obspm.fr}
45: 
46: \shorttitle{21-cm observations of VIRGOHI 21}
47: \shortauthors{Minchin et al.}
48:  
49: 
50: \begin{abstract}
51: Many observations indicate that dark matter dominates the extra-galactic 
52: Universe, yet no totally dark structure of galactic proportions has ever 
53: been convincingly identified.  Previously we have suggested that 
54: \object{VIRGOHI 21}, 
55: a 21-cm source we found in the Virgo Cluster using Jodrell Bank, was a 
56: possible dark galaxy because of its broad line-width ($\sim 200$ km\,s$^{-1}$) 
57: unaccompanied by any visible gravitational source to account for it.  We have 
58: now imaged VIRGOHI 21 in the neutral-hydrogen line and find what could be a 
59: dark, edge-on, spinning disk with the mass and diameter of a typical spiral 
60: galaxy.  Moreover, VIRGOHI 21 has unquestionably been involved in an 
61: interaction with \object{NGC 4254}, a luminous spiral with an odd one-armed 
62: morphology, but lacking the massive interactor normally linked with such a 
63: feature.  Numerical models of NGC 4254 call for a close interaction 
64: $\sim 10^8$ years ago with a perturber of $\sim 10^{11}$ solar masses.  This 
65: we take as additional evidence for the massive nature of VIRGOHI 21 as there 
66: does not appear to be any other viable candidate. We have also used the Hubble 
67: Space Telescope\footnote{Based on observations made with the NASA/ESA Hubble 
68: Space Telescope, obtained at the Space Telescope Science Institute, which is 
69: operated by the Association of Universities for Research in Astronomy, Inc., 
70: under NASA contract NAS 5-26555. These observations are associated with program
71: \#10562.} to search for stars associated with the H\,{\sc i} and find none 
72: down to an $I$ band surface brightness limit of $31.1 \pm 0.2$ mag 
73: arcsec$^{-2}$.
74: \end{abstract}
75: 
76: \keywords{dark matter -- galaxies: individual (VIRGOHI 21) -- radio lines: 
77: galaxies}
78: 
79: 
80: \section{Introduction}
81: 
82: The ability of a galaxy to form stars depends critically on the fraction of 
83: its mass that forms the baryonic gas disk and the temperature of this gas.  
84: High densities shield the gas from ionising radiation, allowing it to cool 
85: while higher intensity ionizing radiation keeps the gas hot.  Models
86: predict that galaxies can form with gas column densities that prohibit
87: star formation yet provide some self shielding from the ionizing
88: background (Davies et al. 2006). Such dark galaxies are potentially
89: detectable by blind 21-cm surveys of the sky. The possible existence of dark 
90: galaxies has previously been discussed by Jimenez et al. (1997), Hawkins 
91: (1997), Verde et al. (2002), and Davies et al. (2006); see also Taylor \& 
92: Webster (2005) for an alternative view.
93: 
94: Objects detected at 21-cm but with no optical counterparts have been
95: known about for many years; these include high velocity clouds
96: (Wakker \& van Woerden 1997), the Leo Ring (Schneider et al. 
97: 1983), and various gas clouds
98: close to bright galaxies (Kilborn et al. 2000; Boyce et al 
99: 2001; Ryder et al. 2001).  However,
100: none of these objects have the characteristics of a galaxy, i.e.
101: detectable emission over galaxy-sized spatial scales and a velocity
102: structure consistent with a rotating and gravitationally bound disk.  Recently
103: two candidate dark galaxies have been reported: VIRGOHI 21 (Minchin
104: et al. 2005) and HVC Complex H (Simon et al. 2006). A third possibility is the 
105: H\,{\sc i} cloud associated with the Local Group galaxy LGS3 (Robishaw et al. 
106: 2002).
107: 
108: In this paper, we present high resolution H\,{\sc i} observations of VIRGOHI 
109: 21, which we believe add further support to our hypothesis that it is 
110: a dark galaxy.  We also describe HST observations that were specifically 
111: designed to search for red giant stars at the distance of the Virgo Cluster 
112: and so place very faint limits on the surface density of stars that might be 
113: associated with VIRGOHI 21.
114: 
115: \section{21-cm Observations and Analysis}
116: 
117: The new 21-cm data were taken in March 2005 at the Westerbork Synthesis Radio 
118: Telescope (WSRT) in two full 12-hour syntheses and reduced using the
119: {\sc Miriad} package.  The data were flagged for shadowing and on two of the 
120: fourteen 25-meter antennae one polarization was flagged due 
121: to problems with the gain.  A spectral bandwidth of 10 MHz covered the velocity
122: range 930 - 3070 km\,s$^{-1}$.  Removal of the noisier end channels left 230 
123: useful channels of width 8.2 km\,s$^{-1}$ each, giving a velocity resolution 
124: of 10 km\,s$^{-1}$ over the range 980 to 2890 km\,s$^{-1}$.  Continuum removal 
125: was carried out in the UV plane using {\sc uvlin}.  The standard source 3C147
126: was used for calibration.  Cleaning used a robust setting of 1, close to 
127: normal weighting (Briggs 1995). The cleaned cube was Gaussian 
128: smoothed spatially and Hanning smoothed in velocity.  This was then used as
129: a template for regions where flux was present in the data in order to carry 
130: out a deeper cleaning of the dirty image, which gave the final cube used in 
131: the analysis.  The synthesized beam was $99\arcsec\times 30\arcsec$ in size 
132: (extended north-south) and the noise was 0.3 mJy\,beam$^{-1}$\,channel$^{-1}$, 
133: giving a $5\sigma$ column-density limit for sources 25 km\,s$^{-1}$ wide of 
134: $2\times 10^{19}$ atoms cm$^{-2}$.
135: %\clearpage
136: \begin{figure*}
137: \plotone{f1.eps}
138: %\resizebox{\columnwidth}{!}{\includegraphics{f1.eps}}
139: \caption{(a) H\,{\sc i} contour map of the 21-cm observations, superimposed
140: on a 1 square degree negative Digitized Sky Survey image. Contours are from
141: $2.5 \times 10^{19}$ to $2 \times 10^{20}$ cm$^{-2}$ at intervals of $2.5 
142: \times 10^{19}$ cm$^{-2}$.(b) Shows the declination-velocity projection of 
143: the data cube.  More detail can be seen in the online animation of the 
144: datacube (Fig. \ref{animation}).}
145: \label{fig1}
146: \end{figure*}
147: %\clearpage
148: Fig. \ref{fig1}a shows a neutral hydrogen (H\,{\sc i}) contour map of the 
149: field superimposed on a negative optical image.  VIRGOHI 21 is the elongated 
150: structure in the center (which is at about 2000 km\,s$^{-1}$).  A faint bridge 
151: can be seen stretching down to the prominent spiral NGC 4254 
152: (2400 km\,s$^{-1}$) while the other two sources, \object{NGC 4262} 
153: (1350 km\,s$^{-1}$, 
154: upper left) and the faint galaxy `C' (1750 km\,s$^{-1}$, immediately to the 
155: left of VIRGOHI 21) appear unconnected.  Fig. \ref{fig1}b shows the 
156: velocity-declination
157: projection of the full 3 dimensional data cube.  Now NGC 4254 is at the bottom 
158: right while VIRGOHI 21 is the angular structure in the center with `C' to its 
159: left.  Far more detail can be seen in the animation of the data cube, Fig. 
160: \ref{animation} (available in the electronic edition of the {\it Astrophysical 
161: Journal}, only the first frame is shown in the print edition).  The 
162: bridge between VIRGOHI 21 and NGC 4254 is clear, as is the lack of any 
163: connection between VIRGOHI 21 and either NGC 4262 or galaxy `C'.  The apparent 
164: alignment of NGC 4262 with VIRGOHI 21 and NGC 4254 in Fig. \ref{fig1}b is 
165: merely a consequence of the particular projection shown.
166: %\clearpage
167: \begin{figure}
168: \plotone{f2.eps}
169: \caption{This figure is available as a color mpeg animation in the electronic 
170: edition of the {\it Astrophysical Journal}, only the first frame is shown 
171: here.
172: The figure shows an animation of the WSRT data cube,
173: oriented with increasing velocity upwards and
174: rotating around this axis.  VIRGOHI 21 can be seen in the center joined by 
175: a stream of neutral hydrogen to NGC 4254 (above and to the side).  Object 
176: C is below VIRGOHI 21 and NGC 4262 is the ring structure near the bottom.  
177: The vertical column is the continuum source 1214+1456 (White \& 
178: Becker 1992) which was incompletely removed by {\sc uvlin}.  The
179: faint vertical striping is due to incompletely removed continuum from 
180: M87, which is detected in the sidelobes of the array.}
181: \label{animation}
182: \end{figure}
183: %\clearpage
184: The bridge is $\sim$ 25 arcmin long and stretches from 2250 km\,s$^{-1}$
185: at $+14^\circ28^\prime$ at the low-velocity (western) edge of NGC 4254
186: to 1900 km\,s$^{-1}$ at $+14^\circ 41^\prime$.  Here it meets a section 
187: of relatively high column-density H\,{\sc i} which is flat in velocity until 
188: $+14^\circ 46^\prime$ where there is a strong velocity gradient in the
189: opposite direction from that seen in the bridge, rising to 2100 km\,s$^{-1}$ 
190: at $+14^\circ 49^\prime$.
191: 
192: Fig. \ref{fig2} is a blow-up of the source region superimposed on a far 
193: deeper CCD optical image.  The CCD image was made by combining two 
194: $B$-band images: a 750 s Isaac Newton Telescope (INT) Wide Field Camera (WFC)
195: survey image taken during photometric conditions in March 2000 
196: and our own 600 s image taken with the same instrument during non-photometric 
197: conditions in May 2004. Both sets of data were reduced using the Wide Field 
198: Survey 
199: pipeline\footnote{http://www.ast.cam.ac.uk/$\sim$wfcsur/technical/pipeline/} 
200: which includes 
201: de-biasing, bad pixel replacement, non-linearity correction, flat-fielding 
202: and gain correction.  The images were combined using the Starlink task 
203: {\sc MakeMos} in the {\sc CCDPack} package and photometry was taken from the 
204: WFC survey image, which has a photometric accuracy of 0.05 mag.  Combined,
205: the two images have a $1\sigma$ sky noise of 26.7 mag arcsec$^{-2}$ which has 
206: been improved to 27.5 mag arcsec$^{-2}$ by binning into 1 arcsec pixels. This 
207: gives a surface brightness detection limit of about 27.5 mag arcsec$^{-2}$ 
208: for an object of diameter 10 arcsec.  We find no optical counterpart down
209: to this surface brightness and size limit.
210: 
211: There is a small, faint galaxy labeled `A' superposed on the highest
212: H\,{\sc i} contour ($10^{20}$ cm$^{-2}$) at declination 
213: $+14^\circ47.4^\prime$. An optical spectrum from the 6.5-m MMT in Arizona 
214: shows that it is at a redshift of $z=0.25$ and is therefore unconnected 
215: with VIRGOHI 21 (see Fig. \ref{fig3}).  The 17th magnitude galaxy `C' to the 
216: left at $+14^\circ45^\prime$ is an H\,{\sc i} point-source at this resolution. 
217: By comparison, VIRGOHI 21 is an extended structure in both dimensions rather 
218: than a collection of discrete compact clouds.  The velocity-declination plot 
219: (right) shows the complex kinematic structure of VIRGOHI 21.  The most 
220: remarkable feature is the tilted portion between $+14^\circ46^\prime$ 
221: (1900 km\,s$^{-1}$) and $+14^\circ49^\prime$ (2100 km\,s$^{-1}$) which 
222: resembles the signature of an edge-on rotating disk (e.g. Kregel et al. 2004).
223: A tentative detection of gas further to the north (in Fig. 3b) could either be 
224: part of the interaction or clumps further out in the disk. Note the lack of 
225: connection between galaxy `C' and VIRGOHI 21.
226: %\clearpage
227: \begin{figure*}
228: \centering
229: \plottwo{f3a.eps}{f3b.eps}
230: \caption{As Fig. \ref{fig1}, but expanded and superimposed on a negative of 
231: our deep
232: CCD $B$-band image (Minchin et al. 2005) with a surface-brightness 
233: limit of 27.5 $B$ mag. arcsec$^{-2}$.  The hatched ellipse in the bottom-left 
234: corner indicates the WSRT half-power beam.}
235: \label{fig2}
236: \end{figure*}
237: 
238: \begin{figure}
239: \centering
240: \plotone{f4.eps}
241: \caption{Optical spectrum of object `A' taken with the 6.5-m MMT.  The H and 
242: K lines give a redshift of $z=0.2535$.}
243: \label{fig3}
244: \end{figure}
245: 
246: \begin{figure*}
247: \centering
248: \epsscale{.77}
249: \plotone{f5_color.eps}
250: \caption{Channel maps of the VIRGOHI 21 -- NGC 4254 system.  Contours are at
251: $-1$ mJy ($-3\sigma$), 1 mJy ($3\sigma$), 2 mJy ($6\sigma$) and 3 mJy 
252: ($9\sigma$) (colored orange, red, magenta and blue in the electronic 
253: edition).  The beam size is shown by the ellipse in the bottom-left and
254: the velocity of each channel (in km\,s$^{-1}$) is given in the top-left.
255: See the electronic edition of the Journal for a color version of this figure.}
256: \label{fig4}
257: \end{figure*}
258: 
259: \begin{figure*}
260: \centering
261: \rotatebox{270}{\includegraphics[width=0.85\linewidth]{f6_color.eps}}
262: %\plotone{f6_color.eps}
263: \caption{Channel maps of the central region of VIRGOHI 21, contours as in Fig.
264: \ref{fig4}. The cross indicates the peak of the zeroth moment map, which is at 
265: the center of the rotating component (see Fig. \ref{fig2}).  It can clearly be 
266: seen that this component (declination range $14^\circ46^\prime$ to 
267: $14^\circ49^\prime$) starts south of this point and moves northwards across it
268: with increasing velocity (as might be expected for rotation).  See the 
269: electronic edition of the Journal for a color version of this figure.}
270: \label{fig5}
271: \end{figure*}
272: %\clearpage
273: Figs. \ref{fig4} and \ref{fig5} show the channel maps from the WSRT 
274: observations.  Fig. \ref{fig4} shows
275: the whole VIRGOHI 21 -- NGC 4254 system, including the bridge connecting
276: the two which is only a $3\sigma$ detection in the individual channels but,
277: taken over all channels, is definitely present (see Fig. \ref{fig1}b and the 
278: animation of the datacube, Fig. \ref{animation}).  Fig. \ref{fig5} shows 
279: the cube between RA $12^h17^m46^s$ and 
280: $12^h18^m03^s$ over the velocity range of VIRGOHI 21.  It can be seen that the
281: section identified in Fig. \ref{fig2}b as the `Rotating Disc' undergoes a 
282: smooth shift northwards with increasing velocity, which resembles (taking into 
283: account
284: the difference in beam sizes) that observed in edge-on rotating systems 
285: (e.g. Uson \& Matthews 2003; Rupen 1991). Note that this section of VIRGOHI 21
286: is unresolved in individual channels, leading to a broadening of the linewidths
287: down each line of sight to $\sim 100$ km\,s$^{-1}$ and a lowering of the
288: column-density as the H\,{\sc i} is smeared by the beam.
289: 
290: \section{HST Observations}
291: 
292: To place a better limit on the presence of stars associated with VIRGOHI 21 we 
293: have obtained HST data
294: with a total exposure time of 23142 s.  The ACS Wide Field Camera was used 
295: with the $I$-band F814W filter. The pointings were centered on the H\,{\sc
296: i}-flux-weighted position of $12^h17^m53^s, 
297: +14^\circ45^\prime25^{\prime\prime}$ (J2000), 
298: with the ACS field of view matching the minimum H {\sc i} size (3.5
299: arcmin) very closely. Each orbital exposure was {\sc cr-split} in two for 
300: cosmic ray removal and was slightly offset to permit drizzling.
301: 
302: Preliminary data reduction was done by the Space Telescope Science
303: Institute with the standard pipeline. The flat-fielded
304: images have been combined with the {\sc Iraf} task {\sc Multidrizzle} to
305: obtain a distortion-corrected and drizzled final image. The pixel
306: size in the resulting images is 0.05 arcsec.
307: 
308: With this depth and spatial resolution we should be able to
309: resolve individual stars in VIRGOHI 21 down to $I \simeq 28$ and thus to
310: detect the tip of the Red Giant Branch which at the distance of
311: the Virgo Cluster would be close to 27 in $I$. The presence of
312: individual stars would not  conflict with the ground-based INT
313: observations: the $B$-band surface-brightness limit of 27.5 mag
314: would (for $B - I \simeq 2$) equate to around 0.5
315: detectable RGB stars per square arcsecond, or 2500 - 3000 over the
316: area of VIRGOHI 21.
317: 
318: Stellar photometry was performed with the {\sc Daophot} package in {\sc Iraf}
319: (Stetson 1987). A preliminary selection of stars with a threshold
320: of 3.5 sigma was performed with the automatic star-finding
321: algorithm {\sc Daofind} and their magnitudes obtained with the task {\sc Phot}
322: using an aperture radius of 3 pixels. A spatially variable model
323: point-spread function (PSF)  -- with a full width at half maximum
324: of 2 pixels --  was built with the task {\sc Psf} by selecting several
325: isolated and bright stars. Finally, {\sc Allstar} was used to fit the
326: model PSF to the stars in the input list and to produce the final
327: catalog.  The conversion to Johnson {\it I} magnitudes and the
328: aperture correction were done following Sirianni et al. (2005).  
329: Artificial star tests were used to determine the completeness of our
330: selection; the 50\% completeness limit is found to be $I=27.9$ mag.
331: 
332: We found that the photometry catalog produced with
333: {\sc Daophot/Allstar} ($\sim$600 objects) included  bad measurements due to
334: misidentifications such as  parts of bright background galaxies or
335: unresolved background galaxies whose contribution to the
336: luminosity function is significant at fainter magnitudes (see
337: Durrell et al. 2002). Given the long exposures required by such
338: observations we did not ask for additional time to look at another
339: field to determine the background density of star-like objects;
340: instead, we have removed objects from our list by
341: individually examining their radial profiles and rejecting those with a
342: FWHM $\gtrsim$ 2.5 - 3 pixels, corresponding to the value of the
343: PSF built with the brightest stars in the field.  The final list
344: consists of 281 stars and the corresponding luminosity function is
345: shown in Fig. \ref{lum_func}.
346: %\clearpage
347: \begin{figure}
348: \plotone{f7.eps}
349: \caption{The luminosity function of the final
350: list of stars. The dotted line shows the output of the edge
351: detection filter used to detect the tip of the RGB. The maximum of
352: the function indicates the position of the tip. The vertical line
353: at $I = 27.9$ mag indicates the 50\% completeness limit. }
354: \label{lum_func}
355: \end{figure}
356: %\clearpage
357: To estimate the contamination due to foreground stars, we compared
358: the number of detections  with a model of the expected number of
359: Milky Way stars in the position of the Virgo Cluster (Castellani 
360: et al. 2002; Cignoni et al. 2003).  This model shows
361: that the contribution of foreground stars is only significant at
362: brighter magnitudes ($I \leq 26.5$) where the number of stars
363: detected is small (Fig. \ref{lum_func}); at fainter magnitudes the larger
364: number of detections makes the foreground unimportant.
365: 
366: To search for the Tip of the Red Giant Branch (TRGB) we have applied a
367: Sobel filter to the binned luminosity functions.  The luminosity
368: function shows a slope change at $I = 27.2\pm 0.1$ mag, 
369: which we identify as the possible tip of the red giant branch.  This would 
370: place these stars at a distance of 16.6 - 18.2 Mpc, which is consistent with 
371: them being in the Virgo Cluster (d=16 Mpc), though the statistics are poor due 
372: to the relatively low numbers of stars in each magnitude bin. If  we assume 
373: this TRGB to be real, despite the low signal to noise ratio (around 2$\sigma$
374: on the Sobel-filter output), we can set the most stringent limit on the 
375: surface-brightness of VIRGOHI 21.
376: 
377: Assuming (very generously) that all of the fainter stars are also at the 
378: distance of the cluster 
379: (119 with $I = 27.2 - 27.9$) they have a combined magnitude
380: of $I = 22.3$, giving a surface brightness of 33.8 $I$
381: mag arcsec$^{-2}$. Using the formula of Durrell et al. (2002) gives a
382: correction for only having the top 0.7 mags of the RGB of $1.5 \pm 0.1$
383: mag arcsec$^{-2}$.  We do not detect any Asymptotic Giant
384: Branch stars; based on the work of Durrell et al. these could contribute
385: another $0.2 \pm 0.1$ mag arcsec$^{-2}$.  Overall, this gives
386: an upper limit for the total $I$-band surface brightness of 
387: $31.1 \pm$ 0.2 mag arcsec$^{-2}$. For normal galaxy
388: colors of $B-I = 1.5 - 2$, this equates to an extremely faint $B$-band surface
389: brightness of 32.4 -- 33.3 mag arcsec$^{-2}$.  
390: 
391: These results can be compared with studies of Virgo Cluster inter-galactic 
392: stars (Durrell et al. 2002) and those used to determine the 
393: color-magnitude diagrams of
394: dwarf galaxies belonging to the cluster (Caldwell 2006). Durrell et al.
395: imaged a field at about 40$^\prime$ northwest of M 87 ($\approx$190 kpc)
396: while the two fields observed by Caldwell were also relatively close to the
397: center of Virgo, one at $1.4^{\circ}$ north ($\approx$400 kpc)
398: and  the other at $1.1^{\circ}$ west ($\approx$315 kpc). These fields are much 
399: closer to M 87 at the center of the cluster
400: than VIRGOHI 21 and the number density of stars found is also much higher 
401: however, there is an apparent decrease in the number of stars 
402: as one moves away from the cluster center (see Fig. 9 in Caldwell 2006 and Fig.
403: 6 in Durrell et al. 2002).  A comparison between the estimated
404: surface brightnesses of the three fields shows that $\mu_I$ ranges
405: from 27.7 $I$ mag arcsec$^{-2}$ (Durrell et al. 2002) for the
406: fields closer to the cluster center to $\mu_I = $ 27.9 and 29.9 $I$ mag 
407: arcsec$^{-2}$ (Caldwell 2006) for the more distant ones.  Our $I$ band
408: surface brightness of 31.3, at a projected distance of 1.1 Mpc from the
409: cluster center (M87), is thus consistent with the known population 
410: intra-cluster stars in
411: Virgo and so there is no evidence to suggest that our detections are 
412: associated with VIRGOHI 21 in any way.
413: 
414: \section{Discussion}
415: 
416: In the previous sections we have described our new observations of VIRGOHI 21. 
417: In this section we try and put them in context and particularly we want to 
418: assess the evidence for and against the dark galaxy hypothesis.
419: \begin{enumerate}
420: \item If attributed to gravitation, changes in velocity of galactic size over
421: galactic scales, as seen here, require masses of galactic proportions.  On
422: dimensional grounds the abrupt change in velocity $\Delta V$ ($\sim 200$ 
423: km\,s$^{-1}$)
424: seen in VIRGOHI 21 over a conservative length-scale $\Delta x$ ($\sim 14$ kpc 
425: $\simeq 5\times 10^{22}$ cm $\simeq 3$ arcmin at the Virgo Cluster distance
426: of 16 Mpc)
427: implies a mass $M \geq (\Delta V)^2 \Delta x/G \simeq 10^{10-11}$ solar masses 
428: if this is due to an object rotating in dynamical equilibrium. If the 
429: velocities are not due to rotation then changes in velocity like this can be 
430: observed if, for example, we are observing an arc of gas and our view is almost
431:  edge-on. These geometrical effects can lead to large velocity widths that may 
432: resemble rotation (e.g. Bournaud et al. 2004). The problem with this latter 
433: scenario (as also pointed out by Vollmer et al. (2005) is that the perturbing 
434: galaxy should be moving in the plane of the arc and so should be projected on 
435: the sky very close to it. We can simply quantify what we mean by this. Imagine 
436: two galaxies with radial velocities, $V_1$ and $V_2$ at either end
437: of an approximately linear tidal bridge of physical length $d$ pitched at an 
438: angle $\theta$ to the line of sight.  A telescope pointed towards it has a 
439: transverse beam diameter of $b$ at the bridge.  The only significant
440: gas motions within the bridge will be streaming velocities along its length.
441: From end to end of the bridge the radial velocity difference is $|V_2 - V_1|$
442: while within the telescope beam the measured velocity width, $\Delta V$, will 
443: be ($b/d\sin\theta)\times|V_2 - V_1|$.  But, bridges of any appreciable
444: size arise only when the total velocity difference between the 
445: interacting galaxies i.e. $|V_2 - V_1|/\cos\theta$, is of order 
446: the circular velocity $V_c$ in the galaxies involved (Toomre \& Toomre 1972).
447: It follows that $\Delta V/V_c \simeq (b/d)\cot\theta$.  Thus broad
448: line widths $\Delta V \simeq V_c$, as here, can only be seen within a beam if
449: both interactors appear to lie within, or very close to the beam
450: i.e. $d\sin\theta \leq b\cos\theta \leq b$.  We can find no potential 
451: perturber within the 
452: Arecibo beam ($b\simeq 3\farcm6$) or close to it. Thus we do not believe
453: VIRGOHI 21 can be tidal debris from an interaction of this nature. 
454: 
455: \item The existence of the H\,{\sc i} bridge indicates that NGC 4254 has 
456: undergone some kind of interaction. 
457: NGC 4254 is a luminous one-armed spiral galaxy 
458: sufficiently peculiar to have attracted several studies 
459: (Iye et al. 1982; Phookun et al. 1993; Vollmer et al. 
460: 2005). According to Vollmer et al. 2005) 
461: the morphological peculiarities of NGC 4254 can be explained by an
462: interaction with a perturbing mass of $\sim 10^{11} M_\odot$.  As it is now 
463: clear
464: that VIRGOHI 21 is involved in this interaction, it should be considered a 
465: candidate for the perturber, particularly as no other candidate can be
466: easily identified.    The 
467: Vollmer et al. simulation implies an 
468: interaction $3\times 10^8$ years ago.  As the projected length of the bridge
469: is 120 kpc this would imply it has been drawn out at a projected speed of 390 
470: km\,s$^{-1}$, which is comparable to the radial velocity difference 
471: between NGC 4254 and VIRGOHI 21 of 400 km\,s$^{-1}$. Thus if VIRGOHI 21 were 
472: massive enough ($\sim 10^{11}$ $M_{\odot}$), it could be the cause of the 
473: tidal bridge.
474: 
475: \item If object C, the H\,{\sc i} galaxy just to the east (left) of VIRGOHI 21 
476: (Fig. \ref{fig2}) were involved its mass (see (a) above) should be 
477: $\sim 10^{11} 
478: M_\odot$.  However, 
479: its measured H\,{\sc i} velocity distribution and size suggest a probable mass 
480: $\leq 10^9 M_\odot$, while it has a luminosity of only $10^8 L_\odot$ and a 
481: $M_{HI}$ of $2\times 10^7 M_\odot$.  Galaxy `C' appears to be two orders of 
482: magnitude too 
483: undermassive and underluminous to explain the one-armed spiral structure of 
484: NGC 4254 and the H\,{\sc i} bridge connecting to it. 
485:  So, it seems very unlikely that `C' has sufficient
486: mass to cause a perturbation of over 200 km\,s$^{-1}$ to the tidal stream,
487: and VIRGOHI 21 passes to its west, not between it and NGC 4262 as might
488: be expected if it had pulled the stream westward.  The radial velocity
489: of `C' means that it must be moving past the stream at a velocity
490: (relative to the putative undisturbed velocity of the stream)
491: of at least 350 km\,s$^{-1}$.  At this speed, it would not have stayed close
492: to the stream long enough to have severely disturbed it.
493: It would also be expected
494: that if `C' were involved in the interaction then there would be gas
495: falling onto it, but this does not appear to be the case as there is no gas
496: seen between VIRGOHI 21 and C.  The H\,{\sc i} in `C' is unresolved, implying
497: that it is confined to the area of the optical galaxy; it shows no signs
498: of infalling gas streams or any other disturbance.  Galaxy `C' is in our a 
499: minds a possible, but highly unlikely participant in this interaction.
500: 
501: \item Recently Bekki et al. (2005) have published numerical simulations on 
502: the basis of which they argue that ``VIRGOHI 21 \ldots\ is likely to be tidal 
503: debris rather than a `dark galaxy'\,''. They 
504: model high-speed, hyperbolic interactions within cluster potentials (at 350 
505: and 700 kpc from the cluster center) over 2 Gyr in order 
506: to draw out very long tidal tails where
507: the H\,{\sc i} and stars have become separated so that the average surface
508: brightness within a 100 kpc$^2$ box is $> 30 B$ mag arcsec$^{-2}$ ($L_B <
509: 6.9\times 10^8 L_\odot$), assuming $M_\star/L_B = 4$.
510: The Bekki et al. stellar density can be converted to $I$-band if we assume
511: a value for $M_\star/L_I$.  Estimates of this vary (e.g. Portinari et al. 2004;
512: Vallejo et al. 2002; McGaugh \& de Blok 1997) but are generally close to 1 and 
513: always less than 2.  We therefore adopt a very conservative (in the sense of
514: producing the lowest $I$-band surface-brightness) value of $M_\star/L_I = 2$.  
515: This gives a value of $B-I$ 
516: for the surface-brightness estimates given in Bekki et al. of 0.75.  For
517: the two double-tailed models presented (M1 and M2) this gives average
518: $I$-band surface-brightnesses of 32.7 and 33.0 mag arcsec$^{-2}$, while for
519: the two single-tailed models presented (M3 and M4) this gives average 
520: $I$-band surface-brightnesses of 29.5 and 29.9 mag arcsec$^{-2}$.  The 
521: two-tailed models are inconsistent with our H\,{\sc i} data and the H\,{\sc i}
522: data of Phookun et al., which clearly show a single tail from NGC 4254, while
523: the single-tailed models give considerably higher surface-brightnesses (1.2 --
524: 1.6 mags, or 6 -- 8 $\sigma$) than the upper limit from our HST observations.
525: In addition to this, Bekki et al. find that the velocity fields produced in 
526: the dark clouds by their models
527: do not resemble rotation, concluding that ``Velocity fields are the key 
528: observational tools to help us determine whether gas clouds are (unbound) tidal
529: debris or are self-gravitating systems embedded within a massive dark matter
530: halo.''  From both the surface-density of stars expected in a single-tail 
531: interaction and from the rotation curve, it appears that high-speed, 
532: hyperbolic interactions within a cluster potential like those proposed by 
533: Bekki et al. are unable to reproduce our new observations of VIRGOHI 21.
534: 
535: \item NGC 4262, the spiral to the North-East, does not appear to be involved. 
536: This is not surprising because it is in totally the wrong place to give 
537: rise to a tidal feature like that observed (NGC 4262, NGC 4254 and VIRGOHI 21 
538: form a well separted triangle on the sky, see also Vollmer et al. 2005). Also,
539: the radial velocity between NGC 4254 and NGC 4262 is far
540: too large (900 km\,s$^{-1}$) to generate tidal features such as bridges and
541: tails.  Furthermore, with such a large  radial 
542: velocity difference between NGC 4254 and NGC 4262, any 
543: interaction must be at high speed and, as seen above from the
544: simulations of Bekki et al., such interactions are not capable of reproducing 
545: our observations as even the most isolated gas clouds formed in single-tail 
546: systems have surface-brightnesses 1.5 - 2 magnitudes brighter than the limit 
547: set by our HST observations.  NGC 4262 is a S0 galaxy with an 
548: H\,{\sc i} ring that is inclined with respect to the optical disk (Krumm et al.
549: 1985).  Krumm et al. present higher-resolution H\,{\sc i} observations than
550: our data and they say that if the H\,{\sc i} has come from an interaction, 
551: ``the narrowness of the ring and the undisturbed appearance of the optical
552: disk suggest that the donor was a gas-rich dwarf galaxy or intergalactic
553: H\,{\sc i} cloud, rather than a massive spiral''. Thus we find it very 
554: difficult to see how NGC 4262 can have given rise to the features we have now 
555: observed.
556: 
557: \item The components of VIRGOHI 21 are connected both spatially and in
558: velocity, making it exceedingly improbable that they could be chance
559: superpositions of smaller hydrogen clouds, while the bridge to NGC 4254 is not 
560: explained by this hypothesis.
561: 
562: \item Phookun et al. (1993) suggested that the distortion of NGC 4254 
563: could be due to infalling gas-clouds.  Could VIRGOHI 21 be a tail, similar to 
564: those seen in \object{UGC 10214} (Briggs et al. 2001) or 
565: \object{NGC 4038}\object[NGC 4039]{/9} (Gordon et al. 2001)?  In
566: NGC 4038/9, in particular, there is what appears to be a tidal dwarf  near 
567: a strong concentration of H\,{\sc i} at the end of the southern tail, around 
568: 60 kpc from the center of the system (for a distance of 13.8 
569: Mpc; Saviane et al. 2004).  However, the major problem with this hypothesis
570:  is that NGC 4254 apparently shows
571: no sign of having undergone a recent merger.  It would be expected that
572: a merger violent enough to have thrown out an arm over 100 kpc in length
573: would leave some imprint on the optical disk, but this is not the case here.
574: Phookun et al. found that ``The galaxy is undergoing an interaction that is
575: not so strong or violent as to disrupt the disk.''  While this fits with
576: the hypothesis of VIRGOHI 21 as a dark galaxy, playing a role similar to 
577: that of the companions of \object{NGC 4027} or \object{NGC 4654} in 
578: exciting the $m=1$ mode, it does not fit the merger scenario.
579: 
580: \item Oosterloo \& van Gorkom (2005) argue that another H\,{\sc i} 
581: cloud in the Virgo Cluster, \object{VIRGOHI 4} (Davies et al. 
582: 2004), is caused by ram-pressure stripping from \object{NGC 4388} 
583: due to an interaction with the hot-gas halo of the \object{M86} sub-group and 
584: suggest a similar origin could
585: be possible for VIRGOHI 21.  This would explain the bridge without the need
586: to invoke a second galaxy, either interacting or merging.  However, we cannot 
587: see how this can
588: give the steep, reversed velocity gradient seen in VIRGOHI 21, nor does 
589: it explain the distortion to the optical disk of NGC 4254. 
590: 
591: \item Single-armed  spirals are normally the result of interactions with 
592: close-by massive companions (Iye et al. 1982; Phookun et al. 1993);
593: the lack of any visible companion to NGC 4254 has thus led to a number of
594: observations and dynamical models.  As mentioned above, recent numerical 
595: models 
596: by Vollmer et al. (2005) suggest that NGC 4254 ``had a close and 
597: rapid encounter
598: with a $10^{11} M_\odot$ galaxy $\sim 250$ Myr ago.  The tidal interaction 
599: caused the spiral structure\ldots''.  Phookun et al. (1993), in a VLA 
600: study of the galaxy, find a trail of gas leading away from it (their Figure 5)
601: which can be identified from its direction and velocity-gradient as the
602: NGC 4254 end of the
603: `bridge' between the two objects.  This supports our  inference that VIRGOHI 21
604: is the aforesaid $\sim 10^{11} M_\odot$ mass which caused the peculiarities in 
605: NGC 4254 - particularly so given the lack of any other candidate.
606: 
607: \end{enumerate} 
608: 
609: If our hypothesis is correct that this is a dark, gravitationally bound, 
610: edge-on rotating disk
611: then its properties are as presented in Table \ref{properties}.  Judging from 
612: visible disk galaxies, whose
613: masses continue to rise beyond their H\,{\sc i} edges (e.g. Salucci \& Persic
614: 1997), the 
615: full size and mass of such a disk could easily reach the $\sim 10^{11} M_\odot$
616: required by the Vollmer et al. (2005) simulations.
617: The very low surface-brightness limits give an upper limit to the luminosity of
618: the disk of $1.9 \times 10^5 L_\odot$, giving a lower limit for $M_{dyn}/L_B$ 
619: of at least $10^6 M_\odot/L_\odot$, whereas normal galaxies have values
620: $< 50$.
621: %\clearpage
622: \begin{table}
623: \caption{Properties of the putative dark disk}
624: \label{properties}
625: \begin{tabular}{ll}
626: Diameter $2R$, [from $+14^\circ 46^\prime$ to $49^\prime$] & 14 kpc\\
627: Circular Velocity $V_c$, [$(2100 - 1900)/2$] & 100 km\,s$^{-1}$\\
628: Spin Period $P$, [$2\pi R/V_c$] & $4\times 10^8$ years\\
629: Total Mass $M_T$, [$RV_c^2/G$] & $2\times 10^{10} M_\odot$\\
630: Face-on Mass-density [$M_T/\pi R^2$] & $2\times 10^{-2}$ g cm$^{-2}$\\
631: Hydrogen Mass $M_{HI}$, [$F_{HI} = 0.6$ Jy\,km\,s$^{-1}$]& $3 \times 10^7
632: M_\odot$\\
633: Face-on gas density $N_{HI}$, [$M_{HI}/\pi m_{H}R^2$] & $3\times 10^{19}$
634: cm$^{-2}$\\
635: Luminosity limit $L_{B}$, [$\mu_B > 32.8$ mag arcsec$^{-2}$]&$<1.9\times 
636: 10^5 L_\odot$\\
637: Total Mass to Blue Light Ratio $M_{dyn}/L_B$& $> 10^6 M_\odot/L_\odot$\\
638: \multicolumn{2}{l}{[assuming a disk 0.5 arcmin $\times$ 3 arcmin]}
639: \end{tabular}
640: \end{table}
641: %\clearpage
642: 
643: %\clearpage
644: \begin{table}
645: \caption{H\,{\sc i} fluxes and masses of the detected components}
646: \label{components}
647: \begin{tabular}{llll}
648: &$F_{HI}$ (Jy km\,s$^{-1}$)&$M_{HI}$ ($M_\odot$)\\
649: NGC 4254&91.5&$5.5\times 10^{9}$\\
650: NGC 4262&7.7&$4.6\times 10^{8}$\\
651: Object C&0.3&$2\times 10^{7}$\\
652: `Rotating Disc'&0.6&$3\times 10^{7}$\\
653: ($14^\circ46^\prime<\delta<14^\circ49^\prime$)\\
654: Low col. density bridge&1.3&$8\times 10^7$\\
655: ($\delta<14^\circ41^\prime$)\\
656: High col. density bridge&0.8&$5\times 10^{7}$\\
657: ($14^\circ41^\prime<\delta<14^\circ46^\prime$)
658: \end{tabular}
659: \end{table}
660: %\clearpage
661: 
662: It is clear from our new observations that VIRGOHI 21 is intimately linked to 
663: NGC 4254. The important question is whether it is the cause of the HI bridge 
664: or the effect of some previous encounter. We believe the weight of the evidence
665: presented above lies on the side of a dark galaxy.  Dark halos have, after
666: all, been predicted by galaxy evolution simulations such as Jimenez et al. 
667: (1997), Verde et al. (2002) and Davies et al. (2006).  That none have been 
668: confirmed so far is most probably due to instrumental limitations. VIRGOHI 21 
669: itself was missed in the normal HIPASS-sensitivity observations of this area, 
670: and only picked up with ten times longer integrations in the  
671: VIRGOHI survey. In any case single-dish angular resolution makes it all too 
672: easy to confuse a dark cloud with a bright object clustered with it (Davies et 
673: al. 2006).  The recent optical identification program for HIPASS 
674: sources (Doyle et al. 2005) can rule out only `isolated', i.e. 
675: unclustered, dark galaxies.
676: 
677: If we accept the total mass given in Table 1 then an oddity about VIRGOHI 21 is
678: its shortage of baryons
679: compared to visible galaxies. Visible disk galaxies typically have,
680: in the form of stars and gas, $\sim 10$ percent of their dynamical mass
681: in baryons, a figure consistent with cosmological models which predict
682: $\sim 7$ times as much Dark Matter as baryonic in the Universe at large
683: (Salucci \& Persic 1997).  Had VIRGOHI 21 ten percent of its 
684: mass in baryons it would contain
685: $\sim 2\times 10^9$ solar masses which would, if it were once in the form
686: of H\,{\sc i} spread evenly across the existing dynamical disk, yield a
687: column density $\sim 10^{21}$ atoms cm$^{-2}$ -- sufficiently above
688: the ``star-formation threshold'' of $\sim 10^{20}$ cm$^{-2}$ (Martin \&
689: Kennicutt 2001),
690: or above the Toomre criterion for gravitational instability in a rotating 
691: disk (Toomre 1964), to form stars (Jimenez et al. 1997;
692: Verde et al. 2002).  Since it apparently contains no stars it must have a 
693: minimum of ten times less baryons - a phenomenon often 
694: apparent in dwarf galaxies (Mateo 1998).  Thus
695: 90 percent of the baryons one might have expected to find in a typical galaxy
696: appear to be missing and so it is not difficult to see how a galaxy like this 
697: may remain dark.
698: %\newpage
699: 
700: %\ \\
701: 
702: \section{Summary}
703: In this paper we have presented new Westerbork high resolution 21-cm 
704: observations of a HI source (VIRGOHI 21) that we have previously observed at 
705: Jodrell Bank and Arecibo.
706: These observations clearly show that VIRGOHI 21 has played a part in some form 
707: of interaction with the bright spiral galaxy NGC 4254. Whether VIRGOHI 21 is 
708: the result of some previous interaction with a third object or it is the 
709: interacting object itself is still open to conjecture. We have argued in this 
710: paper that the weight of evidence now resides on the side of the latter rather 
711: than the former conjecture. This is primarily because the new observations 
712: show a long drawn out tidal feature which has an abrupt velocity change and 
713: we cannot find a candidate galaxy to have caused it. Tidal interactions between
714: optically bright galaxies often draw out stars as well as gas. We have also 
715: used HST to search for stars associated with VIRGOHI 21 and find none down to 
716: very faint surface brightness levels. We look forward to new observations and 
717: numerical models that will add further weight to either side of this debate. 
718: 
719: 
720: \acknowledgements
721: We thank the Netherlands Foundation for Radio Astronomy for use of the
722: WSRT and Rafaella Morganti and Tom Oosterloo for assistance in reducing the
723: WSRT data, the National Astronomy and Ionosphere Center for use of the 
724: Arecibo Telescope, the UK Particle Physics and Astronomy Research Council for 
725: financial support, and the Australia Telescope National Facility for 
726: inspiration and technical support.  We would like to thank the following for
727: useful comments: Virginia Kilborn, Erwin de Blok, Martin Zwaan and Greg 
728: Bothun.  We also thank the anonymous referee for comments that helped improve
729: this paper.  The Digitized 
730: Sky Survey was produced at the Space Telescope Science Institute under U.S. 
731: Government grant NAG W-2166. The images of these surveys are based on 
732: photographic data obtained using the Oschin Schmidt Telescope on Palomar 
733: Mountain and the UK Schmidt Telescope. The plates were processed into the 
734: present compressed digital form with the permission of these institutions.
735: 
736: 
737: \begin{thebibliography}{}
738: 
739: \bibitem[2005]{Bekki05} Bekki K., 
740: Koribalski B. S., Kilborn V. A., 2005, MNRAS, 363, L21
741: 
742: \bibitem[2001]{Boyce01} Boyce P. J. et
743: al., 2001, ApJ, 560, L127
744: 
745: \bibitem[2004]{Bournaud04} Bournaud F., Duc P.-A., Amram P., Combes F., Gach
746: J.-L., 2004, A\&A, 425, 813
747: 
748: \bibitem[1995]{Briggs95} Briggs D. S., 1995, 
749: BAAS, 27, 1444
750: 
751: \bibitem[2001]{Briggs01} Briggs 
752: F. H., M\"oller O., Higdon J. L., Trentham N., Ramirez-Ruiz E., 2001, 
753: A\&A, 380,418
754: 
755: \bibitem[2006]{Caldwell06} Caldwell N., 2006, ApJ, 651, 822
756: 
757: \bibitem[2002]{Castellani02} Castellani V., Cignoni M., Degl'Innocenti S. et 
758: al., 2002, MNRAS, 334, 69.
759: 
760: \bibitem[2003]{Cignoni03} Cignoni M., Prada Moroni P. G., Degl'Innocenti S.,
761: 2003, MSAIS, 3, 143
762:  
763: \bibitem[2005]{Doyle05} Doyle J. et al., 2005, MNRAS, 361, 34
764: 
765: \bibitem[2004]{Davies04} Davies J., Minchin R., Sabatini S., van Driel W., Baes
766: M., Boyce P., de Blok W. J. G., Disney M., Evans Rh., Kilborn V., Lang R., 
767: Linder S., Roberts S., Smith R., 2004, MNRAS, 349, 922
768: 
769: \bibitem[2006]{Davies06} Davies J. I, Disney M. J., Minchin R. F., Auld
770: R., Smith R., 2006, MNRAS, 368, 1479
771: 
772: \bibitem[2002]{Durrelll02} Durrell P. R., Ciardullo R., Feldmeier J. J., 
773: Jacoby G, H., Sigurdsson S., 2002, ApJ, 570, 119
774: 
775: \bibitem[2001]{Gordon01} Gordon S., Koribalski B., Jones K., 2001, MNRAS, 326, 
776: 578
777: 
778: \bibitem[1997]{Hawkins97} Hawkins M. R. S.,
779: 1997, A\&A, 328, 25
780: 
781: \bibitem[1982]{Iye82} Iye M., Okamura S., Watanabe M., 1982, ApJ, 256, 103
782: 
783: \bibitem[1997]{Jimenez97} Jimenez R.,
784: Heavens A. F., Hawkins M. R. S., Padoan P., 1997, MNRAS, 292, L5
785: 
786: \bibitem[2000]{Kilborn00} Kilborn 
787: V. A. et al., 2000, ApJ, 120, 1342
788: 
789: \bibitem[2004]{Kregel04} Kregel M., van der Kruit P., Blok W., 2004, MNRAS, 
790: 352, 787
791: 
792: \bibitem[1985]{Krumm85} Krumm N., van Driel W., van Woerden H., 1985, A\&A, 
793: 144, 202
794: 
795: \bibitem[1997]{McGaugh97} McGaugh S. S., de Blok W. J. G., 1997, ApJ, 481, 689
796: 
797: \bibitem[2001]{Martin01} Martin 
798: C. L., Kennicutt R. C. Jr., 2001, ApJ, 555, 301
799: 
800: \bibitem[1998]{Mateo98} Mateo M. L., 1998, 
801: ARA\&A, 36, 435
802: 
803: \bibitem[2005]{Minchin05} Minchin R.,
804: Davies J., Disney M., Boyce P., Garcia D., Jordan C., Kilborn V., Lang 
805: R., Roberts S., Sabatini S., van Driel W., 2005, ApJ, 622, L21
806: 
807: \bibitem[2005]{Oosterloo05} 
808: Oosterloo T., van Gorkom, J., 2005, A\&A, 437, 190
809: 
810: \bibitem[1993]{Phookun93} Phookun B., Vogel S. N., Mundy L. G., 1993, ApJ, 
811: 418, 113
812: 
813: \bibitem[2004]{Portinari04} Portinari L., Sommer-Larsen J., Tantalo R., 2004,
814: MNRAS, 347, 691
815: 
816: \bibitem[1991]{Rupen91} Rupen M. P., 1991, AJ, 102, 48
817: 
818: \bibitem[2001]{Ryder01} Ryder S. D. 
819: et al., 2001, ApJ, 555, 232
820: 
821: \bibitem[1997]{Salucci97} Salucci P. 
822: \& Persic M., 1997, in Persic M., Salucci P., eds, ASP Conf. Ser. 117,
823: Dark and Visible Matter in Galaxies, Astron. Soc. Pacific, San 
824: Francisco, p. 1
825: 
826: \bibitem[2004]{Saviane04} Saviane I., Hibbard J. E., Rich R. M., 2004, AJ, 
827: 127, 660
828: 
829: \bibitem[1983]{Schneider83} 
830: Schneider S. E., Helou G., Salpeter E. E., Terzian Y., 1983, ApJ, 273, L1
831: 
832: \bibitem[2006]{Simon06}
833: Simon J. D., Blitz L., Cole A. A., Weinberg M. D., Cohem M., 2006, ApJ, 640,
834: 270
835: 
836: \bibitem[2005]{Sirianni05}
837: Sirianni M. et al., 2005, PASP, 117, 1049
838: 
839: \bibitem[2005]{Spekkens05} Spekkens K., Momjian E., Kent B. R., Giovanelli R.,
840: Haynes M. P., Catinella B., Stierwalt S., Saintonge A., 2005, AAS, 207, 179.25
841: 
842: \bibitem[1987]{Stetson87}
843: Stetson P. B., 1987, PASP, 99, 191
844: 
845: \bibitem[2002]{Robishaw02} Robishaw T., Simon J. and Blitz L., 2002, ApJ, 580, 
846: 129
847: 
848: \bibitem[2005]{Taylor05} Taylor E. N., Webster R. L., 2005, ApJ, 634, 1067
849: 
850: \bibitem[1964]{Toomre64} Toomre A., 1964,
851: ApJ, 139, 1217
852: 
853: \bibitem[1972]{Toomre72} Toomre A., 
854: Toomre J., 1972, ApJ, 178, 623
855: 
856: \bibitem[2003]{Uson03} Uson J. M., Matthews L. D., 2003, AJ, 125, 2455
857: 
858: \bibitem[2002]{Vallejo02} Vallejo O., Braine J., Baudry A., A\&A, 2002, 387, 
859: 429
860: 
861: \bibitem[2002]{Verde02} Verde L., Oh S. P., Jimenez R., 2002, MNRAS, 336, 541
862: 
863: \bibitem[2005]{Vollmer05} Vollmer B., Huchtmeier W., van Driel W., 2005, 
864: A\&A, 439, 921
865: 
866: \bibitem[1997]{Wakker97} Wakker, B. P., van Woerden H., 1997, ARA\&A, 35, 217
867: 
868: \bibitem[1992]{White92} White, R. L., Becker, R. H., 1992, ApJS, 79, 331
869: 
870: 
871: \end{thebibliography}
872: 
873: \end{document}
874: