1: %% The command below calls the preprint style

2: %% which will produce a one-column, single-spaced document.

3: %% Examples of commands for other substyles follow. Use

4: %% whichever is most appropriate for your purposes.

5: %%

6: %%\documentclass[12pt,preprint]{aastex}

7: 

8: %% manuscript produces a one-column, double-spaced document:

9: 

10: %%\documentclass[manuscript]{aastex}

11: 

12: %% preprint2 produces a double-column, single-spaced document:

13: 

14: %%\documentclass[preprint2]{aastex}

15: %%\usepackage{graphicx}

16: 

17: \documentclass[preprint2]{emulateapj}

18: %%\usepackage{graphics}

19: \usepackage{graphicx}

20: 

21: 

22: \shorttitle{Bar imprints on M33 kinematics}

23: \shortauthors{Corbelli and Walterbos}

24: 

25: 

26: \begin{document}

27: 

28: 

29: \title{Bar imprints on the inner gas kinematics of M33}

30: 

31: %% Use \author, \affil, and the \and command to format

32: %% author and affiliation information.

33: %% Note that \email has replaced the old \authoremail command

34: %% from AASTeX v4.0. You can use \email to mark an email address

35: %% anywhere in the paper, not just in the front matter.

36: %% As in the title, use \\ to force line breaks.

37: 

38: \author{Edvige Corbelli}

39: \affil{INAF-Osservatorio Astrofisico di Arcetri, Largo E.Fermi, 5, 50125

40: Firenze, ITALY}

41: \email{edvige@arcetri.astro.it}

42: 

43: \and

44: 

45: \author{Rene' A. M. Walterbos}

46: \affil{Department of Astronomy, New Mexico State University, P.O. 

47: Box 30001, MSC 4500, Las Cruces, NM 88003}

48: \email{rwalterb@nmsu.edu}

49: 

50: \begin{abstract}

51: 

52: We present measurements of the stellar and gaseous velocities in the

53: central 5$'$ of the Local Group spiral M33. The data were obtained

54: with the ARC 3.5m telescope. Blue and red spectra with resolutions

55: from 2 to 4\AA\ covering the principal gaseous emission and stellar

56: absorption lines were obtained along the major and minor axes and six

57: other position angles. The observed radial velocities of the ionized

58: gas along the photometric major axis of M33 remain flat 

59: at ~22 km~s$^{-1}$

60: all the way into the center, while the stellar velocities show a

61: gradual rise from zero to 22 km~s$^{-1}$ over that same region.  The

62: central star cluster is at or very close to the dynamical center, with

63: a velocity that is in accordance with M33's systemic velocity to

64: within our uncertainties.  Velocities on the minor axis are non-zero

65: out to about 1$'$ from the center in both the stars and gas. Together

66: with the major axis velocities, they point at significant deviations

67: from circular rotation. The most likely explanation

68: for the bulk of the velocity patterns are streaming motions along a

69: weak inner bar with a PA close to that of the minor axis, as suggested

70: by previously published IR photometric images. The presence of bar imprints

71: in M33 implies that all major Local Group galaxies are barred. The

72: non-circular motions over the inner 200 pc make it difficult to

73: constrain the shape of M33's inner dark matter halo profile. If the

74: non-circular motions we find in this nearby Sc galaxy are present in

75: other more distant late-type galaxies, they might be difficult to

76: recognize.

77: 

78: \end{abstract}

79: 

80: 

81: 

82: \keywords{galaxies:individual(M33)-galaxies:kinematics and dynamics

83: -galaxies:structure,dark matter}

84: 

85: \section{Introduction}

86: 

87: 

88: The central regions of late type spiral galaxies are still quite

89: puzzling places. Often they show a young, massive, compact star

90: cluster and no clear evidence of a stellar bulge or bar

91: \citep[e.g.][]{bok02} leaving open the question of a fueling 

92: mechanism for star formation in their shallow central potential.  

93: The central region of the Local Group galaxy M33 is the posterchild 

94: of such an environment. It hosts a bright blue

95: compact cluster at its center but, unlike other bright 

96: Local Group members, no definitive evidence 

97: of a bar or bulge has been found yet. Being

98: nearby \citep[840~kpc,$1''$=4~pc,][]{fre01} 

99: and at favorable inclination \cite[52$^\circ$,][]{cor00}, M33

100: allows however a detailed look at the spatial distributions and 

101: kinematics of various components in the central region. 

102: 

103: 

104: The mass of the blue compact nuclear cluster in M33 has been estimated

105: to be of order $10^6$~M$_\odot$ \citep{kor93,lau98}. Its small

106: mass-to-light ratio is indicative of a young population of stars (age

107: $\sim 10^7$-$10^8$~yr) concentrated in the compact core

108: \citep{kor93}. An age spread is likely to be present across the

109: cluster, due to a possible sequence of accretion events. However,

110: there are models which predict a single-age stellar population by taking

111: into account the role of dust in processing the UV/optical stellar 

112: spectrum \citep{gor99}.  From the theoretical point of view

113: it is unclear whether compact nuclear clusters are born and manage to

114: survive at the dynamic center or they develop non-steady

115: motions around the mass centroid or migrate towards the center and

116: eventually get disrupted as they approach it

117: \citep{mil92,por03,mil04}. In the latter cases the dynamic center

118: would not coincide with the position of the bright cluster, but given

119: the shallow potential of M33 \citep{vdm06} it is unclear that this 

120: has happened for this galaxy.  There has been no accurate and

121: conclusive spectroscopic determination of the dynamical center via

122: stellar absorption lines or emission line gas mapping of the central

123: 1~kpc region of M33. Rubin \& Ford 1980 noticed a possible small

124: displacement of the dynamic symmetry center with respect to the

125: location of the compact nuclear cluster.  Locating the dynamical

126: center of M33 has been one of the goals of our project even though we

127: shall discuss in this paper the difficulties of reaching a definite

128: conclusion on this issue.  With the nuclear cluster light taken out,

129: the photometric center of M33 is difficult to identify because of the

130: presence of dust.  M33 lacks a supermassive black hole at the center,

131: the upper limit to the mass being as low as 1500~M$_\odot$

132: \citep{geb01,mer01}.  Given the shallow potential well of the center

133: of M33 this is not a surprise although the downward extrapolation of the

134: relationship between stellar velocity dispersion and supermassive

135: black hole mass predicts a black hole mass for the cluster that is

136: well above the quoted limit.

137: 

138: Understanding the innermost region of M33 is important also for

139: possible constraints to cosmological models of structure formation, in

140: particular to distinguish cuspy dark matter halo models from

141: isothermal core models.  In a recent determination of the rotation

142: curve from the CO J=1-0 line data at $R>0.5$~kpc, Corbelli (2003)

143: discusses cosmological models compatible with the data and finds that

144: an additional mass component is required in the innermost 1~kpc. What

145: is the nature of this component? Can measurements of the kinematics of

146: stars in the innermost 1~kpc region help in solving some of these issues?

147: 

148: Photometric evidence for a bulge  has been presented for M33

149: but this is far from being conclusive.  Exponential profiles

150: typically provide a good fit to the stellar surface brightness in the

151: large-scale disk but in the inner few hundred parsecs the brightness

152: exhibits an excess with respect to the inward extrapolation of the

153: disk exponential law \citep[e.g.][]{bot92,reg94}. It is unclear

154: whether this is due to an excess of stellar density distributed in a

155: disk or to a bulge/spheroidal light distribution.  In the latter case

156: Regan \& Vogel (1994) derive a large angular size spheroid (8$'$ or

157: 2~kpc effective radius) from photometric fits to the disk and

158: spheroidal light distribution.  Minniti et al. (1993) have claimed a

159: photometric detection of a smaller central component, a bulge with an

160: effective radius of 0.5~kpc, through JHK-band imaging of the central

161: region of M33.  This is perhaps a pseudobulge since it underwent a

162: star formation episode less than 1~Gyr ago. They find a displacement

163: of the center of symmetry of the bulge with respect to the nuclear

164: cluster by about 20$''$ to the southeast.  They mention dust as a

165: possible origin of this photometric center shift but we note that the

166: large OB stellar association south-east of the nucleus might also

167: cause some displacement of the light center. Recent GALEX far-UV and

168: near-UV mosaics of M33 \citep{thi05} have also shown an excess of

169: light in the innermost 0.5~kpc region, in agreement with optical

170: photometric results and with the high level of star formation detected

171: by infrared and H$\alpha$ surveys (see Magrini

172: et al. 2007 and references therein).

173: 

174: A change in the ellipticity of the isophotes between the inner

175: and the outer regions of a disk galaxy is ofter used to infer

176: the presence of weak bars in nearby galaxies and to measure the

177: bar strength even in more distant systems \citep{abr99}.

178: Larger IR maps of M33 \citep{reg94,blo04} point at a weak bar which becomes

179: visible after subtraction of the exponential disk and a spheroidal

180: light distribution. The bar would have a position angle close to

181: 90$^\circ$ and projected radial extend smaller than 6'.  However an

182: alternative explanation to the bar-like emission is a possible change

183: of the pitch angle of the the spiral arms at smaller radii.  Fourier

184: decomposition of galaxy images in the blue has also revealed a bar

185: structure \citep{elm92}.  The metallicity, which has a negligible

186: gradient in the entire star forming disk of M33 \citep{cro06,mag07},

187: also rises suddenly in the innermost 0.5~kpc indicating a possibly

188: complex star formation history.

189: 

190: It is remarkable that in spite of this large number of photometric

191: studies of the innermost region of M33, there is not a similarly

192: detailed study of the stellar and gas kinematics. Colin \&

193: Athanassoula (1981) noticed an appreciable velocity asymmetry in the

194: inner $5'$ velocity field using the ionized gas as a kinematic

195: tracer. This however has not been confirmed by the CO rotation curve

196: \citep{wil89,cor03}. In a conference paper, Rubin \& Ford (1980)

197: presented emission-line gas spectroscopy at $R<1'$ along the major axis

198: only. They find a constant radial velocity all the way into the

199: center, an intriguing result that has not received further follow-up

200: or interpretation.  The published HI data \citep{new80,deu87} do not

201: have sufficient angular resolution to assess the kinematics at small

202: distances from the center and within 200~pc radius

203: only a few molecular clouds have been detected \citep{wil90,hey04},

204: insufficient to measure a rotation curve. Therefore, the central gas

205: kinematics has to be investigated via line emission from HII regions

206: and diffuse ionized gas, which fortunately is very abundant

207: (see Figure 1).  Stellar velocities, to our knowledge, have

208: never been measured. The intrinsic difficulties of accurate

209: kinematical measurements in the central region of this nearby galaxy

210: compared to e.g. M31, lie in the relatively fainter stellar light

211: (except at the location of the nuclear cluster) and in the slow rising

212: of its rotation curve which implies velocities below 30~km~s$^{-1}$ in

213: the inner ~0.5 kpc region. These are of the same order or lower than

214: the expected stellar velocity dispersions.

215: 

216: In this paper we will discuss the results of a set of observations

217: which outline the complex kinematics of the central region of M33.

218: These will hopefully complement other information available for this

219: nearby galaxy, to further establish it as a template for interpreting

220: observations of less resolved, more distant late type galaxies.  In

221: Section 2 we outline our observations of the stellar and gaseous

222: kinematics in the inner 0.5~kpc of M33. Results are presented in

223: Section 3, while Section 4 discusses possible dynamical mass models 

224: to explain the data. We conclude with a summary of our most relevant 

225: results in Section 5.

226: 

227: 

228: \section{Observations and data reduction}

229: 

230: We obtained spectroscopic observations of the central region of M33

231: with the ARC 3.5-m telescope in various short observing runs in 2002,

232: 2004, 2005, and 2006 using the DIS long-slit spectrograph. The two

233: cameras in DIS covered wavelengths from 3700-5500\AA\ (2004) or

234: 4300-5500\AA (other years) in the blue, and 5600-7200\AA\ in the

235: red. The dichroic which splits the light in blue and red components is

236: around 5400\AA\ which implies low sensitivity in the region

237: 5300-5500\AA. Spectra of the Ca triplet near 8500\AA\ were obtained

238: for the major axis (PA=22$^\circ$) in 2005. The slit width for all

239: spectra was 1.5$''$.  The slit length was 5$'$ which corresponds to 1.2 kpc

240: at a distance of 840~kpc. Between the 2002 and 2004 observations

241: the optics of the spectrograph were replaced resulting in changes to

242: the spatial scale from 0.54$''$/pixel to 0.4$''$/pixel. The 2006

243: observations were taken with new high resolution gratings in

244: place. The spectral resolution of the various spectra (FWHM) varied

245: from 2.0 to 4.0 Angstrom. Seeing was typically between 1$''$ and

246: 2$''$. Spectra were obtained at position angles 18$^\circ$, 22$^\circ$

247: (nominal major axis), 112$^\circ$ (minor axis), 96$^\circ$ (suspected

248: bar major axis), 59$^\circ$, 130$^\circ$, 146$^\circ$, and

249: 160$^\circ$. The PA=18$^\circ$ data were combined with PA=22$^\circ$. The

250: slits are shown overlayed on the H$\alpha$ emission image in

251: Figure 1. In some cases we obtained two spectra shifted in position to

252: extend the spatial coverage to more than 5$'$ along a given PA. Major

253: and minor axis spectra were observed a few times with somewhat

254: different centering for the nucleus. In addition, several spectra

255: parallel to, but offset from the major axis spectrum by 1$''$ to 8$''$, and

256: one at 142$''$ SE offset were obtained. These are not shown in Figure 1

257: to avoid clutter. The net exposure times ranged from 20 to 90 minutes,

258: split in double or triple exposures to eliminate cosmic rays in

259: post-processing. The shorter exposure times apply to the intermediate

260: position angles, while the longer times apply to separate major

261: and minor axis spectra. For all exposures the galaxy was accurately

262: centered in the slit using the slitviewer camera before offsets were

263: done as needed.

264: 

265: Several stars of spectral type F5V through K3III were observed as

266: templates for the stellar velocity measurements.  The blue spectra of M33

267: were predominantly used to measure the stellar velocities, although

268: the H$\beta$ and [OIII] emission lines did provide a check on the

269: H$\alpha$, [NII], and [SII] velocities from the red spectra. In the

270: 2006 run of PA= 130$^\circ$, 146$^\circ$, and 160$^\circ$ the red

271: camera was not available and we include the H$\beta$ emission line

272: velocities for those cuts. The Ca triplet observations provided a more

273: accurate zero point for the stellar velocities than the blue spectra

274: due to the presence of numerous bright sky lines in the IR. Since we

275: wanted to measure the red emission lines, and because the CCD in use

276: for our observations showed significant fringing in the IR and had

277: rather low sensitivity around 8500\AA, we only observed the major axis

278: at this wavelength range. These data did enable us to anchor the

279: velocity of the cluster and bright disk part on the major axis, and to

280: measure stellar velocity dispersions along the brightest part of the

281: major axis.

282: 

283: The large angular size of M33 implies that separate sky exposures had

284: to be obtained to subtract sky lines. We observed blank sky regions

285: each night between M33 exposures. HeNeAr lamp exposures were obtained

286: regularly for wavelength calibration. For the red spectra, velocity

287: zero points were determined from the 6923\AA\ sky line for the red

288: emission lines, and from 4 sky lines near the Ca triplet. The blue

289: region of the spectrum has no bright sky lines near the region of

290: interest and the blue spectra's zero points were set by aligning them

291: with the corresponding red spectra. We estimate the final uncertainty

292: in the zero points to be $\pm 4$~km~s$^{-1}$. 

293: 

294: All data were reduced using standard IRAF processing techniques.

295: Elimination of cosmic rays was done using the crreject option in

296: IRAF. This works well for most of the spectra, but can introduce

297: spurious effects for the bright nuclear star cluster, due to slight

298: changes in seeing, the effects of refraction in the earth's

299: atmosphere, or in exact centering of the nucleus in the

300: slit. Therefore, measurement of the gas velocities close to the

301: nucleus (shown in Figures 8 and 9) and of the Ca triplet data for the

302: star cluster was done on single spectra. Fringing in the Ca triplet

303: data was reduced satisfactorily by flatfielding the spectra with the

304: quartz flat.

305: 

306: In addition to the spectroscopic observations we also have a deep

307: H$\alpha$ image of M33 obtained with the Burrell Schmidt (Hoopes \&

308: Walterbos 2000) and we obtained short-exposure, sub-arcsec seeing

309: H$\alpha$, [OIII], and continuum images of the central 4.6$'$ with the

310: SPICAM imaging camera on the ARC 3.5-m telescope. The latter images

311: provide useful information on the morphology and location of dust

312: lanes, OB assocations, HII regions, and other emission-line

313: sources. Some of the images are shown in Figure~1.

314: 

315: 

316: 

317: From 21-cm data the inferred systemic velocity of M33 is

318: -180~km~s$^{-1}$, the major axis orientation is PA=22$^\circ$, and the

319: disk inclination is $i=52^\circ$ \citep{cor00} and we shall use these

320: values throughout the paper.  The approaching half of M33 is west of

321: the major axis.  In showing the velocities along a slit position, our

322: convention is that negative radial distances refer to the east side of

323: the galaxy and increase towards the west side.

324: 

325: \subsection {Gas velocity and velocity dispersion measurements}

326: 

327: The velocities of the ionized gas were determined on blue and red

328: spectra, measuring the H$\beta$, [OIII]5007, H$\alpha$, [NII]6783,

329: [SII]6716, and [SII]6731 lines. In the red, the four lines were

330: measured simultaneously to determine one redshift, and individually to

331: test internal consistency. Close to the nucleus [NII]6583 remained the

332: only line visible due to a strengthening of the [NII]/H$\alpha$ ratio

333: and the strong stellar H$\alpha$ absorption of the stellar

334: nucleus. Outside the nucleus H$\alpha$ was the strongest

335: line. Velocities from the red spectra typically have better accuracy,

336: although the H$\beta$ velocities generally agreed well with the

337: H$\alpha$ results. Velocities from the [OIII]5007 line line did not

338: always agree with the Balmer line measurements, likely indicating real

339: differences. The [OIII] emission is enhanced in shocks and high

340: excitation regions and it is possible that the [OIII] velocities

341: preferentially probe a less quiescent gas phase; this will be

342: addressed in a second paper where we look at the excitation ratios of

343: the gas in the central region. All the lines were measured with the

344: SPLOT routine in IRAF.  From consistency checks between different

345: exposures of the same P.A., overlapping sections between spectra, and

346: internal consistency between the 4 emission lines in each red spectrum

347: we infer that typical uncertainties in the emission line velocities

348: are of order 4~km~s$^{-1}$ (including zero point uncertainties). 

349: %When

350: %data from different exposures or lines are averaged into bins along

351: %the minor and major axis the uncertainty around the mean are of order

352: %1~km~s$^{-1}$ inside 1$'$ and and 2~km~s$^{-1}$ further out. 

353: 

354: 

355: \subsection{Stellar velocity and velocity dispersion measurements}

356: 

357: In order to derive the M33 stellar velocities along the line of sight

358: we used a cross-correlation technique \citep[as described by][]{kur98}

359: and implemented in the RVSAO set of programs in the IRAF

360: environment. This technique is based on the original power spectrum

361: method developed by Tonry \& Davis (1979) and is suitable for rotation

362: curve measurements when line profiles are Gaussian and no complex

363: velocity structures or multiple components are present \citep{bar00}.

364: The faint and low mass stellar disk of M33 does not show evident signs

365: of complexity and furthermore the high spatial resolution achieved in

366: this nearby galaxy minimizes the possibility of multiple components

367: overlapping along the line of sight. Our spectral resolution is such

368: that no precise information on the stellar velocity dispersion (of

369: order 25-30~km s$^{-1}$, e.g. Kormendy \& McClure 1993) could be derived

370: from the blue data, but the Ca triplet data did provide some results.

371: The stellar disk kinematics were obtained from absorption features

372: present in the highest resolution spectra across the wavelength range

373: running from about 4400\AA\ to 5500\AA.  In order to investigate the

374: reliability of $H\beta$ absorption line, often mixed with emission

375: line gas, we have also used the cross-correlation technique in the

376: restricted wavelength range 5030\AA\ to 5500\AA\ and find consistent

377: results: we shall refer to this restricted wavelength range

378: correlation as the `Mg side' correlation. All velocities were

379: corrected to V$_{helio}$ using the program rvcorrect in IRAF.

380: 

381: A zero point offset in the velocity scale arises in the resulting

382: velocities because of the varying flexure of the spectrograph as it

383: rotates at the Nasmyth focus. The stars observed for cross-correlation

384: are never observed at the same rotator angle of M33 implying a varying

385: relative zero point after wavelength calibration is applied. We

386: aligned the velocity curves setting the cluster velocity to

387: -180~km~s$^{-1}$, as suggested by Ca triplet line analysis.

388: 

389: A number of stellar templates were observed for each instrumental

390: set-up. Spectra were rebinned along the spatial direction to obtain a

391: nearly constant signal-to-noise ratio along the slit. The

392: uncertainties shown in the results are those provided by the XCSAO

393: routine outputs and they do not take into account possible template

394: mismatch. Along the disk, different templates gave consistent results

395: to within a zero point shift only. The stellar nuclear cluster is

396: bluer than the disk, likely different in age, and it may host a

397: mixture of young and old stellar populations. Some template mismatch

398: between nuclear cluster and disk then can cause higher uncertainties

399: in the velocity zero point of the cluster compared to that of the disk

400: when blue spectra are used. This results in an increased scatter

401: eventhough the measured radial dependence of the stellar velocities in

402: M33 was generally consistent along the disk. This is evident for

403: example around the nuclear cluster region and along the minor axis

404: (shown in Figure~3, top panel).  The Ca triplet lines are narrow and

405: strong and vary only slowly with metallicity and star

406: temperature. They can therefore be used even if a mixture of stellar

407: populations is present \citep{kor93}. Outside of the nucleus however,

408: as the lines become weak, sky line contamination and remaining

409: fringing become a problem. We therefore obtained results only for the

410: cluster and central disk region on the major axis.  To best estimate

411: the radial velocity of the star cluster we measured the position of

412: sky lines and the wavelengths of the Ca triplet lines in the

413: single-exposure wavelength calibrated spectrum with the best

414: seeing. This was done through a fit to several sky lines and to the

415: sharp troughs in the triplet absorption lines. This way we avoid any

416: influence of the stellar template that is used in the

417: cross-correlation method on the derived radial velocity. In addition,

418: by tying the absorption velocity to the sky line wavelengths we

419: eliminate effects of a zero point offset in the velocity scale

420: associated with varying flexure in the spectrograph as it rotates at

421: the Nasmyth focus. The resulting velocity we derived for the cluster

422: was indistinguishable within the uncertainty from the systemic

423: velocity of -180 km~s$^{-1}$. We estimate our zero point uncertainty

424: to be about $\pm4$ km~s$^{-1}$. Through cross-correlation with stellar

425: templates in the restricted wavelength range 8480\AA\ 8680\AA\ we also

426: measured stellar radial velocities close to and in the cluster region

427: (Figure 2).

428: 

429: 

430: \section{Kinematics of Gas and Stars in the central region}

431: 

432: The M33 data presented in this paper probe velocities at high spatial

433: resolution and very close in to the nucleus compared to typical

434: rotation curve measurements obtained from the ground for almost all

435: other galaxies.  A few issues that pertain specifically to the central

436: region of this nearby low luminosity disk galaxy need to be

437: underlined. First, the observed radial velocities and the velocity

438: dispersions are quite low. Radial velocities are less than 30 km~s$^{-1}$. 

439: This is of the

440: same order as might be expected for velocity gradients across spiral

441: arms and for velocities in expanding HII shells around regions of star

442: formation. We will focus here on systematic velocity patterns rather

443: than individual features. Second, the typical uncertainty in the zero

444: point determinations is about $\pm 4$ km~s$^{-1}$. This implies that

445: any apparent asymmetries of this order in the measured radial

446: velocities from one side of the center to the other may not be

447: real. Third, the scaleheights of the gaseous and stellar disks are

448: expected to be much larger than the closest distances to the central

449: region we can probe: 1$''$ corresponds to 4~pc. Therefore, projection

450: effects can lead to misinterpretation of velocities very close to the

451: center, because the gas could be further away from the center even if

452: seen projected close in. We don't know the scaleheights of the various

453: components in the mass modeling and this also becomes a bigger problem

454: close to the center where the nuclear cluster is a compact source in a

455: likely much thicker gas disk.

456: 

457: \subsection{Emission line velocities of the ionized gas}

458: 

459: The major and minor axis velocities of the ionized gas in the inner

460: region of M33 are shown in bottom panels of Figures 2 and 3. The most

461: remarkable feature of the major axis velocities is that they are

462: constant at $\sim 22$~km~s$^{-1}$ on average from $2'$ (0.5~kpc) down

463: to within $2''$ ($\simeq 8$~pc) from the center, in agreement with the

464: results by Rubin \& Ford (1980). The minor axis velocities are

465: characterized by significant non-zero velocities within 1-1.5$'$ from

466: the center, and reach maximum values similar to those observed along the

467: major axis.  Even though velocities have a regular pattern along the

468: minor axis, they are far from being constant. Here velocities are not

469: symmetric with respect to the blue cluster center which 

470: appears to be located at or close to the dynamical center (see

471: also Section 3.2 and 3.3). On the west side of the minor axis

472: velocities vary by about 50~km~s$^{-1}$ across the innermost $1'$

473: region.  The minor axis is close to the suspected bar direction

474: (estimated at approximately PA=96$^\circ$ from

475: Fig.1$(c)$ in Block et al.2004 or Fig.1$(b)$ in Elmegreen et al. 1992). 

476: We show the ionized gas velocities

477: for PA=96$^\circ$ in the middle panel of Figure 4. The velocities in this

478: direction are much like that of the minor axis, with

479: somewhat reduced amplitude. The velocities East of -1$'$ drop

480: below systemic as expected, given we see the

481: approaching side of the disk here.

482: 

483: 

484: The other two panels in Figure~4 show velocities along PA=59$^\circ$ and

485: PA=146$^\circ$, while Figure~5 repeats the PA=146$^\circ$ data

486: averaging all four red lines, and adds H$\beta$ measurements along

487: PA=146$^\circ$ and the neigboring directions PA=130$^\circ$ and

488: 160$^\circ$. A few points are worth noticing about the radial

489: velocities at the various intermediate PAs. First, the radial velocity

490: maxima reach values as high as observed on the major axis. This

491: would not be expected if major axis velocities were due to gas in

492: circular motion around the center. Specifically, if the rotation curve

493: were flat inside 0.5 kpc as suggested by the major axis velocities, we

494: ought to see constant velocities with a reduced amplitude and a sign 

495: change at the center, for all the intermediate PAs. In the inner $\pm1'$ 

496: velocities are instead asymmetric with

497: respect to the center, especially along PA=146$^\circ$. Here east side

498: velocities rise to about 30~km~s$^{-1}$ and abruptly decrease around

499: -1$'$ while on the west side they stay close to systemic for D$<1'$.

500: 

501: In Figure~7 we show line of sight velocities observed along directions

502: parallel to major axis. The slits  cut the minor axis at

503: different projected distances from the nuclear cluster center.  

504: Major-axis velocities show the steepest jump when the minor axis

505: is crossed. The systematic shift in the velocity at the minor axis

506: crossing that is evident in the various cuts, is in agreement with the

507: non-systemic velocities observed on the minor axis.  We don't notice

508: any peculiarity in the velocity curve which samples regions at larger

509: galactocentric distances along a direction parallel to major axis

510: (142$''$ east offset). The velocities appear normal for disk rotation

511: and the intersection with the minor axis occurs at the systemic

512: velocity this far out.

513: 

514: We tried to measure the velocity dispersions implied by the widths of

515: the emission lines. The highest S/N exposures of the major and minor

516: axis gas velocities in the red were from 2002. These data have a

517: spectral resolution (FWHM) of about 2.5\AA, corresponding to

518: 115~km~s$^{-1}$, or a 1 sigma dispersion of about 50 km~s$^{-1}$. We

519: measured the dispersions by deconvolving the widths of the 4 principal

520: red lines (H$\alpha$, [NII]6548, [SII]6716+6731) with the widths of

521: nearby sky lines in the same frame, or with the widths of lines in the

522: arc spectrum at exactly the same location of the data. Typically, 

523: in cases where the deconvolution of the line

524: widths gave a consistent result, the derived dispersions were 12 to

525: 20~km~s$^{-1}$ with only very few locations of higher values. There

526: were also several cases where no discernable broadening could be

527: measured. The dispersion due to thermal motion of the gas, for

528: temperatures of 5000 to 10,000K, would be 6 to 8~km~s$^{-1}$ for the

529: Balmer lines.

530: 

531: \subsection{The stellar velocity pattern}

532: 

533: In the top panels of Figures~2 and 3 we show the stellar velocities

534: measured along the major and minor axis. For the major axis, we

535: include results from the Ca triplet and from the blue spectra, while

536: the minor axis has only velocities derived from the blue

537: spectra. The use of the whole wavelength range in the blue or only of

538: the Mg side (see Section 2) gave consistent results. Velocities

539: along the major axis rise much more slowly for the stars than for the

540: gas. On the minor axis, velocities are close to or at systemic,

541: although we do note that average velocities within $\pm 1'$ from

542: the center are positive with respect to the systemic velocity, a trend

543: which is also observed for the gas.  This assumes that the star

544: cluster is at the systemic velocity, a result derived from the Ca

545: triplet measurements. The increased scatter in the minor axis stellar

546: data compared to the major axis data is likely due to increased

547: uncertainty regarding the relative zero points of the nuclear star

548: cluster versus the disk velocities, as discussed in Section 2.

549: 

550: We also measured stellar velocities at PA=130$^\circ$, 146$^\circ$,

551: and 160$^\circ$. These data were obtained with a new blue high

552: resolution grating in place in 2006. The results are shown in the

553: panels of Figure~6. The radial velocity patterns for the stars along

554: these directions are quite similar to those of the gas (Figure 5) but

555: of lower amplitude.

556: 

557: The Ca triplet data also provided some information on the stellar

558: velocity dispersions. We find 20-23 km~s$^{-1}$ for the star cluster,

559: in good agreement with previous results \citep{kor93,geb01}, and

560: values in the range 28 to 35 km~s$^{-1}$ along the inner disk major

561: axis.

562: 

563: 

564: 

565: \subsection{Velocities in the innermost 60~pc region}

566: 

567: To better characterize the emission line velocities close to the

568: nucleus, we measured each of the four red emission line on individual

569: exposures of the major and minor axis. This was done to avoid any

570: smearing or loss of information that might be introduced close to and

571: across the bright nucleus in the combination of frames during the

572: cosmic ray elimination process. We measured the velocities every

573: single spatial pixel over a region of -15$''$ to +15$''$ along the

574: major and minor axes. The velocities derived this way were all shifted

575: in zero point using the 6923\AA\ sky line, and then averaged for each

576: pixel when all frames have consistent values i.e. an rms dispersion

577: around the mean less than 25~km~s$^{-1}$. Measurements are not

578: averaged and refer to single frames when the rms dispersion was larger

579: or lines were hardly measureable due to blending with stellar

580: absorption or to intrinsic weakness. [SII] velocities are an average

581: of the velocities derived for the 6716\AA\ and 6731\AA\ lines when

582: they give consistent values. In the opposite case individual

583: measurements of the two [SII] velocities are shown, rather than the

584: average and rms dispersion. Results are shown in Figures~8 and 9.

585: 

586: Data along the major axis show that velocities remain constant in

587: H$\alpha$ and [NII], while the [SII] velocities seem to increase near

588: the center. The latter result is uncertain given the large scatter in

589: individual measurements. The only emission line that remains visible

590: across the stellar nucleus is [NII]6854. Even this line is however of

591: limited S/N due to underlying very bright stellar continuum; the

592: [NII]6584 line sits just at the edge of the deep stellar H$\alpha$

593: absorption associated with the cluster. The [NII] velocities on the

594: major axis suggest that the systemic velocity is not measured at the

595: position of the cluster but about 1$''$ to the SW of it (along the

596: major axis). This offset was also noted by Rubin \& Ford (1980).  On

597: the minor axis we find the same value of the gas radial velocity we

598: measure at the cluster position on the major axis.  The results are in

599: agreement and point out to a gas velocity of order -200 km~s$^{-1}$

600: which could imply that the dynamical center is not exactly at the

601: cluster center. However, the presence of similar velocity jumps over

602: short distances on the cuts parallel to the major axis (Figure~7) and

603: the possibility of outflows and non-circular motions imply that it may

604: be difficult to pinpoint the exact dynamic center of M33.  The nucleus

605: might be "wandering" around the dynamic center, either falling in or

606: oscillating (e.g. Miller $\&$ Smith 1992).

607: 

608: Another somewhat striking aspect of Figure~9 is the negative velocity

609: feature on the minor axis at D$\simeq 8''$ in western

610: direction.  This is measured quite accurately in H$\alpha$, and

611: reproduced at the intersection with the minor axis on the cut 8$''$

612: offset from the major axis in Figure 7 (top panel). The feature is

613: about 30~km~s$^{-1}$ deep with respect to systemic extending over

614: $\pm3$ arcsec or $\pm$12~pc.  No expanding HII shell is visible at

615: this location. In our H$\alpha$ image there is a combination of

616: diffuse emission and faint filamentary structures in this region, much

617: like the emission over most of the area covered by our spectra, but

618: not a single large shell. It is possible that the feature is part of

619: the general oscillatory patterns we see in the gas motions here, that

620: may be associated with shocks or with the bar (see Section 4). A

621: more speculative option is that the dip is associated with a possible

622: keplerian fall off or rotation connected to a black hole at or near

623: that location. However, at our slit positions we did not notice any

624: evident increse in the stellar velocity dispersion. Any meaningful

625: discussion on this issue requires full 2-D velocity field mapping of

626: the central region with high spatial and velocity resolution.

627: 

628: 

629: 

630: 

631: \section{Mass modelling of the central region}

632: 

633: 

634: A comparison of the stellar and gas major axis velocities points to a

635: systematic difference: the stars seem to rotate more slowly than the gas

636: does.  This

637: behaviour suggests that asymmetric drift reduces the stellar rotation

638: curve, with the effect being most pronounced close to nucleus, as

639: observed.  There are two other noticeable features in the ensemble of

640: gas radial velocity plots shown in the previous Sections.  $(a)$ Going

641: radially outwards from the nuclear cluster there is not a particular

642: direction that has the maximum or minimum radial velocities.  The

643: major axis direction cannot be easily defined then: for each slit

644: position crossing the center the maximum observed radial velocity is

645: of order 20-30~km~s$^{-1}$ in the radial range $\pm 1'$.  Continuous

646: tilting of the orbits i.e. a non defined preferential direction for

647: the major axis is not sufficient to explain the observed pattern.

648: $(b)$ Velocities along a slit passing through the center do not show a

649: clear symmetry with respect to the nuclear star cluster except along

650: PA=22$^\circ$, the major axis of the large scale disk.  Asymmetries

651: with respect to the nuclear cluster might be present in the stellar

652: velocities as well, especially along PA=146$^\circ$, even though

653: uncertainties in the results of the cross-correlation tecnique and in

654: the central cluster velocity limit our ability to firmly establish

655: their amplitude.  Thus models which consider pure circular orbits 

656: might be inappropriate for the innermost region of M33. 

657: However, we will first explore these models, where circular components

658: make the largest contribution to the observed radial velocities, attributing

659: any gas residual velocity to local peculiar motions such as radial 

660: inflows/outflows, shocks, etc. In this

661: framework asymmetric drift should provide the likely explanation for

662: any difference between gas and stellar velocities and gas velocities

663: should be consistent with a dynamical mass model. 

664: In the second part of this Section we shall examine if

665: possible asymmetries in the stellar and gas velocity pattern can be

666: related to the displacement of a small bulge, as suggested by Minniti

667: et al.  (1993), which would be the dominant mass component in the

668: innermost region. At the end of this Section we interpret the emission

669: line gas kinematics inside $\pm 1'$ in terms of large non-circular

670: orbits around the dynamical center and discuss the possible role of a

671: weak stellar bar in determining the gas flow pattern and its

672: deviations from pure circular orbits.

673: 

674: 

675: \subsection{The axisymmetric potential}

676: 

677: In this Section we consider circular orbits for the gas and stars,

678: centered on the nuclear star cluster. The star distribution is

679: symmetric around the azimuthal direction and about the disk equator.

680: We shall assume that the orientation of the orbits along the line of

681: sight follows that of the large scale disk, namely they are inclined

682: by $i=52^\circ$ around the position angle of the major PA=22$^\circ$.

683: We assume at first that the circular rotational velocity is traced by

684: the gas emission lines observed along the nominal major axis,

685: corrected by the inclination of the disk along our line of sight.  For

686: $R<2'$, gas radial velocities imply a flat rotation curve with 

687: an amplitude of 29~km~s$^{-1}$.

688: 

689: 

690: In Figure 10 we show these velocities, averaging and binning red and

691: blue emission line velocities along along PA=22$^\circ$ (triangle

692: symbols). The open stars are stellar rotational velocities along the

693: same direction. In Figure 10 the CO J=1-0 rotation curve is also

694: plotted. This is obtained from azimuthal averages of data 

695: within $\pm 45^\circ$ of the major axis (Corbelli 2003).  The azimuthal

696: averages of the CO velocities are larger than optical line emission

697: velocities at radial distances around 0.5~kpc. This may be due to

698: peculiar motion (e.g. crossing of spiral arms) along various directions 

699: or to the fact that PA=22$^\circ$ may not

700: be exactly the major axis direction.  The

701: rotation curve is traced by optical emission line velocities:  is

702: flat inside $R=2'$ and the mass density should decline as $1/R^2$.  

703: The stellar circular

704: velocities follow the gas rotational velocities after 

705: asymmetric drift corrections are applied.

706: Considering the mean of 

707: the radial velocity times the velocity perpendicular to the plane as

708: constant with height above the plane, from the mass

709: volume density $\rho_s$ we can estimate the drift following equation 4.33 

710: of Binney $\&$ Tremaine (1987) as:

711: 

712: \begin{equation}

713: V^2_c-V^2_s=-{{\sigma_R^2}}\lbrack 1-\sigma^2_\phi/{{\sigma_R^2}}

714: +2\ \partial {\hbox{ln}}\rho_s/\partial {\hbox{ln}}R \rbrack 

715: \end{equation}

716: 

717: \noindent

718: where  $V_c$ and $V_s$ are the circular velocity and observed stellar 

719: rotational velocity respectively,  $\sigma_R$, $\sigma_\phi$, and 

720: $\sigma_z$ are the radial, azimuthal and vertical velocity dispersion. 

721: In the above equation we have assumed that 

722: $\partial {\hbox{ln}}(\sigma_R^2\rho_s)/\partial {\hbox{ln}}R \simeq 

723: 2 \partial {\hbox{ln}}\rho_s/\partial {\hbox{ln}}R $.

724: Using also the assumption that $\sigma_\phi^2\simeq \sigma_z^2 \simeq 

725: 0.45{{\sigma_R^2}}$ (Binney $\&$ Tremaine 1987, eq.3.76), we have:

726: 

727: \begin{equation}

728: \sigma_z^2={V^2_c-V^2_s\over 2.2 \lbrack -0.65

729: -2\ \partial {\hbox{ln}}\rho_s/\partial {\hbox{ln}}R \rbrack}

730: \end{equation}

731:  

732: The observed stellar rotation is well fitted by $V^2_s\simeq 200

733: R^2$, where $V$ is in km~s$^{-1}$ and $R$ in arcmin. Assuming that

734: the observed gas velocities trace $V_c$ and hence the gravitational

735: potential, we can infer the mass density $\rho_s$ from the gas kinematics.

736: The flatness of $V_c$ in the innermost region of M33 and the above

737: equation implies $\sigma_z$ of order 10~km~s$^{-1}$. This is too a

738: small value given the observed stellar dispersions in the disk

739: (of order 30~km~s$^{-1}$), away from the nuclear cluster. Therefore

740: the mismatch between a flat rotation curve for the gas and a rising

741: rotation curve for the stars cannot be only the result of asymmetric

742: drift. Likely the flatness of the radial velocities distribution along

743: PA=22$^\circ$ is not due to pure circular motion but to a

744: combination of rotation and some streaming motions in a

745: non-axisymmetric potential. This picture also provides a natural

746: explanation for the non zero velocities observed along the putative

747: minor axis at PA=112$^\circ$ and we shall discuss it in Section 4.3.

748: A slow rising gas rotation curve reaches a better agreement with the

749: stellar rotation curve affected by asymmetric drift.  Since stellar

750: velocity dispersions in the disk are 30 km~s$^{-1}$ or higher, it is

751: feasible to assume $\sigma_z> 20$~km~s$^{-1}$. Since $V^2_c-V^2_s$

752: is of order 29~km~s$^{-1}$ close to the center and decreases further

753: out, then eq.(1) implies that the derivative of the stellar

754: density should be:

755: 

756: \begin{equation}

757: \partial {\hbox{ln}}\rho_s/\partial {\hbox{ln}}R > -0.8

758: \end{equation}

759: 

760: This condition implies that asymmetric drift can explain the differences

761: between the stellar and the gas rotation curve when the last one is 

762: rising. This is inconsistent with the hypothesis that

763: optical gas velocities observed along PA=22$^\circ$ effectively

764: trace the rotation curve in an axisymmetric potential. An almost flat

765: rotation curve requires in fact the above derivative to be of order $-2$.

766: There are other difficulties related to the interpretation that the

767: optical gas velocities observed along PA=22$^\circ$ define a flat

768: rotation curve for this galaxy. If the constant velocities of order

769: 29~km~s$^{-1}$ along the major axis at PA=22$^\circ$ are due to pure

770: circular rotation then we should find constant velocities of order

771: 14,4,9,13~km~s$^{-1}$ along

772: PA=59$^\circ$,130$^\circ$,146$^\circ$,160$^\circ$ respectively.  We

773: don't have any evidence for these constant values in the inner region

774: except possibly for PA=160$^\circ$.  Moreover, we find no acceptable

775: combination of mass components (bulge, dark matter halo, stellar disk)

776: for fitting a completely flat curve in the innermost 0.5~kpc region of

777: M33 and its sudden rise around 0.5~kpc. In

778: addition, the disk mass to light ratio in the innermost regions is

779: constrained by the fit to the large scale RC and cannot be considered

780: arbitrarily small such as to leave a dominant shallow bulge component

781: out to 0.5~kpc.

782: 

783: In this sub-section we still want to examine another possibility

784: related to the axisymmetric potential, namely that the rotation curve

785: is traced by the CO and HI lines down to 0.3~kpc. We consider only the

786: two innermost points of our optical binned emission line set to

787: complement this further in.  This removes our previous assumption of a

788: flat rotation curve out to R$\sim 0.5$~kpc and that major axis

789: position angle is strictly 22$^\circ$.  The higher azimuthal averages

790: of CO velocities considered might in fact indicate that indeed the

791: rotational velocities are higher along other directions (consistent

792: with our data).  In Figure 10 $(a)$ the circled symbols show the inner

793: data used to define a rotation curve and which we now use fit a mass

794: model.  For $R>1.2$~kpc the rotation curve used is described by

795: Corbelli (2003) and extends out to 19~kpc.  We follow the approach

796: of Corbelli (2003) in a cold dark matter framework and use a dark

797: matter profile as described by Navarro et al (1997).  As mass

798: components we shall use an exponential stellar disk, a dark matter

799: halo, a bulge, a gaseous disk and the nuclear cluster.  We allow the

800: exponential disk to have all possible mass-to-light ratio values. For

801: the bulge we assume a rotational velocity of the following form:

802: 

803: \begin{equation}

804: V_{sph} = G M_{sph}  R^{0.5}/(R+s) 

805: \end{equation}

806: 

807: \noindent where the mass $M_{sph}$ and the scale radius $s$ are free

808: parameters.  We set the stellar disk scalelength equal to the K-band

809: scalelength of 5.6$'$ (Regan $\&$ Vogel 1994) and search for the bulge

810: and dark matter parameters which give the lowest $\chi^2$ compatible

811: with the data.  The lowest bulge mass and scalelength $s$ compatible

812: with the data are $9\times 10^7$~M$_\odot$ and $s=0.15$~kpc.

813: We show the central region of this mass model fit in Figure 10$(a)$.

814: The mass to light ratio of the disk and of the bulge are of order

815: unity even though uncertainties in the bulge luminosity leave this

816: ratio uncertain for the bulge itself. A larger bulge gives acceptable

817: fits and we cannot exclude it from dynamical arguments. The dark

818: matter halo parameters in this fit are similar to those found by

819: Corbelli (2003).  The gas rotation curve in this case is far from

820: flat and it is compatible with the rotation curve inferred from the

821: stellar absorption lines, provided that the last one is affected by

822: asymmetric drift.  The velocity dispersions required by the asymmetric

823: drift correction to stellar rotation in order to agree with gas

824: rotation are of order 20-30~km~s$^{-1}$.

825: 

826: Could the position angle of the major axis be effectively changing for

827: $0.2< R<0.6 $~kpc, as indicated by higher velocities observed in CO

828: along other directions? We cannot exclude that the direction of major

829: axis is changing with radius but this alone does not explain the

830: complex pattern we observe in the emission line velocities.  Despite

831: the good dynamical fit to the slow rising rotation curve of Figure

832: 10$(a)$, compatible with the dynamical model of the extended disk

833: rotation, its projection along other directions does not agree with

834: the observations presented in this paper.  Orthogonal directions such

835: as PA=146$^\circ$ and PA=59$^\circ$ reach similar velocities on the

836: East side for example but much lower value on the West side.

837: Overcoming these discrepancies in the framework of simple models

838: discussed above implies taking into account additional peculiar

839: motions. M33 has its near side west of major axis. Evidence for this

840: is provided by dust lanes, which become more pronounced west of

841: the major axis in a blue image of M33 (e.g. Figure 1). This

842: orientation is also consistent with trailing spiral arms.  In this

843: geometry, a positive anomalous velocity on the near side can be

844: interpreted in terms of local radial infall of gas towards the

845: center. Evidence for local radial gas infall in our sampled areas

846: might come only from a region 1$'$ west on minor axis and along the East

847: side at PA=59$^\circ$.  These peculiar velocities might be connected

848: to shocks along dust lanes or to the beginning of the northern/southern 

849: spiral arm.  The possibility that these features are related to

850: shocks will be analyzed in more detail in a forthcoming paper where

851: emission line ratios are examined.  South-East of the nucleus, towards

852: the strong OB association, the gas shows strong positive velocities

853: which could be interpreted as a large nuclear outflow, possibly

854: intersecting also the innermost 1$'$ East on minor axis.  However,

855: since a similar anomaly is observed also in the stellar velocity

856: pattern, even though of smaller amplitude, we cannot give any definite

857: conclusion on the existence of such outflow.

858: 

859: In the rest of this Section we will discuss the possibility that the

860: anomalous velocities we observe are not related to gas inflow or

861: outflow patterns superposed on pure circular rotational velocities

862: in an axisymmetrical potential, but to large scale distortions of the

863: potential, such as given by a displaced bulge or by a weak bar.

864: 

865: 

866: \subsection{A displaced bulge}

867: 

868: The ordered motion detected in central regions of M33 and the high

869: star formation rate (Thilker et al. 2005; Magrini, Corbelli, $\&$

870: Galli 2007) point out that if any bulge is in place this is more

871: likely a pseudo-bulge, intermediate between a spheroid and a disk.

872: 

873: Looking in detail at the gas and stellar radial velocities along

874: PA=130$^\circ$,146$^\circ$,160$^\circ$ we notice that between -1$'$

875: and 0.5$'$ velocities are symmetric around a point located at

876: $R\simeq -0.25'$ where velocities are of order

877: -170~km~s$^{-1}$. Abrupt changes in the velocity are noticeable

878: outside the -1$'$ -0.5$'$ region. These asymmetries around a displaced

879: center with respect to nuclear cluster are possibly in line with a

880: suggestion by Minniti et al. (1993) about a bulge that is offset from

881: the nuclear cluster. They reached this conclusion from looking at the

882: V and H band images of the center of M33.  However the presence of

883: dust lanes in this region and of the bright OB association in the

884: south-east did not allow any definitive conclusion to be drawn. The

885: displacement in the kinematic symmetry point we find is of order

886: 15$''$ to the South-East, compatible with the suggestion of Minniti et

887: al. (1993). The excess light they found in the innermost 0.24~kpc, is

888: compatible again with what is required by the mass modeling of the

889: large scale rotation curve.  The excess of light in the innermost

890: 0.5~kpc region detected by azimuthal averages of near infrared J,H,K

891: emission by Regan $\&$ Vogel (1994) points to a larger spheroidal

892: component (8$'$ effective radius).  This conclusion however relies on

893: the assumption that the exponential disk scalelength does not change

894: going from the extended disk to the innermost region and that an

895: $r^{1/4}$ light distribution is effectively in place for the

896: spheroid. In reality, departures from $r^{1/4}$ law are evident both

897: in the blue and in the near infrared bands (Bothun 1992; Regan $\&$

898: Vogel 1994).  A possible displacement of the bulge however does not

899: explain the peculiar symmetric steep rise and flatness of the gas

900: velocity observed along PA=22$^\circ$ centered on the stellar nuclear

901: cluster.

902: 

903: 

904: \subsection{An oval bar}

905: 

906: As stated in the Introduction, photometric evidence for a bar has been

907: given by near-IR photometry but also by the Fourier transform

908: technique on galaxy images in the blue band (Elmegreen, Elmegreen $\&$

909: Montenegro 1992). We shall discuss here whether any anomalous velocity

910: with respect to the axisymmetric potential can be interpreted in terms

911: of a triaxial potential, and if this triaxial figure is rotating, as

912: is the case for a bar. Near-IR images point to a bar which we

913: initially estimated to be roughly aligned along PA=96$^\circ$. 

914: This position angle estimate is close to

915: the minor axis and that makes the bar more difficult to see because

916: projection effects make it appear rounder. To investigate the light

917: distribution we downloaded the 2MASS K-band image, rotated it by 

918: 22$^\circ$ clockwise to align the major axis with y-axis, and stretched

919: the x-pixel size by 1.62 to attempt to deproject the image to what it

920: might look face-on if the components are all in a thin disk.  The

921: central portions of the rotated and of the deprojected images are

922: shown in Figure~11. The stretched image needs to be interpreted with

923: caution since e.g. a spherical light distribution would be stretched

924: out into an oval shape. Nevertheless, a central broad oval

925: concentration of light appears to be visible in the non-stretched

926: image too.

927: 

928: We measured the ellipticities of the isophotes on a non-stretched

929: median filtered K-band image. The ellipticity is low averaging 

930: about 0.15 from 50 to 150~arcsec distance from the center, and the PA

931: of the isophotes is $\simeq 80^\circ$ in the non-rotated frame,

932: somewhat different from our previous estimate 

933: (96$^\circ$ by inspecting stretched images). Inside

934: 50~arcsec but still outside the central star cluster the ellipticity

935: is higher, about 0.3. Outside 150~arcsec visible spiral arms

936: converging onto the center limit the validity of the isophote fitting

937: procedure.  The deprojected bar radial extent can then estimated to be

938: at least 2.5$'$ or 0.6~kpc, and it can possibly be even larger.  Regan

939: \& Vogel (1994) estimated a deprojected bar length of about 0.8~kpc

940: while Elmegreen et al. (1992) claim a longer bar, extending out to

941: about 1.5~kpc.  It is not yet clear what limits the extent of weak

942: bars.  According to Combes $\&$ Elmegreen (1993) in late-type galaxies

943: the bar extent is comparable to the radial scalelength of the stellar

944: disk; the K-band radial scalelength of the stellar disk of M33 is

945: about 1.4~kpc\citep{reg94} so this theory predicts a bar length in

946: close agreement with the estimate of Elmegreen et al. (1992).  The bar

947: in M33 is weak according to most classification schemes.  For example

948: following Abraham et al. (1999) and using the 2-MASS image of Figure

949: 11 we computed an intrinsic bar axial ratio of 0.57, and a bar

950: strength parameter $f_{bar}=0.2$. These values place M33 at the

951: boundary between barred and unbarred systems and marginally classify

952: M33 as a barred galaxy. Measuring and interpreting the inner

953: kinematics is therefore especially important for this galaxy in order

954: to find kinematic signatures of the oval bar.

955: 

956: The region along the minor

957: axis where we see anomalous velocities has an extent of about 

958: 1.2$'$ on the sky  i.e. of about 0.5~kpc in radius. This is close to

959: the smallest effective bar length estimate. The positive velocities of

960: the gas East of the star cluster on the minor axis, and the negative

961: velocities on the opposite side, at about 8$''$ away from the stellar

962: cluster, can be interpreted in term of streaming motion of elongated

963: orbits parallel to the bar. It is likely in fact that streaming

964: motions become more pronounced closer to the center, where light

965: contours are more elongated.  The remarkably constant velocities for

966: the emission line gas observed on both sides of the nucleus at

967: PA=22$^\circ$ are also consistent with streaming motions along a bar:

968: For a bar oriented close to the minor axis as is the case for M33,

969: streaming motions along the bar add to the circular velocities on the

970: major axis to produce a sharp discrete jump in velocity across the

971: nucleus. This is seen  clearly in the galaxy NGC6300 which

972: has a bar close to the minor axis and inclination similar to M33 (Buta

973: etal. 2001, their Figure 21). Major and minor axis velocity

974: peculiarities in the inner $\pm 1'$ are then consistent with having

975: streaming motions along a bar.

976: 

977: Let us now assume that the central flat plateau of the velocity

978: curve along PA=22$^\circ$ is mostly due to streaming motion and that

979: the average rotational velocities are lower, somewhere in between the

980: stellar velocities and the CO velocities. For fitting a mass model we

981: take the average velocities traced by optical emission lines 

982: along the northern and southern side of the major axis between 0.2 and 

983: 0.8~kpc. We consider the

984: rotation curve of Corbelli (2003) at larger radii.  In Figure 10$(b)$

985: we show with circled symbols the data points in the inner region which

986: we shall use to fit a mass model.  Since we don't require rotational

987: velocities to be high at $R<0.2$~kpc, we can achieve acceptable fits

988: without putting any relevant mass in the bulge. Best fitting

989: parameters for the stellar disk and for the dark matter halo are

990: similar to those of the model shown in Fig. 10$(a)$. The

991: additional non axisymmetric mass component in the bar is what causes

992: the streaming motion i.e. the larger velocities observed in the gas

993: (at $R<0.2$~kpc along the major axis and in the azimuthal

994: averages traced by the molecular component out to 0.8~kpc) 

995: with respect to the pure circular rotation (continuous line in Figure 

996: 10$(b)$). The difference

997: between the gas and the stellar rotation curve in this case is

998: compatible with the asymmetric drift hypothesis even though in the

999: presence of a bar, stellar and gas orbits might be different.

1000: 

1001: Using this rotation curve and the epicyclic approximation we can then

1002: locate the inner/outer Lindblad resonances, (hereafter ILR/OLR) and the

1003: corotation radius, if a bar pattern speed can be inferred. The

1004: positive radial gradient of the rotation curve shown in Figure 10$(b)$

1005: implies that there is no ILR or that this is close to the cluster

1006: location, at very small radii. In Figure~12 we show the angular

1007: frequency $\Omega$ corresponding to the rotation curve of Figure

1008: 10$(b)$.  After computing the epicyclic frequency $\kappa$ we draw the

1009: $\Omega \pm \kappa/2$ curves needed to locate the inner and outer

1010: Lindblad resonances.  If the bar in M33 ends at its own corotation

1011: (0.6-1.5~kpc) then the bar pattern speed is 10$\pm 2$~km~s$^{-1}$

1012: (shaded region in Figure 12). The intersection between the $\Omega

1013: -(+) \kappa/2$ curve and the bar pattern speed defines the ILR (OLR).

1014: The requirement of a bar ending at the IRL would imply instead a

1015: very low bar pattern speed, of order 1-2~km~s$^{-1}$.  The absence of

1016: an ILR (possibly occurring at very small radii), or a low bar pattern

1017: speed, confirm the weakness of the M33 bar. The likely situation is

1018: that the bar of M33 is at its dawn, and still in the process of

1019: aligning the orbits in a preferential direction.  Our finding is

1020: linked to the argument of Binney $\&$ Tremaine (1987) about the growth

1021: of weak bars: namely, that when a weak bar instability arises in a

1022: disk, most of the orbits are aligned parallel to the bar, streaming

1023: motion are large in the direction perpendicular to the bar and there

1024: is no ILR.

1025: 

1026: Evidence for weak bars or triaxial structures has been

1027: found for two other late-type spirals recently: NGC 628 (Fathi et

1028: al. 2007) and NGC 2976 (Spekkens \& Sellwood 2007). They show

1029: that late-type spirals may not be the simple dynamical systems we once

1030: might have thought them to be. In the case of M33, streaming motions

1031: are difficult to quantify given the small amplitude of the observed

1032: velocities and the small size of the region where these motions take

1033: place.  Even though a weak bar can likely drive non-circular motion

1034: and cause the non zero velocities along minor axis and the flat

1035: velocity profile along the major axis of M33, our impression is that

1036: it cannot account for all peculiar velocities observed in other

1037: directions.  Any more detailed model for a non axisymmetric flow

1038: pattern requires a full 2-D velocity map of the central region.

1039: 

1040: 

1041: \section{Summary and discussion}

1042: 

1043: In this paper we have presented and analyzed optical observations of

1044: gas and stellar radial velocities in the innermost 0.5~kpc region of

1045: M33. We summarize below the main results and their

1046: implication for the mass distribution in the central region of this

1047: nearby galaxy.

1048: 

1049: \begin{itemize}

1050: 

1051: \item[({\em i}\/)]The radial velocity of the bright nuclear

1052: cluster places it close to the systemic velocity of this galaxy.

1053: We see only a small offset in the dynamical center as determined

1054: from the required symmetry in emission line radial velocities 

1055: across the central steep velocity gradient observed along PA=22$^\circ$

1056: (major axis of the large scale disk). The lack of a dominant

1057: gravitational entity in the center of this

1058: galaxy implies that the exact dynamical center cannot be yet defined.

1059: 

1060: \item[({\em ii}\/)]Gas radial velocities indicate a relatively flat 

1061: and symmetric curve along PA=22$^\circ$ between 4$''$

1062: and 2$'$ radial distances from the nuclear cluster. Some

1063: local displacements around the mean value of 22~km~s$^{-1}$ are visible.

1064: No other direction, between all those observed, presents evidence 

1065: for a very steep central gradient 

1066: and for a flat and symmetric velocity distribution such as observed 

1067: along PA=22$^\circ$ (about a 45~km~s$^{-1}$ jump over a few pc).

1068: For each slit crossing the nuclear cluster the maximum observed radial 

1069: velocity however is of order 

1070: 20-30~km~s$^{-1}$ (as along PA=22$^\circ$) but it can occur as far as 

1071: 1.5$'$ from the nuclear cluster. Particular relevant are the non

1072: zero velocities observed for the gas along the minor axis within

1073: projected radial distances of  $\pm 1'$. Given the large scale

1074: orientation of the M33 disk these 

1075: cannot be explained in terms of a gas inflow pattern.

1076: 

1077: \item[({\em iii}\/)]The stellar velocities show smaller radial

1078: gradients than the gas, even though the overall pattern

1079: along each direction agrees with that of the gas. We have discussed

1080: the differences between the gas and stellar velocities in terms

1081: of asymmetric drift correction. We estimate that the asymmetric drift

1082: correction cannot explain the difference between a flat gas rotation

1083: curve and a rising stellar rotation curve for an axisymmetric

1084: potential. Moreover we have not found a mass distribution compatible

1085: with a flat rotation curve inside 0.5~kpc region and with the large scale

1086: rotation curve.

1087: 

1088: \item[({\em vi}\/)]We have discussed the likely possibility that the

1089: flat gas velocity curve observed along PA=22$^\circ$ and the non

1090: zero velocities observed  in the orthogonal direction are connected

1091: with streaming motion due to a weak bar. This bar has been postulated

1092: previously from photometric measures but here we have

1093: presented the first dynamical evidence consistent with this. 

1094: With M33 included, it now seems all major Local

1095: Group galaxies (including the Large Magellanic Cloud) are barred!

1096: 

1097: %\item[({\em v}\/)]Dynamical models to the data leave open the 

1098: %possibility of a small bulge, likely a pseudo bulge, which could 

1099: %perhaps take the place of the bar

1100: %if it is triaxial. A possible displacement of its center

1101: %with respect to the nuclear cluster cannot be ruled out.

1102: 

1103: \end{itemize}

1104: 

1105: 

1106: 

1107: The presence of a weak bar and the uncertainties in the

1108: amplitude of the streaming motion close to the center of this galaxy

1109: limit our ability to constrain the density profile of the dark matter

1110: halo in the center and any related cosmological issues.

1111: The other limit comes from what Figure 10 shows, namely that

1112: the likely baryonic mass components in the central region of M33 are

1113: never so small as to leave dark matter as the main ingredient of the

1114: gravitational potential. The presence of relevant baryonic components

1115: with their related uncertainties limits the ability to constraint the

1116: shape of dark matter density distribution.

1117: If a galaxy like M33 were 5-10 times further away,

1118: the non-circular motions we have found might have gone unnoticed. In

1119: this sense, M33 provides a sobering reminder that even a nearby

1120: disk-dominated system still does not provide a unique answer to this

1121: cosmological issue.

1122: 

1123: 

1124: \acknowledgments We thank the referee for her/his useful comments

1125: to the original manuscript and Josie Fenton for her work on an initial 

1126: data set taken with the first version of the DIS spectrograph before its

1127: upgrade to new CCD detectors. This paper is

1128: based on observations obtained with the Apache Point Observatory 3.5-m

1129: telescope which is owned and operated by the Astrophysical Research

1130: Consortium. We appreciate the support from the APO

1131: observatory staff and the support, both technical and financial, that

1132: has led to upgrades to DIS that have been very beneficial for this

1133: project. This publication makes use of data products from the 2MASS, 

1134: which is a joint project of the University of Massachusetts and 

1135: IPAC/California Institute of Technology funded by the National

1136: Science Foundation and the National Aeronautics and Space Administration.

1137: 

1138: \newpage

1139: 

1140: \begin{thebibliography}{}

1141: \bibitem[Abraham et al. 1999]{abr99} Abraham, R. G., Merrifield, M. R.,

1142:    Ellis, R. S., Tanvir, N. R., \& Brinchmann, J. \mnras, 308, 569

1143: \bibitem[Barton et al. 2000]{bar00} Barton, E. J., Kannappan, S. J., Kurtz,

1144:    M. J., \& Geller, M. J. 2000, \pasp, 112, 367

1145: \bibitem[Block et al. 2004]{blo04} Block, D. L., Freeman, K. C., Jarrett, T. H.,

1146: Puerari, I., Worthey, G., Combes, F., \& Groess, R. 2004, \aap, 425, L37

1147: \bibitem[Boker et al. 2002]{bok02} Boker, T., Laine, S., van der 

1148:     Marel, R. P., Rix, H. W., Ho, L. C., \& Shields, J. C. 2002, \aj, 123, 1329

1149: \bibitem[Bothun 1992]{bot92} Bothun, G. D. 1992, \aj, 103, 104

1150: \bibitem[Buta 2001]{but01} Buta, R., Ryder, S. D., Madsen, G. J., 

1151:     Wesson, K., Crocker, D. A., \& Combes, F. 2001, \aj, 121, 225

1152: \bibitem[Colin \& Athanassoula 1981]{col81} Colin, J., \& Athanassoula,

1153:     E. 1981, \aap, 97, 63

1154: \bibitem[Corbelli \& Salucci 2000]{cor00} Corbelli, E., \& Salucci, P.

1155:    2000, \mnras, 311, 441

1156: \bibitem[Corbelli 2003]{cor03} Corbelli, E. 2003, \mnras, 342, 199

1157: \bibitem[Crockett et al. 2006]{cro06} Crockett, N. R., Garnett, D. R.,

1158:    Massey, P., \& Jacoby, G. 2006, \apj, 637, 741

1159: \bibitem[Deul \& van der Hulst 1987]{deu87} Deul, E. R., \& van der Hulst, J. M.

1160:     1987, \aaps, 67, 509

1161: \bibitem[Elmegreen et al. 1992]{elm92} Elmegreen, B. G.,

1162:    Elmegreen, D. M., \& Montenegro, L. 1992, \apjs, 79, 37

1163: \bibitem[Fathi et al., 2007]{fat07} Fathi, K., Beckman, J. E., Zurita, A., 

1164:    Relano, M., Knapen, J. H., Daigle, O., Hernandez, O., \& Carignan, C. 2007,

1165:    \aap, in press (astro-ph/0703018)

1166: \bibitem[Ferrarese et al. 2006]{fer06} Ferrarese, L., et al., 2006, \apjl,

1167:     644, L21

1168: \bibitem[Freedman et al. 2001]{fre01} Freedman, W. L., Wilson, C. D., \&

1169:     Madore, B. F. 1991, \apj, 372, 455

1170: \bibitem[Gordon et al. 1999]{gor99} Gordon, K. D., Hanson, M. M., Clayton, G. C.,

1171:     Rieke, G. H., \& Misselt, K. A. 1999, \apj, 519, 165

1172: \bibitem[Gebhardt et al. 2001]{geb01} Gebhardt, K., et al., 2001, \aj, 122,

1173:     2469

1174: \bibitem[Heyer et al. 2004]{hey04} Heyer, M. H., Corbelli, E.,  Schneider, S. E.,

1175:    \& Young, J. S. 2004, \apj, 602, 723

1176: \bibitem[Kormendy \& McClure 1993]{kor93} Kormendy, J., \& McClure, R. D.

1177:     1993, \aj, 105, 1793

1178: \bibitem[Kurtz \& Mink 1998]{kur98} Kurtz, M. J. and Mink, D. J. 1998, \pasp,

1179:    110, 934

1180: \bibitem[Lauer et al. 1998]{lau98} Lauer, T. R., Faber, S. M., Ajhar, E. A., 

1181:     \& Grillmair, C. J. 1998, \aj, 116, 2263

1182: \bibitem[Magrini et al. 2007]{mag07} Magrini, L., Corbelli, E., \& Galli, D. 2007

1183:     \aap, submitted

1184: \bibitem[Merrit et al. 2001]{mer01} Merritt, D., Ferrarese, L., \& Joseph, C. L.

1185:     2001, Science, 293, 111

1186: \bibitem[Miller \& Smith 1992]{mil92} Miller, R. H., \& Smith, B. F. 1992,

1187:     \apj, 393, 508

1188: \bibitem[Milosavljevic 2004]{mil04} Milosavljevic, M. 2004, \apjl, 605, L13

1189: \bibitem[Minniti et al. 1993]{min93} Minniti, D., Olszewski, E. W., \& 

1190:    Rieke, M. 1993, \apjl, 410, L79

1191: \bibitem[Navarro et al. 1997]{nav97} Navarro, J. F., Frenk, C. S. F., 

1192:       \& White, S. D. M. 1997, \apj, 490, 493

1193: \bibitem[Newton 1980]{new80} Newton, K. 1980, \mnras, 160, 689

1194: \bibitem[Portegies-Zwart et al. 2003]{por03} Portegies Zwart, S. F., McMillan, 

1195:    S. L. W., \& Gerhard, O. 2003, \apj, 593, 352

1196: \bibitem[Regan \& Vogel 1994]{reg94} Regan M. W.,  \& Vogel S. N., 1994, \apj, 

1197:    434, 536

1198: \bibitem[Rubin \& Ford 1980]{rub80} Rubin, V. C. \& Ford, W. K. 1985, 

1199:    Carnegie Inst. Washington Ann. Rept., No. 85, p. 66

1200: \bibitem[Spekkens \& Sellwood 2007]{spe07} Spekkens, K., \& Sellwood, J.A. 2007,

1201:    \apj, in press (astro-ph/0703688)

1202: \bibitem[Thilker et al. 2005]{thi05} Thilker, D. A. et al. 2005, \apjl, 619, 67

1203: \bibitem[Tonry \& Davis 1979]{ton79} Tonry, J. L. \& Davis, M. 1979, \aj, 

1204:    43, 393

1205: \bibitem[van der Marel et al. 2006]{vdm06} van der Marel, R. P., Rossa, J., 

1206:   Walcher, C. J., Boeker, T. Ho, L. C., Rix, H.-W., \& Shields, J. C, 2007, 

1207:     to appear in 

1208:     the Proceedings of IAU Symposium 241 "Stellar Populations as Building Blocks 

1209:     of Galaxies", A. Vazdekis and R. Peletier., eds. (astro-ph/0702433) 

1210: \bibitem[Wilson \& Scoville 1989]{wil89} Wilson C. D., \& Scoville, N. 1989, 

1211:    \apj 347, 743

1212: \bibitem[Wilson \& Scoville 1990]{wil90} Wilson, C. D., Scoville, N. 1990,

1213:     363, 435

1214: \end{thebibliography}

1215: 

1216: \newpage

1217: 

1218: \begin{figure}

1219: \includegraphics[width=85mm]{figure1a.ps}

1220: \includegraphics[width=85mm]{figure1b.ps}

1221: \caption{\label{fig1}$Left$ Blue image of the central region 

1222: of M33 obtained with the 

1223: ARC 3.5m telescope through a 100\AA\ wide filter centered at 

1224: 4750\AA. M33's disk has a position angle of 22$^\circ$. The area to 

1225: the right of the major axis is the near side which may explain why dust 

1226: clouds are best seen there. $Right$ H$\alpha$ image of the same

1227: region. The continuum emission has been subtracted from this

1228: image. The various slit positions at which spectra were obtained are

1229: shown. They are, anti-clockwise from top, PA=18$^\circ$,22$^\circ$,

1230: 59$^\circ$,96$^\circ$,112$^\circ$,130$^\circ$,146$^\circ$, and 160$^\circ$. 

1231: In addition to the bright HII regions, there is diffuse ionized

1232: gas present across the region, enabling measurements of emission line

1233: velocities throughout each slit.

1234: North is up and East is left in both images. The frames measure

1235: about 280$''$ on the side, where 1$''$ corresponds to 4~pc at 840~kpc distance.}

1236: \end{figure}

1237: 

1238: \begin{figure} 

1239: \includegraphics[width=124mm]{fig2.eps} 

1240: \caption{\label{fig2}$Bottom~panel$. Radial

1241: velocities of the ionized gas in km s$^{-1}$ along PA=22$^\circ$, the nominal

1242: major axis, as a function of projected distance (D) from the center (nuclear

1243: star cluster) in arcmin ($1'$=240~pc). 

1244: Filled circles refer to simultaneous fits of the 4

1245: red emission lines, filled squares to H$\alpha$ measurements, open

1246: squares to [NII]6584, open triangles to [SII]6716, and inverted "Y" to

1247: [SII]6731. Note the flat velocity profile and sudden velocity jump

1248: across the central region which occurs over just a few arcsec (see

1249: also Fig.7 and Fig.8). The vertical dashed line marks the position of the star

1250: cluster in this and subsequent panels.  $Top~panel$. Stellar radial

1251: velocities along the same PA derived from the Ca triplet (open star symbols) 

1252: and blue spectra (cross and asterix symbols, each relative to one observing

1253: run). Stellar velocities show a smoother increase with distance from the center 

1254: than ionized gas velocities.}  

1255: \end{figure}

1256:  

1257: \begin{figure}

1258: \includegraphics[width=124mm]{fig3.eps}

1259: \caption{\label{fig3} $Bottom~panel$. Radial velocities  

1260: of the ionized gas along PA=112$^\circ$, the nominal minor axis, 

1261: as a function of projected distance from the center. 

1262: Symbols as defined in Figure 2. Notice the strong non-zero velocities 

1263: out to a radius of about 1$'$ on each side.

1264: $Top~panel$. Stellar velocities on the minor axis. Different symbols refer to 

1265: velocities derived from correlation technique on the Mg side of blue spectra 

1266: obtained during different observing runs. Average 

1267: stellar velocities seem higher than V$_{sys}$ = $-180$ km~s$^{-1}$ in

1268: those regions where gas velocities are also higher. 

1269: }

1270: \end{figure}

1271: 

1272: \begin{figure}

1273: \includegraphics[width=124mm]{fig4.eps}

1274: \caption{\label{fig4} From $top$ to $bottom$: Radial velocities of 

1275: the ionized gas along PA= 59$^\circ$, 96$^\circ$, and 146$^\circ$. 

1276: Symbols as defined in Figure 2.  The regions with peculiar velocities are 

1277: again confined to $\pm$ 1-1.5$'$ from the center.}

1278: \end{figure}

1279: 

1280: \begin{figure}

1281: \includegraphics[width=124mm]{fig5.eps}

1282: \caption{\label{fig5}

1283: Radial velocity measurements for the emission line gas at

1284: PA=130$^\circ$, 146$^\circ$, and 160$^\circ$. For PA=146$^\circ$, 

1285: the 2005 H$\alpha$ and 2006 H$\beta$ (open circles) measurements are

1286: shown. Zero points for the H$\beta$ measurements have been

1287: determined through comparison with  PA=146$^\circ$ H$\alpha$

1288: velocities. 

1289: }

1290: \end{figure}

1291: 

1292: \begin{figure}

1293: \includegraphics[width=124mm]{fig6.eps}

1294: \caption{\label{fig6}From $top$ to $bottom$: Radial velocities of 

1295: stars along PA= 130$^\circ$, 146$^\circ$, and 160$^\circ$

1296: from cross-correlation with template star spectrum on the Mg side

1297: wavelength range. The nuclear cluster is centered at -180~km~s$^{-1}$.

1298: }

1299: \end{figure}

1300: 

1301: 

1302: 

1303: \begin{figure}

1304: \includegraphics[width=144mm]{fig7.eps}

1305: \caption{\label{fig7}

1306: $Top~panel$: Emission line velocities observed along the major axis

1307: (PA=22$^\circ$) and along

1308: slit positions parallel to it. NE is to the left and SW

1309: to the right.  Open circles refer to H$\beta$ lines, filled circles

1310: to averages over the four red lines.

1311: The perpendicular distance from the major axis axis for

1312: each cut is listed, as well as the velocity offsets added to separate

1313: the curves. The dashed lines indicate the systemic velocity and location 

1314: of the center. 

1315: $Bottom$: Radial velocities measured along a slit position

1316: parallel to the major axis but offset 142$''$ in the SE direction along

1317: the minor axis. The intersection with the minor axis is shown. The

1318: points around -3$'$ are centered on a bright HII region in the disk. }

1319: \end{figure}

1320: 

1321: 

1322: \begin{figure} 

1323: \includegraphics[width=134mm]{fig8.eps} 

1324: \caption{\label{fig8}Gas radial velocities along

1325: the nominal major axis, PA=22$^\circ$, in the innermost 

1326: region of M33. Lines have been measured

1327: in single frames and pixel by pixel. The spatial sampling is 0.54$''$

1328: but the typical seeing was about 1.5$''$ so neighboring points are not

1329: independent.  Filled squares correspond to average values of the

1330: velocity from measurements in different frames. Averages are not

1331: computed when line velocities from different frames are inconsistent

1332: i.e. when their dispersion around the mean (errorbars) is larger than

1333: 25~km~s$^{-1}$. In this case and when lines are hardly measurable due

1334: to blending with stellar absorption or to intrinsic weakness, single

1335: frame measurements are plotted (open squares).  }  

1336: \end{figure}

1337: 

1338: \begin{figure}

1339: \includegraphics[width=134mm]{fig9.eps}

1340: %\epsscale{.80}

1341: %\plotone{fig2.eps}

1342: \caption{\label{fig9}Same as Figure 8 but for the minor axis, PA=112$^\circ$.}

1343: \end{figure}

1344: 

1345: 

1346: \begin{figure}

1347: \includegraphics[width=124mm]{fig10.eps}

1348: %\epsscale{120}

1349: \caption{\label{fig10} Rotational velocities assuming 

1350: PA=22$^\circ$ for the major axis direction and $i=52^\circ$.

1351: Triangles are gas average velocities traced by optical emission lines in the   

1352: red and in the blue frame along PA=22$^\circ$. 

1353: Errorbars indicate the uncertainties

1354: in the mean in each bin. Open stars are stellar rotational

1355: velocities along the same direction, measured using the MgII side of the blue 

1356: frame only. Filled circles show the CO J=1-0 rotation curve from azimuthal averages 

1357: of data points within $\pm45^\circ$ of major axis (Corbelli 2003). 

1358: Open squares are 

1359: from emission lines in one red frame along the south-west side of major axis. 

1360: The continuous line shows the resulting best fit to the rotation curve defined 

1361: by the circled symbols in this inner region and by data shown 

1362: in Corbelli (2003) at larger radii. Dashed lines mark the contribution of the 

1363: various mass components to the rotation curve assuming pure circular motion:

1364: stellar and gaseous disk, bulge or nuclear cluster(nc), dark matter halo. 

1365: In $(a)$ the fitted rotation curve is traced by optical emission lines 

1366: inside 0.2~kpc and by the CO and 21-cm lines at larger radii.

1367: In $(b)$ the fitted rotation curve is traced by averaged optical emission 

1368: lines from 0.2 to 0.8~kpc and by CO and 21-cm lines further out. 

1369: Inside 0.2~kpc optical emission lines are assumed to trace

1370: the shear or non circular component due to a possible bar (see Section 4.3).

1371: In both cases, $(a)$ and $(b)$, the dynamical model  

1372: for pure circular motion predicts  rotational velocities which

1373: are not constant but rise continuously outside 

1374: the nuclear cluster region.}

1375: \end{figure}

1376: 

1377: \begin{figure}

1378: \includegraphics[width=85mm]{fig11a.ps}

1379: \includegraphics[width=85mm]{fig11b.ps}

1380: \caption{\label{fig11} The left panel shows a 25'(horizontal) by 23.5' 

1381: (vertical) region of the 2MASS K-band image of M33 rotated by -22$^\circ$ 

1382: to align the disk's major axis with the vertical axis. 

1383: The grey-scale stretch is chosen to emphasize the oval distortion 

1384: of the light in the inner 3' diameter region. 

1385: The right panel shows 

1386: the same image deprojected to face-on by stretching the X-axis

1387: by 1/cos(i=52$^\circ$). The deprojection is only representative of the face-on

1388: light distribution if all components are intrinsically disk-like.

1389: }

1390: \end{figure}

1391: 

1392: \begin{figure}

1393: \includegraphics[width=124mm]{fig12.eps}

1394: \caption{\label{fig12} The angular frequency, $\Omega$, derived  from

1395: the rotation curve of Figure 10$(b)$ (filled circles) and the

1396: $\Omega \pm \kappa/2$ curves. The intersection between the latter curves

1397: and $\Omega_p$ defines the location of the inner and

1398: outer Lindblad resonances. The shaded region indicates the possible 

1399: values of the bar pattern speed $\Omega_p$ if the bar ends at its 

1400: own corotation.

1401: }

1402: \end{figure}

1403: 

1404: \end{document}

1405: 

1406: %%

1407: