Gas distribution, kinematics and star formation in faint dwarf galaxies

Monthly Notices of the Royal Astronomical Society, Feb 2006

We compare the gas distribution, kinematics and the current star formation in a sample of 10 very faint (−13.37 < MB < −9.55) dwarf galaxies. For five of these galaxies we present fresh, high-sensitivity, Giant Metrewave Radio Telescope H i 21-cm observations. We find that the large-scale H i distribution in the galaxies is typically irregular and clumpy, with the peak gas density rarely occurring at the geometric centre. We also find that the velocity fields for all the galaxies have an ordered component, although in general, the patterns seen do not fit that expected from a rotating disc. For all our galaxies we construct maps of the H i column density at a constant linear resolution of ∼300 pc; this forms an excellent data set to check for the presence of a threshold column density for star formation. We find that while current star formation (as traced by Hα emission) is confined to regions with relatively large [NH i > (0.4–1.7) × 1021 cm−2] H i column density, the morphology of the Hα emission is in general not correlated with that of the high H i column density gas. Thus, while high column density gas may be necessary for star formation, in this sample at least, it is not sufficient to ensure that star formation does in fact occur. We examine the line profiles of the H i emission, but do not find a simple relation between regions with complex line profiles and those with ongoing star formation. Our sample includes examples of regions where there is ongoing star formation, but the profiles are well fitted by a single Gaussian, as well as regions where there is no star formation but the line profiles are complex. Finally, we examine the very fine scale (∼20–100 pc) distribution of the H i gas, and find that at these scales the emission exhibits a variety of shell-like, clumpy and filamentary features. The Hα emission is sometimes associated with high-density H i clumps, sometimes the Hα emission lies inside a high-density shell, and sometimes there is no correspondence between the Hα emission and the H i clumps. In summary, the interplay between star formation and gas density in these galaxies does not seem to show the simple large-scale patterns observed in brighter galaxies.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

http://mnras.oxfordjournals.org/content/365/4/1220.full.pdf

Gas distribution, kinematics and star formation in faint dwarf galaxies

Ayesha Begum 1 Jayaram N. Chengalur 1 I. D. Karachentsev 0 S. S. Kaisin 0 M. E. Sharina 0 0 Special Astrophysical Observatory, Nizhnii Arkhys 369167, Russia 1 National Centre for Radio Astrophysics , Post Bag 3, Ganeshkhind, Pune 411 007, India A B S T R A C T We compare the gas distribution, kinematics and the current star formation in a sample of 10 very faint (13.37 < M B < 9.55) dwarf galaxies. For five of these galaxies we present fresh, high-sensitivity, Giant Metrewave Radio Telescope H I 21-cm observations. We find that the large-scale H I distribution in the galaxies is typically irregular and clumpy, with the peak gas density rarely occurring at the geometric centre. We also find that the velocity fields for all the galaxies have an ordered component, although in general, the patterns seen do not fit that expected from a rotating disc. For all our galaxies we construct maps of the H I column density at a constant linear resolution of 300 pc; this forms an excellent data set to check for the presence of a threshold column density for star formation. We find that while current star formation (as traced by H emission) is confined to regions with relatively large [N H I > (0.4-1.7) 1021 cm2] H I column density, the morphology of the H emission is in general not correlated with that of the high H I column density gas. Thus, while high column density gas may be necessary for star formation, in this sample at least, it is not sufficient to ensure that star formation does in fact occur. We examine the line profiles of the H I emission, but do not find a simple relation between regions with complex line profiles and those with ongoing star formation. Our sample includes examples of regions where there is ongoing star formation, but the profiles are well fitted by a single Gaussian, as well as regions where there is no star formation but the line profiles are complex. Finally, we examine the very fine scale (20-100 pc) distribution of the H I gas, and find that at these scales the emission exhibits a variety of shell-like, clumpy and filamentary features. The H emission is sometimes associated with high-density H I clumps, sometimes the H emission lies inside a high-density shell, and sometimes there is no correspondence between the H emission and the H I clumps. In summary, the interplay between star formation and gas density in these galaxies does not seem to show the simple large-scale patterns observed in brighter galaxies. - In the currently popular hierarchical models of galaxy formation, star formation starts in small objects; these in turn later merge to form larger galaxies. In such a model, extremely small nearby galaxies are likely candidate primeval galaxies, in the sense that they may represent the earliest units of star formation in the Universe. There is some observational support for these models, even in the very local Universe, viz. (i) the Milky Way itself appears to be still growing via the accretion of small companions like the Sagittarius dwarf galaxy (see e.g. Majewski et al. 2003), and (ii) nearby dwarf galaxies have stellar populations that are at least as old as the oldest stars in the Milky Way (see Grebel 2005, for a recent review). In detail, however, the star formation history of nearby dwarf galaxies appears to be extremely varied. At the two extreme ends, dwarf spheroidals have little gas or ongoing star formation while the relatively rare dwarf irregulars are gas-rich and also generally have measurable ongoing star formation. Their past star formation histories also appear to have been different at a given luminosity dwarf spheroidals are more metal-rich than dwarf irregulars, indicative of rapid chemical enrichment in dwarf spheroidals in the past (Grebel 2004). Why is it that dwarf irregulars, despite having a substantial reservoir of gas, have resisted converting it into stars? What keeps the gas in dwarf irregulars from collapse? It is widely believed that the smallest dwarf irregular galaxies have chaotic gas velocity fields (e.g. Lo, Sargent & Young 1993), in this case the crucial question then becomes, what sustains these chaotic gas motions? In this context, it is interesting to note that for galaxies which have been observed with sufficient sensitivity and velocity resolution, the velocity field has invariably turned out to have a measurable ordered component, (Begum, Chengalur & Hopp 2003; Young et al. 2003; Begum & Chengalur 2004). Does this generally hold for extreme dwarf irregulars, or do some of them genuinely have no ordered components in their velocity fields? Irrespective of the exact nature of the velocity fields, the question of why dwarf irregulars have been unable to convert their gas into stars remains. In spiral galaxies, the current star formation rate (SFR) appears to depend on at most two parameters: (i) the gas surface density and (ii) some measure of the dynamical time. In practice, models which depend only on the gas surface density, such as the Schmidt star formation law, or those which depend on both these parameters, such as the instability criterion of Toomre (Toomre 1964), appear to provide an equally good fit to the observations (Kennicutt 1998b). For irregular galaxies, Skillman (1987) has proposed that star formation occurs only above a threshold column density, and that this threshold may be related to a critical amount of dust shielding required for molecular gas formation. Can any of these models be extrapolated to the faintest dwarf irregulars? We present here deep, high velocity resolution (1.6 km s1) Giant Metrewave Radio Telescope (GMRT) H I observations, as well as H observations of a sample of faint (MB > 13.0 mag) galaxies, aimed at addressing the above issues. The rest of this paper is divided as follows. The dwarf galaxy sample is presented in Section 2, the GMRT observations are detailed in Section 3, while the results are presented in Section 4 and discussed in Section 5. 2 D WA R F G A L A X Y S A M P L E The optical properties of our sample of 10 galaxies are given in Table 1. Fresh H I observations for five galaxies in the sample, viz. KDG 52, UGC 4459, CGCG 269 049, UGC 7298 and KK 230, are presented in this paper. GMRT H I data for KK 44 (Camelopardalis B), GR 8 and DDO 210 were presented in our previous papers (Begum et al. 2003; Begum & Chengalur 2003, 2004), although we include here fresh maps and measurements at angular scales that are relevant to the issues discussed in this paper. GR 8 and DDO 210 (2) Right ascension (J2000) 04h53m06.9s 08h23m56.0s 08h34m06.5s 09h59m26.4s 12h15m46.7s 12h16m28.6s 12h58m40.4s 14h07m10.7s 19h29m59.0s 20h46m51.8s +6705 57 +7101 46 +6610 45 +3044 47 + 5223 15 +5213 38 +1413 03 +3503 37 1740 41 1250 53 were also observed with the Very Large Array (VLA); the VLA data for these galaxies are presented in Young et al. (2003). VLA H I data for Sag DIG and Leo A were obtained from the VLA archive. These observations have been discussed earlier by Young & Lo (1996) (Leo A) and Young & Lo (1997) (Sag DIG); once again we present here only those maps and measurements that are relevant to this paper. 3 O B S E RVAT I O N S A N D A N A LY S I S 3.1 Optical observations and data analysis H observations of some of our sample galaxies, viz. KK 44, KDG 52, CGCG 269 049, UGC 7298 and KK 230, were carried out at the 6-m Special Astrophysical Observatory (SAO) telescope using a 2048 2048-pixel CCD camera. The scale was 0.36 arcsec pixel1, and the total area imaged was 6 6 arcmin2. The H + [N II] emission-line fluxes were obtained by observing each galaxy through two filters: a narrow (75 ) interference filter centred on 6567 , and a middle-width filter ( = 6063 , = 167 ) to determine the nearby continuum level. The integration times were 2 300 s in the middle-width filter and 2 600 s in H. Because the range of radial velocities was small, we used the same H filter for all objects. The images were bias-subtracted and flat-fielded following standard procedures. After flat-fielding, the next step was to subtract the sky emission from both continuum and narrow-band filter images. The continuum filter images were then scaled relative to the narrow-band images using 510 unsaturated stars, and then subtracted from the narrow-band filter images. The continuum-subtracted H images were flux-calibrated using observations from the same night of two or more of Feiges photometric standards. Corrections for the Galactic extinction were made assuming A(H) = 2.32 E (B V ) using the data from Schlegel, Finkbeiner & Davis (1998). The SFR for these galaxies was calculated from the derived H luminosities, using the conversion factor from Kennicutt (1998a) SFR = 7.9 1042 L(H) M yr1. The calculated SFR for our sample galaxies is given in Table 5. In the case of KDG 52 and KK 230, no H emission was detected; the derived limits on the SFR for these galaxies are also listed in Table 5. For Leo A, Sag DIG and GR 8, H images were downloaded from the NASA/IPAC Extragalactic Database (NED). Details of these images can be found in Hunter & Elmegreen (2004). The H image of UGC 4459 was kindly provided by U. Hopp; details can be found in Schulte-Ladbeck & Hopp (1998). For DDO 210, van Zee (2000) detected a single source of H emission in the galaxy; however, follow-up observations suggested that it does not arise in a normal H II region, but probably comes from dense outflowing material from an evolved star. In all the figures of DDO 210 in this paper, we show the location of this emission by a star, but caution the reader that it may not actually represent a star-forming region. Except for KK 230, broad-band optical observations of all our sample galaxies were available in the literature. For KK 230, V- and I-band Hubble Space Telescope (HST) Advanced Camera for Surveys (ACS) images were used to obtain the total magnitude of the galaxy. The derived magnitude is I (R < 40 arcsec) = 15.6 0.15 mag, and the integrated (V I ) colour inside the same radius is 0.90. Assuming a typical color (B V ) = 0.50 for KK 230, we estimated its integrated blue magnitude to be B = 17.0 0.25 mag. 3.2 H I observations and data analysis H I 21-cm observations of KDG 52, UGC 4459, CGCG 269 049, UGC 7298 and KK 230 were conducted with the GMRT (Swarup et al. 1991) between 2001 November and 2002 November. KK 44, GR 8 and DDO 210 were also observed with the GMRT; details can be found in Begum et al. (2003) and Begum & Chengalur (2003, 2004). Data for Sag DIG and Leo A were obtained from the VLA archive. These observations are also discussed in Young & Lo (1996, 1997). Here we briefly describe only the fresh GMRT observations. For all galaxies, the observing bandwidth of 1 MHz was divided into 128 spectral channels, yielding a spectral resolution of 7.81 kHz (velocity resolution of 1.65 km s1). The setup for the observations is given in Table 2. The flux and bandpass calibrations were done using 3C48, 3C147 and 3C286. The phase calibration was done once in every 30 min by observing the VLA calibrator sources 0831 + 557 (UGC 4459), 1216 + 487 (UGC 7298), 1216 + 487 (CGCG 269 049), 3C286 (KK 230) and 0834 + 555 (KDG 52). The galaxies UGC 7298 and CGCG 269 049 are close in space (12 arcmin) as well as in velocity, hence both were included in a single GMRT pointing (the field of view of the GMRT 24 arcmin). The data were reduced in the usual way using standard tasks in classic AIPS. For each run, bad-visibility points were edited out, after which the data were calibrated. The GMRT does not do online Doppler tracking any required Doppler shifts have to be applied during the offline analysis. For UGC 7298 and CGCG 269 049, the differential Doppler shift over our observing interval was much less than the channel width, hence, there was no need to apply any offline correction. On the other hand, the differential shifts for UGC 4459, KK 230 and KDG 52 were significant, hence, for each of these galaxies, the calibrated (u, v) data set for each day was shifted in the frequency space to the heliocentric velocity of the galaxy, using the task CVEL in AIPS. For each galaxy, data for all the runs were then combined using the AIPS task DBCON. The GMRT has a hybrid configuration (Swarup et al. 1991) which simultaneously provides both high angular resolution (3 arcsec, if one uses baselines between the arm antennas) as well as sensitivity to extended emission (from baselines between the antennas in the central array). Data cubes were therefore made using various (u, v) cut-offs to get the images of H I emission at various spatial resolutions (see Table 2 for details). Except for the highest-resolution H I data cubes for each galaxy, all the data cubes were deconvolved using the AIPS task IMAGR. For the highest-resolution data cubes in each galaxy, the signal-to-noise ratio (S/N) was too low for CLEAN to work reliably. Despite this, the low S/N of the images implies that the inability to deconvolve does not greatly degrade the dynamic range or fidelity of these images. The morphology of the emission in these galaxies should hence be accurately traced, apart from an uncertainty in the scaling factor (this essentially arises because the main effect of deconvolving weak emission at about the noise level corresponds to multiplying by a scale factor; see e.g. J orsater & van 1995; Rupen 1999). Continuum images were also made for all the galaxies by averaging the line free channels. No extended (26 22 arcsec2) or compact (3 3 arcsec2) emission was detected from any of the galaxies. The 3 limits for each galaxy are given in Table 2. Moment maps were made from the data cubes using the AIPS task MOMNT. Maps of the velocity field and velocity dispersion were also made in GIPSY using single Gaussian fits to the individual profiles. The velocity field produced by Gaussian fitting is in reasonable agreement with that obtained from moment analysis. The velocity dispersion ( obs), as estimated by fitting single Gaussian component to the line profiles is given in Table 3. In all cases, no measurable variation of velocity dispersion was seen (within the error bars) across each galaxy. This lack of substantial variation of across each galaxy is typical of such faint dwarf irregular galaxies (e.g. Skillman et al. 1988; Begum et al. 2003; Begum & Chengalur 2004). As discussed in more detail in Section 5.3, single Gaussian profiles are not necessarily a good fit throughout the galaxy; there are regions where the emission profile is skewed or is otherwise more complex than a single Gaussian. CGCG 269 049 2002 November (60)130 42 39, 26 23 723 671, 447 396 1.7, 1.5 16 15, 6 6 275 258, 103 103 1.3, 0.9 45 38, 29 27 777 656, 500 466 1.9, 1.6 18 16, 3 3 310 276, 52 52 1.4, 1.2 42 39, 28 24 692 642, 461 396 2.0, 1.8 18 17, 4 3 297 280, 66 50 1.7, 1.2 42 37, 26 24 857 755, 530 490 2.0, 1.8 16 15, 4 4 326 306, 82 82 1.6, 1.1 48 45, 34 31 442 415, 313 286 1.6, 1.4 26 24, 4 3 240 221, 37 28 1.2, 0.8 21.4 (1.0) 18.8 (0.7) 20.6 (1.7) 29.6 (1.8) 26.6 (2.2) 21.4 (1.7) 26.0 (1.2) 17.0 (2.0) 19.4 (0.8) 19.1 (1.0) 7.3 (0.8) 9.5 (1.3) 9.0 (1.0) 9.0 (1.6) 9.5 (1.0) 8.5 (1.3) 9.0 (0.8) 7.5 (0.5) 7.5 (1.7) 6.5 (1.0) 4 R E S U LT S 4.1 Large-scale H I distribution and kinematics The global H I profiles for our sample galaxies, obtained from the coarsest-resolution data cubes (see Table 2), are shown in Fig. 1. Columns (2)(7) in Table 3 list the parameters derived from the global H I profiles. Column (1) gives the galaxy name, Column (2) gives the integrated H I flux along with the error bars, Column (3) gives the velocity width at 50 per cent of the peak ( V 50), along with the error bars, Column (4) gives the central heliocentric velocity (V ) and its error bars, Column (5) gives the H I mass along with its error bars, Column (6) gives the H I mass-to-light ratio (M H I/L B), Column (7) gives the ratio of the GMRT flux to the single-dish flux (FI/FISD). The single-dish fluxes for all the galaxies are taken from the tabulation in Karachentsev et al. (2004). In the case of CGCG 269 049, single-dish data are not available. The parameters measured from the GMRT H I profiles are in good agreement with those values obtained from the single-dish observations, in particular, the H I flux measured at the GMRT agrees with the single-dish flux for all the galaxies. This indicates that no flux was missed because of the missing short spacings in our interferometric observations. Column (8) shows the velocity dispersion, along with error bars, as measured from a single Gaussian fit to the line profiles, Column (9) represents the H I radius at a column density of 1019 atoms cm2, Column (10) gives the inclination as measured from the H I moment zero maps, Column (11) gives the ratio of the H I diameter to the Holmberg diameter. For all the galaxies the H I emission extends to 23 times the optical diameter, a typical ratio for dwarf irregular galaxies. The integrated H I emission of our sample galaxies overlaid on the optical Digitized Sky Survey (DSS) images is shown in Figs 3(a) 7(a). The H I distribution in CGCG 269 049 and KK 230 is dominated by a single clump of high column density, while the H I distribution in UGC 4459 and UGC 7298 is concentrated in two high-column density regions, separated by a low-column density region in the centre. In the case of KDG 52, the H I is distributed in a clumpy, incomplete ring. Inclinations (i H I) of our sample galaxies (except for KDG 52) were estimated from the H I moment zero maps by fitting elliptical annuli to the two lowest-resolution images. For KDG 52, only the lowest-resolution H I distribution is sufficiently smooth to be used for ellipse fitting. For all other galaxies, the inclinations derived from these two resolution images match within the error bars. The estimated inclination for each galaxy (assuming an intrinsic thickness q o = 0.2) is tabulated in Table 3. Comparing this value to the optical inclination (Table 1) shows that the two inclinations are in agreement for UGC 4459 and KDG 52, whereas for the rest of the sample galaxies the optical inclination is found to be either much higher (UGC 7298 and CGCG 269 049) or lower (KK 230) than the inclinations derived from the H I morphology. Using the derived H I inclination, the de-projected H I radial surface mass density (SMD) profiles for each galaxy were obtained by averaging the H I distribution over elliptical annuli in the plane of the galaxy. The derived SMD profiles for each galaxy are given in Fig. 2. We next discuss in detail the large-scale H I distribution and kinematics for the five galaxies for which fresh GMRT observations are presented in the current paper. For similar details on the other galaxies in our sample, the reader is referred to Begum et al. (2003), Begum & Chengalur (2003, 2004) and Young & Lo (1996, 1997). 4.2 Notes on individual galaxies 4.2.1 KDG 52 KDG 52 (also called M81DwA) was discovered by Karachentseva (1968) and was later detected in H I by Lo & Sargent (1979). The neutral hydrogen in this galaxy is distributed in a clumpy, broken ring surrounding the optical emission (Fig. 3a). The central H I hole has a diameter of 40 arcsec2 (688 pc); similar central H I holes are seen in other faint dwarf galaxies (e.g. Sag DIG, Young & Lo 1997; DDO 88, Simpson, Hunter & Knezek 2005). The H I hole is not exactly centred on the optical emission; the H I column density at the eastern side of the optical emission is N H I 4 1020 atoms cm2, while the rest of the optical emission lies inside the H I hole. Prior to this work, there have been two H I interferometric studies of KDG 52. It was observed with the Westerbork Synthesis Radio Telescope (WSRT) by Sargent, Sancisi & Lo (1983) with a velocity resolution of 8 km s1 and later re-observed with a high-velocity resolution in the C array of the VLA (Westpfahl et al. 1999). The overall morphology of the earlier images compares well with that of our image. Our coarsest-resolution H I distribution and velocity field (not shown) shows faint emission in the centre and in the northern region of the galaxy, a feature that is not visible at the higher resolutions. One may suspect that this H I emission is not real but is the result of beam smearing. To check for this possibility, the individual channel maps in the 42 39-arcsec2 data cube were inspected. In the channel maps, the peak of the diffuse emission in the central as well as in the northern region of the galaxy occurs at the same heliocentric velocity as that of nearby H I clumps, suggesting that they may arise due to beam smearing. As a further check, the CLEAN components from the 42 39-arcsec2 resolution data cube were convolved with a smaller restoring beam of 30 30 arcsec2, to generate a new data cube. The diffuse emission which was visible in the 42 39-arcsec2 data cube is not seen in the channel maps in this cube, that is, no CLEAN components were found in the region of diffuse emission. Finally, the H I flux measured from a genuine 30 30arcsec2 resolution data cube [i.e. made from the visibility data by applying the appropriate (u, v) range and taper] is the same as that measured from the 42 39-arcsec2 data cube. All these indicate that the diffuse emission in 42 39-arcsec2 is entirely due to beam smearing. The velocity field obtained from 26 23-arcsec2 resolution data cube is given in Fig. 3(b). The velocity field shows a large-scale gradient across the galaxy with a magnitude of 1.7 km s1 kpc1. However, the velocity field is clearly not consistent with pure rotation. One can still crudely estimate the maximum possible circular velocity in the following way: the velocity difference from one edge of the galaxy to the other is 6 km s1, this implies that the magnitude of any circular velocity component must be limited to Vrot sin(i ) 3 km s1. Puche & Westpfahl (1994) have tried to model this velocity field, and find that a combination of rotation (with a magnitude of 7 km s1) and expansion (with a magnitude of 5 km s1) provides a reasonable fit. A similar combination of rotation and expansion was found to provide a good fit to the kinematics of another of our sample galaxies, viz. GR 8 (Begum & Chengalur 2003). KDG 52 is a member of the M81 group of galaxies. Bureau et al. (2004) have suggested that this galaxy is probably a tidal dwarf, formed through gravitational collapse of the tidal debris from the previous interactions of Holmberg II with UGC 4483. In Section 5.1, we estimate the dynamical mass of this galaxy from the virial theorem; this mass estimate implies that the galaxy has a significant amount of dark matter. This would argue against a tidal dwarf origin for KDG 52, since tidal dwarfs are generally not expected to be dark matter dominated (e.g. Braine et al. 2002). 4.2.2 UGC 4459 UGC 4459 is a member of the M81 group of galaxies. It is relatively metal-poor, with 12 + log(O/H) 7.62 (Kunth & Otlin 2000). The optical appearance of UGC 4459 is dominated by bright blue clumps, which emit copious amounts of H (Figs 4a, 9 and 12). The two high-density peaks seen in the integrated H I map coincide with these star-forming regions (Fig. 4a). The velocity field of UGC 4459 (Fig. 4b) shows a large-scale gradient (aligned along the line connecting the two star-forming regions) across the galaxy. The magnitude of the average velocity gradient across the whole H I disc is 4.5 km s1 kpc1. However, we note that the gradient is not uniform across the galaxy. The receding (south-eastern) half of the galaxy shows a rapid change in velocity with galacto-centric distance, while the approaching (northwestern) half of the galaxy shows a much more gentle gradient. UGC 4459 is a fairly isolated dwarf galaxy with its nearest neighbour UGC 4483 at a projected distance of 3.6 (223 kpc) and at a velocity difference of 135 km s1. Being a member of the M81 group, it is possible that interaction with intragroup gas could produce such disturbed kinematics. To check for this possibility, we estimated the ram pressure required to strip gas from this galaxy. The threshold condition for ram pressure stripping is given by (Gunn & Gott 1972) where IGM is the density of the intragroup medium (IGM) and v is the relative velocity of the galaxy moving through the IGM. and g are stellar and gas surface density, respectively. Taking v 190 km s1, typical for the M81 group (Bureau & Carignan 2002), and values for and g from the location in the galaxy where the velocity field begins to look perturbed, we find that the IGM volume density required to strip the interstellar medium (ISM) from UGC 4459 is n IGM 8 105 cm3. UGC 4459 is located at a projected separation of 8.4 (520 kpc) to the south-west of M81 (which we can take to be the centre of the M81 group). The n IGM required for ram pressure stripping of UGC 4459 is much higher than n IGM expected at this location (1.4 106 cm3; assuming that 1 per cent of the virial mass of the group is dispersed uniformly in a hot IGM within a sphere just enclosing UGC 4459; Bureau & Carignan 2002). Hence, it seems unlikely that the peculiar kinematics of the galaxy is due to IGM ram pressure. Given the kinematical asymmetry between the two halves of the galaxy, it is not surprising that a tilted ring fit does not give consistent results for the two halves. The difference in the peak velocities for the rotation curves derived from the two halves is 15 km s1. This difference is significant compared to the peak value of 25 km s1 obtained for the receding half of the galaxy. One can crudely estimate the maximum possible circular velocity in the following way; the velocity difference from one edge of the galaxy to the other is 18 km s1, this implies that the magnitude of any circular velocity component must be limited to V rot sin(i ) 9 km s1. Pustilnik et al. (2003) found substantial small-scale velocity gradients in the H emission along a slit placed parallel to the optical major axis (i.e. also along the direction of maximum velocity gradient in the H I velocity field), as well as a large-scale gradient, with magnitude somewhat larger than what we observe in the H I. UGC 4459 has the largest SFR of all the galaxies in our sample. Pustilnik et al. (2003) estimate very young ages (38 Myr) for the star-forming knots in the galaxy. Since they find no nearby galaxy that could have triggered this recent starburst, they suggest that it could be triggered by tidal interaction with the M81 group as a whole, or by interaction with the IGM. As we argued above, the ram pressure of the IGM is likely to be small. The velocity field of UGC 4459 is, however, qualitatively very similar to that of DDO 26 (Hunter & Wilcots 2002) and IC 2554 (Koribalski, Gordon & Jones 2003), both of which are suspected to be late-stage mergers. It seems possible therefore that UGC 4459 too represents a recent merger of two still fainter dwarfs. 4.2.3 CGCG 269 049 CGCG 269 049 is an extremely metal-poor dwarf galaxy with 12 + log(O/H) 7.43 (Kniazev et al. 2003). It is a member of the Canes Venatici I cloud. The optical emission in CGCG 269 049 shows two components, a central compact component and an outer faint extended component; both elongated in the north-west direction. The H I distribution of the galaxy also shows an elongation in the same direction. However, a misalignment of 10 is seen between the optical and the H I major axis. The H I distribution is regular and shows a slightly off-centred peak; this signature is more prominent in the high-resolution H I images. CGCG 269 049 is undergoing a burst of star formation as indicated by strong emission lines in its spectra. It has been suggested that starbursts in dwarf galaxies could be triggered by tidal interaction with a companion (Taylor 1997; Walter & Brinks 2001). While CGCG 269 049 does have a nearby companion (viz. UGC 7298 as discussed in Section 4.2.4), the H I distribution in neither of these galaxies is suggestive of tidal interaction. The velocity field of the galaxy shows a large-scale gradient, roughly aligned with the morphological major axis and with a magnitude of 5.2 km s1 kpc1 (Fig. 5b). Of all the galaxies in this subsample, CGCG 269 049 has a velocity field that most resembles that expected from a rotating disc. Substantial deviations from simple rotation can however be seen, and a tilted ring fit to the velocity field does not yield meaningful results. CGCG 269 049 has one of the highest current SFRs among our sample galaxies; H imaging of this galaxy shows a bright H core near its centre (see Figs 9 and 12). From Fig. 5(b), one can see kinks in the velocity field in the regions near this star-forming knot. Hence, it is likely that the energy input from the ongoing star formation in the galaxy is responsible for at least some of the distortions seen in the velocity field. The maximum velocity difference from one edge of the galaxy to the other is 16 km s1, hence Vrot sin(i ) 8 km s1. 4.2.4 UGC 7298 UGC 7298 is a member of the Canes Venatici I cloud of galaxies. The velocity field of UGC 7298 (Fig. 6b) shows a large-scale gradient roughly aligned with the line joining the two high-density gas clumps. The magnitude of the gradient is 3.5 km s1 kpc1. The velocity field is broadly similar to that in UGC 4459. The optical properties of the two galaxies are however very dissimilar. UGC 4459 is currently undergoing a starburst and its optical appearance is dominated by bright star-forming knots. UGC 7298, on the other hand, has a very small current star formation rate, as inferred from very faint H emission in the galaxy (Figs 9 and 12). The maximum velocity difference from one edge of the galaxy to the other is 16 km s1, implying that Vrot sin(i ) 8 km s1. 4.2.5 KK 230 KK 230, the faintest dwarf irregular galaxy in our sample, is yet another member of the Canes Venatici I cloud of galaxies (Karachentsev et al. 2003). The velocity field (Fig. 7b) shows a gradient in the eastwest direction (i.e. roughly perpendicular to the H I and optical major axis) with a magnitude of 6 km s1 kpc1. Even apart from this misalignment, the velocity field bears little similarity to that expected from a rotating axisymmetric disc. The origin of the velocity gradient in KK 230 is rather puzzling. This galaxy has no measurable ongoing star formation, and no H emission was detected from the galaxy. It also lies at the periphery of the Canes Venatici I cloud group of galaxies; Karachentsev et al. (2004) found its tidal index to be 1.0, meaning that it is a fairly isolated galaxy. This, along with the fairly regular H I distribution, makes it unlikely that tidal forces are responsible for the observed velocity field. The maximum velocity difference from one edge of galaxy to the other is 10 km s1, which gives V rot sin(i ) 5 km s1. 5 D I S C U S S I O N 5.1 Dynamical mass of our sample galaxies As discussed above, large-scale systematic gradients are seen across all the newly mapped galaxies. In fact, all the 10 galaxies in our sample have velocity fields with a measurable ordered component, contrary to the general belief (e.g. Lo et al. 1993) that the velocity fields of faint dwarf galaxies are chaotic. Some of our sample galaxies overlap with those in Lo et al. (1993), and from a comparison of the new and old determinations of the velocity fields, it appears that high sensitivity and high velocity resolution (1.6 km s1 as opposed to the earlier used 6 km s1) are crucial to discern systematic kinematical patterns in such faint galaxies. The origin of these ordered fields is unclear. One would expect that in the absence of external forces, or internal energy input, gas with non-zero angular momentum would settle down into a rotating disc. Tidal forces and/or energy input from star formation could profoundly disrupt the gas velocity fields. Indeed it has often been suggested that a strong starburst could drive out the ISM of such small galaxies (see e.g. Dekel & Silk 1986; Efstathiou 2000; Ferrara & Tolstoy 2000). For two of the galaxies in our sample, viz. GR 8 (Begum & Chengalur 2003) and KDG 52 (Puche & Westpfahl 1994), detailed modelling shows that the velocity field can be fitted by a combination of circular and radial motions of the gas. In general though, there does not seem to be any particular correlation between the current SFR (see Table 5) and the distortion of the velocity fields. For example, the velocity field of CGCG 269 049 (Fig. 5b) shows relatively mild deviations from that expected from rotation, as compared to that of UGC 7298 (Fig. 6b), even though both galaxies have comparable luminosities and the SFR in CGCG 269 049 is an order of magnitude more than that of UGC 7298. Tidal interactions are also not clearly implicated, as several of our galaxies are relatively isolated, and none of them shows morphologies typical of tidal interactions. Interestingly, some compact high-velocity clouds (notably M31 HVC 1, M31 HVC 16; Westmeier et al. 2005) show similar velocity gradients. Westmeier et al. (2005) argue against these velocity gradients being due to tidal forces, and suggest that they may be indicative of dark matter in these objects. Given their peculiar kinematics, it is difficult to accurately determine the total dynamical mass of our sample galaxies. We instead compute an indicative dynamical mass using the virial theorem, assuming that the H I distribution is spherical and has an isotropic velocity dispersion and negligible rotation. We realize that these assumptions are unlikely to be rigorously justifiable in the current situation, and therefore this mass estimate is at best indicative. Under the assumptions mentioned above, the virial mass estimate is (e.g. Hoffman et al. 1996): MVT = 5 RH I t2rue where R H I is the H I radius of the galaxy at a column density of 1019 atoms cm2 (from Table 3). true is the H I velocity dispersion corrected for the instrumental broadening as well as for the broadening due to the velocity gradient over the finite size of the beam. This correction is applied using where true is the true velocity dispersion, v is the channel width, b characterizes the beam width (i.e. the beam is assumed to be of the form ex2/b2 ) and v o is the observed rotation velocity. obs is the observed velocity dispersion in the H I gas given in Table 3. Table 4 lists the estimated dynamical mass (M VT) for our sample galaxies. The columns in the table are: Column (1) the galaxy name, Column (2) the estimated true, and Column (3) the stellar mass-to-light ratio; B is obtained from the observed B V colour for each galaxy, using the low-metallicity Bruzual and Charlot stellar population synthesis (SPS) model for a stellar population with metallicity Z = 0.008, a Salpeter initial mass function (IMF) and an exponentially declining SFR of age 12 Gyr (Bell & de Jong 2001). In the absence of any colour information for CGCG 269 049, we assume B = 1 for it. Column (4) the stellar mass obtained from the assumed mass-to-light ratio and Column (5) the virial mass, as obtained from equation (3). Fig. 8(a) is a plot of the ratio of H I mass to the dynamical mass against absolute blue magnitudes for a sample of dwarf irregular galaxies. The references from which the data are taken are listed in the figure caption. For our current sample, the dynamical masses are taken from Table 4, Begum et al. (2003), Begum & Chengalur (2004) (for KK 44 and DDO 210) and Young & Lo (1997) (for LGS-3 and Sag DIG). The ratio of M H I/M VT for UGC 4459 is found to be 0.34, which is larger than a value typically seen in dwarf galaxies. Such a high value of M H I/M VT is also seen in some blue compact dwarf galaxies (e.g. van Zee, Skillman & Salzer 1998). Fig. 8(b) shows M VT/M lum for the same sample, plotted as a function of M B. The luminous mass, M lum, is the sum of the stellar and gas mass. The stellar mass for all galaxies was computed in exactly the same way as for our sample galaxies; the gas mass is obtained by taking into account the contribution of primordial He, that is, M gas = 1.4 M H I. We note that although M VT/M lum does not show any correlation with M B for the full sample, there is a trend of an increase in the M VT/M lum ratio with the decrease in M B, seen in our sample galaxies. Further, a jump in M VT/M lum ratio is seen at the faintest luminosities (M B > 10.0). While this might be indicative of increased baryon loss from the haloes of these galaxies (e.g. Gnedin & Zhao 2002), we caution that there is considerable uncertainty in the dynamical mass for the galaxies in this magnitude range, and also that the total number of galaxies is too small to substantiate such a claim. 5.2 H I column density and star formation Our sample galaxies have widely different SFRs (see Table 5), and range from having non-detectable ongoing star formation (e.g. KK 230, KDG 52), to having an optical appearance that is dominated by bright star-forming knots (e.g. UGC 4459, CGCG 269 049). As such, this is a well-suited sample for trying to determine the connections (if any) between the H I distribution and kinematics and star formation. For spiral galaxies, the SFR appears to be quite well correlated with the gas column density, though it is unclear if the gas column density is the only relevant parameter, as in the Schmidt law, or whether a combination of gas column density and the dynamical time is important (as in the instability criteria of Toomre, see Kennicutt 1998b). For dwarf irregular galaxies, it has been suggested that star formation occurs only above a critical threshold column density (1021 cm2, when measured at a spatial resolution of 500 kpc) and that this may be because a critical amount of dust shielding is required before star formation can commence (Skillman 1987). Since the observed column density is resolution-dependent and the distance of our sample galaxies varies from 0.70 Mpc (Leo A) to 4.2 Mpc (UGC 7298), one would require maps at angular resolutions varying by a factor of 6. This problem is well suited to a telescope like the GMRT, where because of the hybrid configuration, maps at a range of angular resolutions can be made from a single observing run. For all our galaxies, we produced CLEANed H I maps at an angular resolution corresponding to a linear scale of 300 pc at the distance to the galaxy. We thus have a unique data set which spans a wide range of SFRs, but for which maps are available at a similar linear resolution. The results from a comparison of the H I distribution with the sites of H emission (see Fig. 9) are given in Table 5. Column (1) shows the galaxy name, Column (2) shows the absolute blue magnitude (M B) of the galaxy, Column (3) shows the resolution of the H I column-density map, Column (4) shows the corresponding linear resolution in pc, Column (5) shows the observed H I peak column density, Column (6) shows the peak gas surface density. The gas surface densities are obtained by correcting the observed H I column densities for inclination and for primordial He content (which we take to be 10 per cent of H I by number). Note that the peak gas density need not occur at the centre of the galaxy, Column (7) shows the column density of the H I contour that just encloses all the H II emission in the galaxy, Column (8) shows current SFR (for KDG 52 and KK 230 the limits on the SFR are listed), Column (9) shows the metallicity of the galaxy (for some of our sample galaxies no measurement of the metallicity exists) and Column (10) shows the references for the SFR and the metallicity. DDO 210 and KK 44 are the only galaxies in our sample which show systematic rotation; the rotation curves of these galaxies have been presented in Begum et al. (2003) (KK 44) and Begum & Chengalur (2004) (DDO 210). For these two galaxies we show in Fig. 10 the ratio of the azimuthally averaged gas density to the threshold density predicted from the instability criterion of Toomre. For both of these dwarf galaxies, this ratio is everywhere smaller than the threshold ratio for star formation in spiral galaxies ( g/ crit 0.7; Kennicutt 1989). A similar result was obtained by van Zee et al. (1997), albeit for brighter (and more rotation dominated, V / > 5) dwarfs. While this low ratio of gas density to critical density is interesting, it is unclear whether the instability criteria of Toomre are relevant in a situation where the rotation speed is comparable to the velocity dispersion. From the H overlays, one can see that if there does exist a threshold H I column density for star formation, it is only in the very loose sense that one can find a (relatively) high H I column density contour that just encloses all the star-forming regions. The actual value of the column density delineating the star-forming regions varies by more than a factor of 4 between different galaxies in our sample. Further, the morphologies of the H emission and the high-column density H I are quite dissimilar in several cases (e.g. UGC 7298, GR 8 and KK 44). Thus, while high H I column density may be necessary for star formation, it clearly is not, in this sample at least, a sufficient criterion for star formation. From Table 5 one can also see that there is no particular correlation between the threshold column density and the metallicity. The metallicity of our sample galaxies is somewhat lower than that in the original sample of Skillman (1987). In that sample the galaxies, with one exception [Sextans A, with 12 + log(O/H) = 7.49] (Kunth & Otlin 2000), have 12 + log(O/H) between 8 and 8.34. For the galaxies in our sample for which measurements exist, the metallicity is typically 1 dex lower, while the star formation threshold density is similar to that noted by Skillman (1987). It has also been suggested that the threshold density for star formation is more related to the presence of a cold phase; in this case, the value of this threshold does not change much with metallicity (Schaye 2004). To further explore the connection between the amount of highcolumn density gas in the galaxy and the star formation, we show in Figs 11(a) and (b) the SFR as a function of the total H I mass as well as the SFR as a function of the mass of H I which has a column density greater than the threshold density defined in Table 5. The star formation rate actually correlates slightly better with the total H I mass of the galaxy (correlation coefficient 0.34, excluding those galaxies where no H emission was detected) as compared to the mass of the gas at high H I column density (correlation coefficient 0.25). Clearly, the efficiency with which gas is converted into stars in these dwarf galaxies is not a function of the amount of high column density of the gas alone. The strongest correlation between the gas distribution and indicators of current or past star formation that we find in our sample is that between the peak gas density (recall that this need not occur at the centre of the galaxy) and the absolute magnitude (Fig. 11c). The reason for the existence of such a correlation is unclear. The most straightforward interpretation is that bigger galaxies are more able to support high-column density gas; this in turn made them, on the average, more efficient at converting their gas into stars. On the other hand, as noted above, the current SFR itself does not correlate particularly strongly with the local column density. One way to reconcile this would be if in such small galaxies, feedback processes rapidly destroy the correlation between the local gas column density and the local SFR. If feedback from star formation is important, one might expect to see the strongest evidence for this on small scales. Hence, we compare the highest-resolution H I images of our sample galaxies with the sites of active star formation, that is, H emission. Our highestresolution images have beam sizes 34 arcsec, which corresponds to linear scales between 19 and 100 pc for our sample galaxies. At this high resolution, the emission could not be CLEANed; moment maps were instead made from the dirty cubes. As discussed earlier, this leads to a scaling uncertainty, which means we cannot translate the observed flux distributions into corresponding H I column densities. We can, however, still use our maps to search for correspondences between the morphologies of the H and the highcolumn density H I. The overlays are shown in Fig. 12; for all the galaxies in our sample, the H I emission shows substantial fine scale structure, with shell-like, filamentary as well as discrete clumplike morphologies being visible. The few other dwarf galaxies that have been imaged at similar linear scales, for example, IC10 (20 pc; Wilcots & Miller 1998), the Small Magellanic Cloud (SMC) (28 pc; Staveley-Smith et al. 1997), the Large Magellanic Cloud (LMC) (15 pc; Kim et al. 2003) also show a similar wealth of small-scale structure. At these scales, the H emission is sometimes seen coincident with high H I column densities (e.g. the northern star-forming region in UGC 4459, the north-eastern region in KK 44, the H knot in CGC 269 049), sometimes the high H I column density gas forms a shell around the H emission (e.g. the south-western star-forming region in KK 44, the high-density H I clumps in GR 8) and sometimes there seems to be no connection at all between the high-density H I gas and the H emission (e.g. UGC 7298). In general, while high H I column density gas is present in the vicinity (as measured on linear scales <100 pc) of current star formation, there does not seem to be a simple, universal relationship between the H-emitting gas and the high-column density neutral gas. 5.3 H I line profiles and star formation Young and collaborators have found that in faint dwarfs, H I line profiles in regions of active star formation differ substantially from a simple Gaussian shape (Leo A, Young & Lo 1996; Sag DIG, Young & Lo 1997; UGCA 292, Young et al. 2003). Young et al. (2003) used a two-Gaussian fit as well as a fit using GaussHermite polynomials to parametrize the line profiles. To facilitate easy comparison, we fit the line profiles to our sample galaxies using the same two models. Apart from Leo A and Sag DIG [where we use essentially the same VLA data as used by Young et al. (2003)] our sample has two galaxies, viz. DDO 210 and GR 8, in common with the earlier sample. We include these two galaxies in the analysis that we do in this section, and compare the results obtained from the GMRT data with those obtained earlier with the VLA. Ideally, one would like to fit profiles to data cubes which correspond to the same linear resolution at the distance to the galaxy. However, fitting to the line profile requires a good S/N, and for the fainter galaxies, the S/N is adequate only in the lowest-resolution images. The linear resolution of the data cubes used for profile fitting is given in Table 6. For all the sample galaxies, the line profiles at each location were first fitted with a single Gaussian component and the residuals were inspected. H I profiles in some cases were found to deviate measurably from a simple Gaussian in such cases the profiles were often either asymmetric or symmetric, but with narrower peak and broader wings than a Gaussian. The profiles were then fitted with a double Gaussian and also separately with a GaussHermite polynomial. The GaussHermite polynomial used for the profile fitting is given as: where y = (x b)/c. Parameters a, b and c are equal to the amplitude, mean and dispersion, respectively, for a Gaussian [to which the equation (4) reduces, when parameters h3 and h4 are zero]. Parameter h3 is related to the skewness of the line profile, that is, in the case of an asymmetric line profile h 3 = 0. If h 4 = 0, the line profiles either have a more pointed top with broader wings (h 4 > 0) or have a flatter top (h 4 < 0) than a Gaussian. In the case of GaussHermite fits, the profiles for which both h3 and h4 parameters were less than three times the uncertainty in these parameters were rejected as bad fits. Similarly, in the case of doubleGaussian fits, the profiles for which the width of the fitted narrow component was less than the velocity resolution of 1.65 km s1 (within the error bars) were rejected. Following Young & Lo (1996), we use the F-test to distinguish between profiles that are adequately fitted by a single Gaussian and those that are not. Locations where the null hypothesis (viz. that a single Gaussian provides an equally good description of the line profile as compared to a double Gaussian or GaussHermite polynomial) was rejected at the 90 per cent or higher confidence level were compared with the locations of ongoing star formation. The results of the line-profile fitting are given in Table 6. Column (1) gives the galaxy name, Column (2) shows the resolution of the H I distribution used for the profile fitting, Column (3) shows the linear resolution in pc, Column (4) shows the minimum gas surface density (the observed H I column density corrected for inclination and He content) enclosing the regions with non-Gaussian H I profiles, Column (5) shows h3 and h4 parameters of the best-fitting GaussHermite polynomial, Column (6) shows the range of the velocity dispersion of the narrow component in the double-Gaussian fit and Column (7) shows the range of the velocity dispersion of the broad component in the double-Gaussian fit. In the case of UGC 4459 and GR 8, results of the line-profile fitting from separate regions (as marked in Fig. 13), are described separately in Table 6. For KDG 52, the two separate regions showing deviation from the Gaussian, near the eastern and western clump, gave similar results. The regions in our sample galaxies where the double Gaussian gave a better fit to the line profiles than a single Gaussian are marked as crosses on the H I column density distribution in Fig. 13. The regions where the GaussHermite polynomial gave a better fit than the single Gaussian are almost similar to the regions where the double Gaussian gave a good fit, hence are not shown separately. We note that for most galaxies (with the exception of DDO 210) the extent of these regions is comparable to our spatial resolution. To allow easy cross-comparison with regions having ongoing star formation, the H-emitting regions are represented as grey-scales in Fig. 13. The line profiles for KK 44 and KK 230 throughout the galaxy are found to be well described by a single Gaussian component, hence are not shown. As seen in Fig. 13, no particular correlation is seen between the location of H emission and the deviation of H I line profiles from single Gaussians. Not all star-forming regions in our sample galaxies show deviation of the line profiles, for example, UGC 7298, KK 44 and the eastern clump in GR 8. Conversely, not all regions which show deviations of line profiles are associated with the starforming regions, for example, KDG 52, DDO 210 and UGC 7298. In this sample at least, the correlation found by Young et al. (2003) from their analysis of three dwarf irregular galaxies does not seem to hold. We also do not find any correlation between the presence of asymmetric profiles and the global star formation activity. The dwarf galaxies DDO 210 and GR 8 are common between our sample and that of Young et al. (2003). The results derived by Young et al. (2003) for DDO 210 are similar to our results. On the other hand, for GR 8, Young et al. (2003) found few line profiles with h 3 = 0 in the southern clump, (albeit with a very small magnitude of h3), and almost none in the eastern clump, while we found all the profiles associated with the southern and eastern clump have h 3 = 0 (within the 3 uncertainty of the parameter). However, this difference is not pronounced, and one should also note that Young et al. (2003) used 14 14 and 18 18-arcsec2 resolution data cubes, whereas we have used a 30 30-arcsec2 resolution data cube. In general, therefore, there is relatively good agreement between the fits obtained with the GMRT and VLA data. 6 S U M M A R Y We compare the H I distribution, kinematics and current star formation in a sample of 10 extremely faint nearby dwarf galaxies. For five of these galaxies, fresh GMRT H I data are presented in this paper. The large-scale gas distribution in the galaxies is generally clumpy, and the peak H I column density is often well removed from the geometric centre. For all galaxies we find a large-scale ordered velocity field, although the patterns are mostly not reconcilable with that expected from a rotating disc. From a simplistic virial theorem-based estimate of the dynamical mass, we find very tentative evidence that the faintest dwarf irregulars have a somewhat smaller baryon fraction than brighter galaxies. We compare the regions of ongoing star formation with regions of high H I column density, with the column density being measured at a uniform linear scale (300 pc) for all galaxies in our sample. We find that while the H emission is confined to regions with relatively high column density, in general the morphology of the H emission is not correlated with that of the high-column density H I gas. Thus, while high gas column density may be a necessary condition for star formation, it is not, in this sample at least, a sufficient condition. We also examine the line profiles of the H I emission, and check if deviations from a simple Gaussian profile are correlated with star formation activity. We do not find any such correlation in our sample there are regions with ongoing star formation but with simple Gaussian line profiles, as well as regions with complex line profiles but no ongoing star formation. Finally, we look at the distribution of H I gas at linear scales 20100 pc. All our sample galaxies show substantial small-scale structures with shell-like, filamentary as well as clumpy features being identifiable in the images. H-emitting regions are sometimes associated with clumpy features; sometimes the H emission lies inside a shell-like feature in the H I, and sometimes there is no particular H I column density enhancement seen near the H emission. The interplay between star formation and gas density and kinematics in these galaxies hence appears to be very varied, and the general unifying patterns seen in larger irregulars and spiral galaxies seem to be absent. Star formation and feedback are complex processes, and perhaps it is the presence of simple large-scale correlations in big galaxies that should surprise us more than the absence of such correlations in small galaxies. AC K N OW L E D G M E N T S We would like to thank Dr U. Hopp for providing the optical images of UGC 4459 and Dr S. Pustilnik for useful discussion on UGC 4459. AB thanks the Kanwal Reiki Scholarship of TIFR for partial financial support. The observations presented in this paper would not have been possible without the many years of dedicated effort put in by the GMRT staff in order to build the telescope. The GMRT is operated by the National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research.


This is a preview of a remote PDF: http://mnras.oxfordjournals.org/content/365/4/1220.full.pdf

Ayesha Begum, Jayaram N. Chengalur, I. D. Karachentsev, S. S. Kaisin, M. E. Sharina. Gas distribution, kinematics and star formation in faint dwarf galaxies, Monthly Notices of the Royal Astronomical Society, 2006, 1220-1234, DOI: 10.1111/j.1365-2966.2005.09817.x