Local tadpole galaxies: dynamics and metallicity
Tadpole galaxies, with a bright peripheral clump on a faint tail, are morphological types unusual in the nearby universe but very common early on. Low mass local tadpoles were identified and studied photometrically in a previous work, which we complete here analyzing their chemical and dynamical properties. We measure H velocity curves of seven local tadpoles, representing 50 % of the initial sample. Five of them show evidence for rotation ( %), and a sixth target hints at it. Often the center of rotation is spatially offset with respect to the tadpole head (3 out of 5 cases). The size and velocity dispersion of the heads are typical of giant HII regions, and three of them yield dynamical masses in fair agreement with their stellar masses as inferred from photometry. In four cases the velocity dispersion at the head is reduced with respect to its immediate surroundings. The oxygen metallicity estimated from [NII]6583/H often shows significant spatial variations across the galaxies (0.5 dex), being smallest at the head and larger elsewhere. The resulting chemical abundance gradients are opposite to the ones observed in local spirals, but agrees with disk galaxies at high redshift. We interpret the metallicity variation as a sign of external gas accretion (cold-flows) onto the head of the tadpole. The galaxies are low metallicity outliers of the mass-metallicity relationship. In particular, two of the tadpole heads are extremely metal poor, with a metallicity smaller than a tenth of the solar value. These two targets are also very young (ages smaller than 5 Myr). All these results combined are consistent with the local tadpole galaxies being disks in early stages of assembling, with their star formation sustained by accretion of external metal poor gas.
Subject headings:galaxies: abundances – galaxies: dwarf – galaxies: evolution – galaxies: formation – galaxies: kinematics and dynamics – galaxies: structure
Tadpole galaxies consist of a large star-forming clump at one end (the head) and a long diffuse region to one side (the tail). This asymmetric morphology is rather common at high redshift but rare in the local universe. For example, tadpoles constitute 10% of all galaxies larger than 10 pixels in the Hubble Ultra Deep Field (UDF) (Elmegreen et al. 2007; Elmegreen & Elmegreen 2010), and they represent 6% of the UDF galaxies identified by Straughn et al. (2006) and Windhorst et al. (2006) using automated search algorithms. In contrast, Elmegreen et al. (2012b, hereafter Paper I) find only 0.2% tadpoles among the star-forming local galaxies of the Kiso survey by Miyauchi-Isobe et al. (2010). This decrease suggests the tadpole morphology to represent a common but transit phase during the assembly of some galaxies. Since local tadpole galaxies are very low mass objects compared to their high redshift analogues (Elmegreen et al. 2012b), the phase must be already over for the local descendants of high redshift tadpoles.
Another independent observation also suggests that the tadpole morphology characterizes a very early phase of evolution. From the point of view of their chemical content, Extremely Metal Poor (XMP) galaxies are the least evolved objects in the local universe (e.g., Pagel et al. 1992; Kunth & Östlin 2000; Izotov & Thuan 2004). They represent only 0.1 % of the galaxies in an arbitrary nearby volume (e.g., Morales-Luis et al. 2011), but a significant fraction of these chemically primitive objects turn out to have tadpole or cometary shape (75 %; Papaderos et al. 2008; Morales-Luis et al. 2011). This association between low metallicity and tadpole shape suggests that they are attributes characteristic of very young systems.
The tadpole structure could have a variety of origins. Elmegreen & Elmegreen (2010) showed lopsided ring-like galaxies that would look like tadpoles if viewed edge-on. Tadpoles could also result from mergers or close galaxy-galaxy interactions (e.g., Baldwin et al. 1982; Corbin et al. 2005; Straughn et al. 2006; Windhorst et al. 2006; D’Onghia et al. 2010). However, this merger scenario cannot be the universal pathway to tadpole formation, at least not locally, since the class of dwarf local tadpoles tends to be relatively isolated (Campos-Aguilar et al. 1993; Kunth & Östlin 2000) and lacks obvious tidal features (Papaderos et al. 2008). Another possibility is that the lopsided starburst results from ram compression by motion through the intergalactic medium (Elmegreen & Elmegreen 2010). Tadpoles could be disks with star formation triggered on the leading side and the rest visible as a red tail of older stars. Alternatively, they could be heavily stripped galaxies with star formation and old stars at the leading edge, and a tail made from star formation in the stripped gas (e.g., Chung et al. 2009). The scenario in which dwarf galaxies are converted into dwarf spheroidal galaxies in the halos of larger galaxies (Lin & Faber 1983; van den Bergh 1994; Mayer et al. 2006) or in galaxy clusters (Boselli et al. 2008; van Zee et al. 2004) could involve a tadpole phase as the gas and young stars are pulled behind. A third possibility is that some tadpoles are normal disk galaxies with a large turbulent Jeans length for gravitational collapse of their interstellar medium. This happens in galaxies that have either small rotational velocities or large turbulent motions, and in this case only large star forming clumps can be produced (Förster Schreiber et al. 2006; Elmegreen et al. 2009). If there is only one clump at any particular time, then it will appear as a tadpole viewed from the right perspective. A fourth possibility is that the head-tail structure results from propagating star formation across a disk (e.g., Franx et al. 1997; Papaderos et al. 2008). Finally, as we propose in this paper, the head of the tadpole may result from accretion of external flows of pristine gas, that penetrate the dark matter halo and hit and heat a pre-existing disk. Cosmological simulations predict cold-flow buildup to be the main mode of galaxy formation (Dekel et al. 2009; Genel et al. 2012), and the incoming gas is expected to form giant clumps that spiral in and merge into a central spheroid (Noguchi 1999; Genzel et al. 2008; Elmegreen et al. 2008). In addition, inflow of low-metallicity gas seems to be suggested by distorted HI velocities and morphologies of some local star-forming galaxies (see López-Sánchez et al. 2012, and references therein).
It is important to realize that the different formation mechanisms mentioned above are not mutually exclusive, but exhibit a large degree of overlap between them. For example, the cold-flow accretion may be considered as a (minor-)merger accretion, and it may also be regarded as the interaction of the galaxy with the (filamentary) intergalactic medium.
In this context of galaxy formation, we examined a sample of fourteen tadpole galaxies in the local universe for comparison with high-redshift tadpoles (Paper I). These local tadpoles seem to form a continuous sequence with the UDF tadpoles studied by Elmegreen & Elmegreen (2010). With regards to their photometric properties, local tadpoles occupy the low mass end in sequences such as star formation, surface density and mass-to-light ratio. In addition, the radial intensity profiles of the local tadpoles show an exponential decrease at large galactocentric distances, which was interpreted as evidence for an underlying disk. The work in Paper I is followed up in the present paper. In order to determine the dynamical properties and metallicities of local tadpoles, we measure H spectra along the head-tail direction in a representative fraction of the original sample. The results reported here show the galaxies to be rotating structures with an unexpected metallicity pattern, which may shed light into the nature of the tadpole morphology.
The paper is organized as follows: Sect. 2 describes the observations, the main steps of the reduction, and the main properties of the resulting spectra. Section 3 details the approximations used along the work to derive physical parameters from spectra. Section 4 analyzes the dynamical properties of the galaxies, their rotation curves (RCs) and linewidths. These are used to infer dynamical masses, which are then compared with the stellar masses derived in Paper I (Sect. 6). Section 5 studies the light profiles both in continuum and H emission. The variation of the oxygen abundance along the galaxy is measured in Sect. 7. Notes on individual galaxies are given in Sect. 8. The results of our observation in the context of tadpoles as dynamically young disks are discussed in Sect. 9.
2. Observations and data reduction
The sample of local tadpole galaxies selected in Paper I comes from the Kiso survey of UV intense galaxies (Miyauchi-Isobe et al. 2010). Among those galaxies labelled in the Kiso catalog as clumpy or having conspicuous HII regions, we visually inspected 158 with images in SDSS-DR7 (Abazajian et al. 2009). Then tadpoles were subjectively selected as those objects with lopsided light distribution, so that the brightest clump is far to one end and the rest of the galaxy is mostly featureless. This inspection rendered only 13 targets, which we completed with one additional galaxy with similar features from the University of Michigan survey of emission-line objects (MacAlpine & Lewis 1978). Half of the sample in Paper I was randomly selected for the follow-up work presented in this paper. The observing program was aimed at determining the dynamical properties of the galaxies based on long slit spectroscopic observations along the well-defined head-tail direction (see Fig. 1). The project was planned for the spectrograph IDS of the 2.5 m Isaac Newton Telescope (INT) at the Observatorio del Roque de Los Muchachos (Laing & Jones 1985). It provides adequate spectral resolution to determine centroids of emission lines with an accuracy better than 10 km s, which suffices to characterize RCs even for dwarf galaxies. The original project could not be completed due to poor weather conditions, so it was finished using service time of the 2.5 m Nordic Optical Telescope (NOT), operated in the same observatory (Andersen 1985).
The observing logbook is summarized in Table 1. The used IDS@INT setup includes a grating with 1800 lines mm, which provides a scale in the spectrograph focal plane of 0.32 Å pix equivalent to 044 pix. The spectral resolution is set by the width of the 1″ slit, which corresponds to 2.2 pix, or 0.73 Å, or 33 km s in H. The camera covers some 700 Å around the central wavelength, which was tuned to H, and which automatically included the line [NII]6583 used in our metallicity estimates (Sect. 3). The tadpoles kiso3193, kiso3867, kiso5149 and kiso8466 were observed with this setup (Table 1). Figure 2 shows one of them as an example. The image contains the spectral region around H. The reduction procedure included the standard bias and flatfield corrections, cosmic ray elimination, as well as removal of sky emission lines. The typical signal-to-noise (S/N) ratio in the peak H emission exceeds 500, and it decreases down to one in the outskirts of the galaxy (Fig. 2, bottom panel). This S/N is achieved after integrating for some one and a half hours (Table 1).
Seeing was of the order of 1″ (Table 1).
The other three galaxies, kiso5639, kiso6669, and kiso6877 were observed with the NOT telescope with an observational setup equivalent to the previous one. We used the ALFOSC spectrograph which, together with a high resolution grism, provided a scale of 0.26 Å pix 019 pix at H. This pixel size over-samples the spectrum, whose resolution was set by the 09 wide slit. In order to match the pixel size to the actual resolution, the original data were re-binned , yielding the final resolution indicated in Table 1. The data reduction followed the standard process mentioned above, with a small difference worthwhile mentioning. The standard procedure includes using images of the sky taken during twilight to characterize (and then to correct for) large-scale illumination gradients in the images. Such sky-flat images corresponding to kiso6877 were not available. For reduction we used those from the previous night, but this rendered spectra with small residual sky lines. To estimate the uncertainties introduced by this problem, the reduction was independently repeated by two of us (JSA and JMA). Thus different spectral regions and interpolations were used to construct the flatfield and to subtract the sky background. The two reductions are in good agreement and provide consistent results. Table 2 contains the final physical parameters inferred from the two parallel reductions, and they agree to the point that their small differences do not affect the results in the work. Seeing during observations was sub-arcsec. The NOT spectra have S/N similar to the INT spectra; the spectral resolution is slightly worse, and the angular resolution slightly better.
Determining the actual spectral resolution is central for the proper interpretation of the observed linewidths. As usual, it was determined from the width of the telluric lines in the spectra since their intrinsic widths are just a few km s and thus negligible (e.g., Balthasar et al. 1982). Gaussian functions were fitted to the lines, and their mean Full Width Half Maximum (FWHM) are quoted in Table 1 as the spectral resolution, with the error bars representing the rms fluctuations along the field of view. We average over all the targets observed with the same instrumental configuration to estimate the resolution corresponding to that particular configuration. The weighted average for the IDS@INT spectra turns out to be km s, whereas in the case of ALFOSC@NOT the measured resolution is km s. The error bars represent the weighted rms fluctuations.
|NameaaNamed as in Elmegreen et al. (2012b), Paper I.||RA DEC||bbTotal exposure time.||Date||Instrument||Slit||SeeingccMean value for the night from RoboDIMM@WHT; see http://catserver.ing.iac.es/robodimm/.||Pixel||Scale||TelluricddFWHM of the observed telluric lines, used as proxy for spectral resolution.|
|kiso3193||08:56:08 +39:52:09||6000||Feb 10||IDS@INT||10||10||0.32Å044||124||26.01.9|
|kiso3867||09:40:13 +29:35:30||6000||Feb 10||IDS@INT||10||10||0.32Å044||36||25.01.0|
|kiso5149||11:16:08 +23:29:16||6000||Feb 10||IDS@INT||10||10||0.32Å044||834||28.21.5|
|kiso5639||11:41:07 +32:25:37||5000||May 29||ALFOSC@NOT||09||08||0.51Å038||119||53.95.7|
|kiso6669||12:31:50 +27:23:13||5000||May 29||ALFOSC@NOT||09||08||0.51Å038||299||55.68.7|
|kiso6877||12:46:11 +26:15:01||5000||May 31||ALFOSC@NOT||09||07||0.51Å038||128||54.36.6|
|kiso8466||16:03:27 +19:09:46||4800||Feb 10||IDS@INT||10||10||0.32Å044||313||24.90.7|
3. Determination of physical parameters
Velocities, masses, abundances and other physical parameters are determined from the spectra. This section explains how these parameters are computed, including the underlying hypotheses and their uncertainties.
Bulk velocities are measured from the displacement of H. We compute the displacement both as the barycenter of the emission line, and as the center of a Gaussian function fitted to the profile. Errors are estimated from the S/N measured in the continuum and then propagated to the centroids (e.g., Martin 1971, Sect. 5.3). The FWHM of the profiles are also measured directly from the profile and from the Gaussian fit. Their errors are also inferred from the noise measured in the continuum by error propagation.
The variation of the velocity with position along the slit is usually referred to as the velocity curve. The velocity curves of our targets can be ascribed to rotation. In order to characterize such rotation, we fit an analytic RC to the velocity curve. We wanted a function that is simple but produces a good match to the observations. The universal RC advocated by Salucci et al. (2007) does a good job representing the observed variation. Specifically, we adopted the dark matter component of the universal RC given by,
where is the velocity observed at a distance , and , , and are free parameters to be determined from a non-linear fit. The curve is simple – yields a center for the RC, gives its amplitude ( when ), and provides the spatial scale for the central gradient. We stress that the hypotheses behind the analytic expression (1) are of little importance in our context. The formula is used here because it provides a good representation of the observations, and so provides a smoothed version of the velocity curve. Moreover, it allows us to determine the center of rotation . The center thus determined does not depend so much on the actual expression used to parameterize the RC, but on the fact that the curvature changes sign at the center of rotation, i.e.,
with the second derivative of . This is a general property that any anti-symmetric RC satisfies.
Dynamical masses are estimated from RCs and linewidths. Assuming the mass distribution to be spherically symmetric (e.g., a bulge or a dark matter halo), the circular velocity that balances the gravitational pull depends only on the mass enclosed within the radius , ,
with the gravitational constant. We assume that the measured macroscopic velocities are the circular velocities affected by the inclination of the galaxy plane, i.e., . We also assume the spectrograph slit to be oriented along the galaxy major axis, implying . Then the inner mass of the galaxy up to distance has the usual expression,
which using astrophysical units turns out to be
with measured in kpc and U in km s. Equation (5) is an approximation; however, including more realism in the mass distribution (e.g., pressure support, Dalcanton & Stilp 2010; or non-spherical components, Salucci & Persic 1997) would only modify the scaling factor as a correction of order one. Masses of individual clumps are inferred from linewidths assuming virial equilibrium. If the isotropic velocity distribution that balances the clump gravity has a dispersion , then (Bosch et al. 2009, Sect. 4.2)
The symbol stands for the half-light radius of the clump, so that Eq. (6) is equivalent to Eq. (5) after including the appropriate scaling factors. As in the case of Eq. (5), distances are measured in kpc and linewidths in km s. The linewidth in Eq. (6) is not the observed width , but the width corrected for instrumental spectral resolution , thermal motions in the nebula , and natural width of H ,
see, e.g., Terlevich & Melnick (1981); Melnick et al. (1999). We use the measured widths of the telluric lines as proxy for instrumental broadening (Table 1). The thermal broadening is assumed be the same for all galaxies at all positions. Its value has been set to a representative round number FWHM km s, which approximately corresponds to H atoms at 14000 K, a temperature typical of HII regions. The actual FWHM is of secondary importance since the range of possible values is significantly smaller than the FWHM resulting from Eq. (8)111 FWHM varies only from 19 to and 26 km s for temperatures between 8000 and 15000 K. Moreover, we did the exercise of calculating masses also with the extreme value of FWHM=0, to check that this assumption does not modify the conclusions in the work. Setting FWHM=0 is equivalent to including thermal motions as part of the virial equilibrium represented by Eq. (6). The exercise implies that we do not have to worry about whether the thermal motions contribute or not to the virial equilibrium.Including or not thermal motions as part of the kinetic energy does not significantly modify the masses estimated in the work. . Finally, the natural width of H is of the order of 7 km s (e.g., Rozas et al. 2006). The radius used to estimate dynamical masses from line widths using Eq. (6) is computed fitting a 1-D Gaussian to the light distribution across the galaxy head. The observed light profile is assumed to represent a 1-D cut across the center of a 2D-Gaussian, which readily provides the half-light radius from the width of the fitted Gaussian – in a 2D Gaussian, the FWHM is twice the half-light radius. The observed radii are corrected for seeing using a formula similar to Eq. (8), namely,
where stands the measured effective radius and represents the seeing, i.e., the FWHM of the seeing disk as given in Table 1.
The metallicity of the gas is commonly inferred by combining emission-line fluxes of several atomic species to derive their relative abundances (e.g., Pagel & Edmunds 1981; Osterbrock 1989). This approach is the so-called direct method or temperature-based method, and it is to be preferred whenever possible. However, it involves measuring fluxes of lines spread throughout the visible-IR spectrum, so is expensive observationally. Fortunately, we have alternatives called strong-line methods (e.g., Shi et al. 2005; Kewley & Ellison 2008), where the metallicity is estimated empirically by relating the ratio of a few selected line fluxes with the abundance of a particularly relevant metal (typically oxygen). The one proposed by Denicoló et al. (2002) turns out to be ideal in our case, when only the spectral region around H is available. It yields the oxygen abundance from the ratio of [NII]6583 to H, and [NII]6583 automatically appears in the spectra next to H (see Fig. 2). We use the calibration by Pérez-Montero & Contini (2009),
particularly suited for low metallicity targets (c.f., Pettini & Pagel 2004). Equation (10) provides the O metallicity from the flux in a N line, therefore, it may be biased in objects having unusual N/O. In order to discard this potential bias in our O metallicities, we also estimate the ratio of N to O using the sulfur lines [SII]6717 and [SII]6731 present in most of our spectra. We use the calibration
also by Pérez-Montero & Contini (2009).
The spectral flux of the galaxy is computed by integrating the observed spectra between their two extreme wavelengths (from 6242 Å to 6935 Å for INT spectra, and from 6368 Å to 6840 Å for NOT spectra). The flux in H results from integration of the emission line profile around its maximum, once the underlying continuum was removed. The integration includes a 10 Å wide region around the peak emission, whereas the continuum was obtained by fitting a linear function to two continuum windows, 10 Å wide, outside the line. The H equivalent width (EW) is inferred from the H flux dividing by this continuum. Continuum and emission lines are combined in the spectral flux for simplicity, and it suffices to compare the limited extension of the emission line region inferred from H with the rest of the galaxy.
4. Velocity curves and linewidths
Figure 3 shows the velocity curves of the tadpole galaxies included in our study. Five out of the seven targets show velocity gradients interpreted as rotation. The figure includes the best fit to the analytic RC in Eq. (1), which does a good job reproducing the observations. We use for fitting the portion of the velocity curve indicated in red in Fig. 3, which includes all positions but the extremes with large error bars or obvious distortions. Sometimes the interpretation of the velocity curve as a RC is obvious (e.g., kiso8466), but other times the curve looks more like a perturbed RC (e.g., kiso5639). kiso3193 and kiso3867 have a rather flat velocity curve, and therefore no obvious rotation. However, one of them, kiso3867, shows a systematic line shift of the order 10–20 km s between the two extremes of the galaxy (Fig. 3). The amplitude is of the order of the error bars, but the displacement is in the raw data as judged by inspection of the individual H profiles. Table 2 contains the amplitudes of the RCs as assigned by the fit (i.e., in Eq. ). Figure 3 represents velocities obtained from the barycenter of H (Sect. 3). The velocities from the Gaussian fit are not shown because the two estimates differ only by a few km s, a difference always smaller than the error bar assigned to each velocity measurement.
|NameaaFrom Elmegreen et al. (2012b), Paper I. DistanceaaFrom Elmegreen et al. (2012b), Paper I.||FWHMbbH width at the tadpole head.||DccDynamical mass of the head – from Eq. (6).||ddHalf-light radius of the head.||PheePhotometric mass of the head, from Paper I, Table 3.||ffAmplitude of the rotation curve – see Eq. (1).||ggCenter of rotation relative to the tadpole head – see Eq. (1).||DhhDynamical mass of the galaxy – see text for its computation.||PhiiPhotometric mass of the galaxy, from Paper I, Table 3.||12+log(O/H)jjMetallicity at the head – from Eq. (10).|
|[Mpc]||[km s]||||[kpc]||||[km s]||[kpc]|||||
|kiso6877kkSecond independent reduction. 26.3|
Table 2 also contains the center of rotation , which often differs from zero, i.e., from the position of the tadpole head. As argued in Sect. 3, its estimate is fairly robust since it comes from the point where the curvature of the RC changes sign – accordingly, it has the small formal error bars provided in Table 2. Figure 3 shows that three out of the five rotating galaxies, explicitly kiso6669, kiso6877 and kiso8466, have their center of rotation displaced with respect to the head by more than the 1″ uncertainty introduced by seeing (see Table 1). Considering that the rotation center points out the center of the galaxy, our results indicate that often the star-forming region at the head is displaced with respect to the center of the galaxy. This suggests that the head is not a bulge-like central spheroid. The fact that the center of rotation is sometimes displaced from the heads may be easier to appreciate in Fig. 4.
Figure 5 shows the FWHM of H as inferred directly from the emission line profile (the solid lines) and from the Gaussian fit (the dashed lines). A significant part of the observed linewidths is due to instrumental broadening. The widths corrected for instrumental effects and thermal motions are given as dotted lines in the figure. We use the directly inferred widths to estimate the intrinsic widths, but using them or the widths from the Gaussian fit render similar results since the two measurements are in very good agreement (Fig. 5). First note that the widths of all galaxies, large and small, are of the order of 20–70 km s. These widths are typical of giant HII regions such as 30 Doradus rather than the widths of the HII regions observed in large spirals (e.g., O’dell & Townsley 1988; Muñoz-Tuñon 1994). Figure 6 shows the variation of the tadpole head size as a function of its velocity dispersion. The two quantities are known to be correlated in giant HII regions, so that the larger the dispersion the larger the size (e.g., Muñoz-Tuñon 1994; Fuentes-Masip et al. 2000). We find the tadpole heads to follow such trend as characterized by Terlevich & Melnick (1981), Roy et al. (1986), or more recently by Wisnioski et al. (2012) (see Fig. 6). In the case of kiso6877, the error bars of the line width at the head are so large that the width is actually an upper limit, but even with this caveat in mind, the measurement is consistent with its head being a giant HII region (Fig. 6).
The FWHM also varies along the galaxies, and the fluctuations are correlated neither with the spectral flux nor with the H flux (c.f. Fig. 4 and 5). If anything, there is a tendency for the tadpole heads to coincide with local minima of linewidth (see kiso5149, kiso5639, and kiso6877 in Fig. 5). Similarly, the center of rotation of the rotating galaxies does not seem to be associated with extremes of the FWHM curve.
Interpreting the decrease of linewidth associated with the head is not straightforward. One may try to explain its origin in the context of forming stars in a highly turbulent galaxy. Only where the turbulence is low enough the conditions that trigger star formation are met, and this preference for low turbulence regions is what we detect.
The square of the ratio between velocity dispersion and rotation gives the Jeans length relative to the galaxy size (e.g., Bournaud & Elmegreen 2009). In other words, it provides the relative size of the clumps to be produced by gravitational instability. Using Eq. (7) and the data in Table 2, the ratio dispersion to rotation turns out to be between 0.2 and 0.6 at the tadpole heads, which yields expected clump sizes between 0.03 % and 30 % galaxy radii. The relative size increases with decreasing galaxy mass, and it may be a coincidence, but the largest ratios correspond to the two XMP galaxies in the sample (see Sect. 7). In some cases the predicted clumps have sizes comparable to those of the observed heads. The ratio also measures whether the object is supported by random motions or rotation. Our disks have ratios similar to the turbulence-supported clumpy galaxies observed at high redshift (e.g., Ceverino et al. 2012).
5. Fluxes and equivalent widths
Figure 4 shows the variation across the galaxies of the spectral flux and the H flux (the solid lines and the dashed lines, respectively). The position where the two fluxes are largest coincides. (As we already pointed out, the origin of distances used along the paper has been set at the maximum of the spectral flux distribution.) Note also the obvious lopsidedness of all light curves, as expected in tadpole galaxies. Another important feature is the extension of the H emission as compared to the spectral flux emission which includes H plus continuum. H is more concentrated, implying that the region producing the total emission extends further away from the star-forming regions. In other words, it shows the existence of an underlying galaxy, with old stellar populations that contribute to the spectral flux but not to H.
The dotted line in Fig. 4 represents a Gaussian plus a constant term fitted to the H flux around the head. The fits are good, and they allow us to assign a radius to the head. Table 2 lists the effective radii of the heads , defined as the radius enclosing half of their light (see Sect. 3). They span a wide range of values from 1.8 kpc to 50 pc, reflecting the wide range of intrinsic galaxy sizes.
Figure 7 shows the variation of H EWs across the galaxies. The largest EWs tend to coincide with the maximum flux (Fig. 4), although not always (kiso3193 in Fig. 7). In general, EWs are fairly moderate, with maxima 300 Å. The two exceptions correspond to the two extremely metal poor targets to be described in Sect. 7 – kiso5639 and kiso6877. Their large H EW implies the extreme youth of the star-forming regions at the galaxy head. We know from modeling that the H EW of an HII region must be smaller than some 3000 Å, and it drops down very quickly so that a coeval starburst reaches EW200 Å in just 10 Myr (e.g., Leitherer et al. 1999, Fig. 83). There is also a dependence of the EW on metallicity, but age is by far the dominant factor. The large observed H EWs correspond to ages of only a few Myr, and these ages are upper limits to the stars responsible for the ionization since the (old) underlying galaxy produces continuum emission that reduces the observed EW. If kiso5639 and kiso6877 are as young as we infer from their H EWs, one expects to find Wolf-Rayet (WR) star features in the spectra, which are characteristics of extremely young starbursts ( Myr; e.g., Crowther 2007). These features are distinctive broad bumps at 4600–4680 Å and 5650–5800 Å (e.g., Schaerer et al. 1999; Brinchmann et al. 2008). Unfortunately, the WR features lie outside our observed spectral range. However, our galaxies also have SDSS spectra, which cover a wider range (Stoughton et al. 2002; Abazajian et al. 2009). We inspected them for WR features but we did not see any (see Fig. 8). Moreover, our targets were not found in the systematic search for WR galaxies in SDSS carried out by Brinchmann et al. (2008). kiso6877 does not show WR features, most probably because the SDSS spectrum was taken away from the tadpole head, in a region which is not particularly young. kiso5639 does not show the broad WR features either. Instead, its spectrum contains narrow high excitation emission lines including HeII4686 (Fig. 8). These lines are supposed to be excited only by the hard UV-radiation of WR stars, which makes interpreting spectra with HeII4686 but without WR bumps puzzling (see, Shirazi & Brinchmann 2012). However, there is a significant number of star-forming galaxies without WR features showing HeII4686 emission. The reason is unknown, but these galaxies are usually metal-poor (Shirazi & Brinchmann 2012), so that the presence of high excitation narrow lines seems to reflect the extreme youth of a metal-poor starburst. According to Shirazi & Brinchmann (2012), the stellar populations at very low metallicities can have much higher temperatures than is currently expected in models. Then even main sequence O stars may excite HeII4686. Alternatively, in low metallicity environments the winds of WR stars are weak, and so, optically thin in the He continuum, allowing the ionizing radiation to escape creating an HeIII region responsible for the observed emission (see Kehrig et al. 2011, and references therein).
In short, even though kiso5639 lacks WR bumps, the presence of high excitation lines such as HeII4686 reinforces our conclusion that its head contains an extremely young metal-poor starburst.
6. Dynamical masses
Figure 9 shows the relationship between the dynamical mass and the photometric mass for the tadpole heads and the full galaxies. The photometric masses are from Paper I, whereas the dynamical masses come from applying Eqs. (5) and (6). In particular, Eq. (5) has been integrated until the last point used to compute the RC. The actual values used for plotting are listed in Table 2.
Note that all dynamical masses are larger than the photometric masses. The difference is less important in the heads (except for the case of kiso3867, discussed in Sect. 8). Three of the heads have almost identical dynamical and photometric masses, so in these cases there appears to be no dark matter in the heads (kiso5149, kiso6877, and kiso8466). Keep in mind that the dynamical masses of the galaxies derived from RCs are actually lower limits. First, there is a factor in Eq. (5). It is probably unimportant since our galaxies are elongated suggesting large inclinations and so Second, and more critical, is the fact that the RC has been integrated only to the largest radii having velocities. Since the dynamical mass of the galaxies thus derived exceeds the photometric mass, we can conclude that galaxies are objects with significant amounts of non-stellar matter.
If the star-forming regions at the head of the tadpole were self-gravitating, one would expect them to hold some degree of internal rotation. If this rotation significantly differs from the galaxy rotation pattern, and if the head is massive enough, then the rotation of the head could perturb the rotation curve of the galaxy producing a noticeable distortion (e.g., Immeli et al. 2004). We examined the observed RCs for such signals and did not find them, except perhaps in the case of kiso6877, discussed in Sect. 8.
The metallicity was estimated as explained in Sect. 3, using the ratio [NII]6583 to H. Figure 10 shows the variation across the galaxies of the oxygen abundance, including their error bars. (Points with errors larger than 1 dex have being excluded.) The galaxies tend to have sub-solar metallicity (the thick horizontal line marks the solar oxygen abundance given by Asplund et al. 2009, ). The galaxies also present significant abundance gradients, with the lowest abundances tending to coincide with the largest H emissions (e.g., kiso6669 and kiso6877 in Fig. 10, keeping in mind that the vertical dotted lines mark the position of the peak H fluxes). We also note that two targets, kiso5639 and kiso6877, have metallicities well below one-tenth the solar value, therefore, they belong to the selected club of XMP galaxies (e.g., Kunth & Östlin 2000; Guseva et al. 2003). They are really rare objects: one out of a thousand galaxies in the local universe according to Morales-Luis et al. (2011). Therefore the fact that we observe two in a sample of seven cannot be a coincidence. It is known that a significant fraction of XMP galaxies turn out to be cometary or tadpole (Papaderos et al. 2008; Morales-Luis et al. 2011). Here we find that the reverse holds too, i.e., that tadpole galaxies have a significant probability of being XMP. As we discuss in Sect. 9, this fact supports the idea that the tadpole morphology is a sign of dynamical youth, as the low metallicity is a sign of being chemically young.
The observed gradients in metallicity are one of the central results of this work and, therefore, deserve a separate discussion. Our abundance determinations are based on N2=[NII]6583/H rather than on the direct method, and this may be a source of systematic error (e.g., Shi et al. 2005). The spatial gradients in metallicity may be artificially due to gradients in excitation. As Morales-Luis et al. (2013, in preparation) discuss, the excitation and (to a lesser extent) N/O change N2 at . The higher the excitation the smaller N2, and the excitation is expected to change with time since the number of ionizing photons drops down quickly in young starbursts (e.g., Leitherer et al. 1999, Fig. 77). However, the bias the excitation produces is much too small to account for the 0.5 dex gradients we detect (Fig. 10). Sánchez Almeida et al. (2009, Appendix A) studied the difference between the oxygen abundance derived from the direct method and from N2 in a large set starburst galaxies with spectra similar to our tadpoles. There were no systematic differences within 0.2 dex for , which secures the reliability of the abundances found in most locations. As for the points with , N2 overestimates the oxygen abundance, which again secures the low metallicity values we find. Since N2 provides the O metallicity based on the flux of a N line, the metallicity may be biased in objects with unusual N/O. The effect of varying N/O seems to be unimportant too. Figure 11, shows the ratio as derived from [SII]6717,6731/[NII]6583 – Eq. (11). (The [SII] lines are not available in two targets and therefore we cannot estimate their N content.) Even though the error bars are large, the observed N/O looks rather constant along the galaxies, at approximately the plateau typical of low metallicity galaxies (; e.g., Pérez-Montero & Contini 2009). Obviously, if N/O is constant then it cannot fake the observed metallicity drop. However, since the error bars of N/O are large, we decided to run a chi-squared test (e.g., Press et al. 1986) to further discard N/O as the source of the measured O variations in kiso5639 and kiso6877. We compared the observed O with the fake variations to be expected if O is constant but N2 varies as N/O in Fig. 11. The test shows how the two variations are inconsistent with 90% confidence.
Figure 12a shows the mass-metallicity relationship for our galaxies, plotting the oxygen abundance of the head versus the photometric mass of the galaxy. The most massive tadpoles tend to have the largest metallicities, although the metallicity values are displaced downward with respect to the mass-metallicity relationship found in the local universe. The dashed line shows the divide above which 97.5 % of the local galaxies are found according to Tremonti et al. (2004). The tadpoles appear just below this line, so they are metal poor for their masses222As we already pointed out, the metallicity scale is not free from uncertainties. However, even taking them into account, the tadpoles are metal poor. The triple-dotted dashed line in Fig. 12a shows the solid line transformed to our metallicity scale using the prescription by Kewley & Ellison (2008). The observed metallicities are well below this line too. . The mass-metallicity relationship varies with redshift so that the same galaxy mass corresponds to lower oxygen metallicities at higher redshifts (a dex drop at redshift 2.3, according to Erb et al. 2006). Our tadpoles are on the mass-metallicity relationship observed at higher redshifts (see the dotted line in Fig. 12a).
We note that the metallicity of the heads also scales with their dynamical masses (Fig. 12b). The relationship is even tighter than the relationship with galaxy photometric mass (Fig. 12a), but we do not have enough points to judge whether or not the improvement is statistically significant. From a practical point of view, the reduction of scatter associated with the use of dynamical masses supports the reliability our estimate of this physical parameter.
8. Notes on individual galaxies
The general properties of the sample are discussed in detail in the preceding sections. Here we focus on a few specifics of the individual galaxies.
kiso3193. This is the only tadpole that shows no rotation, and its line-width versus position curve has a curious double-hump shape (Fig. 5). Its light distribution is less lopsided than for the other tadpoles (Figs. 1 and 4). The absence of rotation combined with the increase of linewidth toward the outskirts and the mild lopsidedness may be consistent with a face-on disk, and in this sense kiso3193 differs from the rest of the sample. The observed linewidths are fairly large, comparable to those of other larger galaxies in the sample. Consequently, the dynamical mass of the head inferred from linewidths is significantly larger than its stellar (photometric) mass.
kiso3867. In both mass and size, this is the smallest tadpole in the sample (see the photometric masses of the full galaxies in Table 2, and also the 1kpc scales given in Fig. 4). Its linewidths do not differ so much from the linewidths of other tadpoles, which contrasts with the low photometric mass inferred in Paper I for the head of the tadpole. Consequently, the ratio of dynamical mass to photometric mass is particularly high in this case – the dynamical mass of the head turns out to be a hundred times larger than the stellar mass derived from photometry (Table 2). The comparatively large oxygen abundance we derive for the head () is probably an overestimate due to observational errors. A few pixels away from the head the abundance drops by 0.3 dex (Fig. 10), which would bring the galaxy down to a more natural location in the mass-metallicity relationship (Fig. 12a).
kiso5149. The asymmetry of its RC is remarkable: see Fig. 3. It extends much further out to one side, which would imply that half of the disk is missing from observation. We cannot explain how this happens, unless the galaxy has a cigar-like shape rather than an axi-symmetric structure. It is unlikely that large reddening obscures half the disk, since the SDSS image does not show a large color gradient across the galaxy. It may be a merger, but then it would have to produce an unlikely large scale velocity field resembling a RC. The oxygen abundance is fairly constant along the galaxy, which seems to be associated with being a massive object (it has the largest mass of the sample – Table 2). The galaxy presents a second bright knot separated from the rotation center.
kiso5639. It is one of the XMP galaxies in the sample, with the head being a young starburst (a few Myr old). It also presents a very irregular RC, particularly in the outer parts.
kiso6669. This galaxy has a second bright knot, not far from the head (Fig. 4). The two knots are of low metallicity, which contrasts with the metallicity of the rest of the galaxy (Fig. 10). The center of the RC stays in between the two emission peaks of the galaxy (and so, off-centered; see Fig. 3). The linewidth curve shows a significant dip which does not coincide with any of the two emission peaks (Fig. 5).
kiso6877. This tadpole has the lowest metallicity and is the youngest object in the sample, and its bright head does not seem to contain dark matter. It also presents unusual high-excitation emission lines in the spectral region of the WR bumps (Sect. 5). As we explain in Sect. 6, we sought distortions in the RCs that could be associated with massive self-gravitating heads. Perhaps kiso6877 represents the only example. Although we cannot assess the reliability of the distortion present in its RC at the head position (see Fig. 3), we carried out the academic exercise of reproducing the observed velocity curve with two RCs combined, one for the galaxy plus one for the head. The center of rotation of the head has been forced to be given by the position of the head. The combined RC is shown as the dashed line in Fig. 13, and it improves the fit to the data points (compare with the solid line). The head is modeled with a counter-rotating disk with a maximum velocity of 6 km s, which according to Eq. (5) corresponds to a dynamical mass of . The mass thus derived is similar to both the photometric mass of the head and the dynamical mass inferred from linewidth (see Table 2).
We also note the coincidence of this wiggle in the RC with an obvious decrease in line-width associated with the head (see Fig. 5).
9. Discussion and conclusions
Galaxies with a bright peripheral clump on a fainter tail are called cometary or tadpole (Fig. 1). The origin of the shape is unknown, but it may trace a transit phase in the assembly of many disk galaxies (Sect. 1). Low mass local tadpoles were identified and studied photometrically in Paper I. Here we follow up the study, and analyze the chemical and dynamical properties of seven such targets inferred from long-slit spectra around H.
Five out the seven observed tadpoles show evidence for rotation ( %), and a sixth target hints at it. Often the center of rotation is spatially offset with respect to the tadpole head (three out of five cases). The RCs of the smaller targets are not smooth but present fluctuations, suggesting a complex dynamics (e.g., a counter-rotating head – see Fig. 13). The size and velocity dispersion of the heads are typical of giant HII regions and follow the scaling relationship known to exist between these two quantities (Fig. 6). We find changes of velocity dispersion along the galaxies, but they are not correlated with intensity variations. If anything, there is a tendency for the tadpole heads to coincide with local minima of velocity dispersion. The head is defined as the position on the galaxy with the largest surface brightness. Observationally, we find it to coincide with the region of largest H flux, and so, of largest SFR in the galaxy (e.g., Kennicutt 1998). Moreover, we also find the continuum flux to extend further out as compared to the flux in H, which is concentrated around the head. Thus the bright heads seem to be large starbursts with their random motions reduced with respect to the rest of the galaxy.
Using the observed RCs and velocity dispersions, we estimate the dynamical masses of the galaxies and their heads. The dynamical masses of the full galaxies are between three and ten times larger than the stellar mass inferred from photometry. The dynamical masses of the heads also exceed their stellar masses, but to a lesser extent than the full galaxies. Actually, the photometric mass and the dynamical mass of three heads agree within error bars. In two other cases, however, the dynamical mass of the head exceeds the photometric mass of the full galaxy.
The oxygen metallicity estimated from [NII]6583/H often shows significant spatial gradients across the galaxies (0.5 dex), being lowest at the head and increasing in the rest of the galaxy, tail included. So far as we are aware of, this is the first time that a metallicity growing away from HII regions has been reported in local galaxies. The sense of the resulting metallicity gradient is at variance with the observation of local disk galaxies, where the gas-phase metallicity increases toward the galaxy centers (Vilchez et al. 1988; Garnett et al. 1997) or is just constant (Moran et al. 2012). However, the type of variation we measure, with a minimum metallicity at the most intense star-forming region, has been observed in galaxies at redshift around 3 by Cresci et al. (2010) where it is interpreted as evidence for infall of pristine gas triggering star formation. Once systematic errors are disregarded (as we did in Sect. 7), it is difficult to avoid such interpretation, also in the case of our tadpole galaxies. We considered and then discarded the following two possibilities: (1) assume a regular metallicity distribution decreasing outward. If the head was formed from gas in the galaxy outskirts, but has spiraled in toward the center by dynamical friction (e.g., Elmegreen et al. 2012a), then the head would naturally present a metallicity lower than its immediate surroundings and similar to the galaxy outskirts. However, this prediction of metallicity gradients induced by internal migration is inconsistent with our observation, where the metallicity at the head is not just lower than the surroundings but the lowest (see Fig. 10). (2) Metal-rich supernova (SN) driven winds remove metals from shallow gravitational potential wells, producing metal-poor galaxies (e.g., Mac Low & Ferrara 1999; Recchi et al. 2004). This mechanism explains the overall low gas metallicity of some galaxies with significant old stellar populations, but it does not account for the presence of a region like the tadpole head, with metallicity lower than the rest of the galactic gas. The metals ejected by the winds are those created by the stars that explode as SNe. They are not the metals of the ambient gas. Then the loss of these ejecta reduces the efficiency of the galaxy to retain metals, but it does not reduce the metallicity of a particular region of the galaxy. In short, the interstellar medium of the tadpole galaxies is not well mixed but shows significant metallicity gradients. Since the mixing time scale is expected to be relatively short (shorter than a few Myr; e.g., Tenorio-Tagle 1996; de Avillez & Mac Low 2002), the recent infall of metal-poor gas seems to be the only viable alternative to explain the metallicity drop observed at the tadpole heads.
The observationally motivated interpretation of external gas infall fits in well the cold-flow gas accretion scenario arising from cosmological numerical simulations (e.g., Kereš et al. 2005; Dekel et al. 2009). It predicts localized accretion in clumpy streams of pristine gas ready to create stars. The streams may directly form giant clumps that we detect as tadpole heads or, alternatively, feed the disks with turbulent gas that eventually fragments into giant clumps by gravitational instability. In both cases the massive clumps are prone to migrate toward the galaxy centers and become progenitors of central spheroids (Elmegreen et al. 2008; Ceverino et al. 2010). Other details of the observed tadpole properties are also consistent with the cold-flow accretion scenario. The process is expected to be ubiquitous at high redshift. Then the flows fade away gradually in a process that has not being completed in small galaxies yet (e.g., Kereš et al. 2009). This prediction is consistent with the absence of low metallicity HII regions in large local spirals, as well as in our most massive tadpoles (kiso5149 and kiso8466; see Fig. 10), for which the infall would be already over. The metallicity drop is observed only in the low mass objects, reflecting the downsizing process in galaxy formation.
The geometrical displacement of some of the tadpole heads with respect to the centers of rotation also favors the cold-flow scenario. The expected streams of cold gas never end up at the galaxy center. The clumps are formed in the disk, and require time to be transported inward.
Extremely metal poor (XMP) galaxies are rare. Tadpole galaxies are also rare. The fact that we observe two XMP galaxies in a sample of seven tadpoles cannot be casual (Sect. 10). It is known that a significant fraction of XMP galaxies turns out to be tadpole or cometary (Papaderos et al. 2008; Morales-Luis et al. 2011). Here we find the reverse to be true as well, i.e., tadpoles have a significant chance of being XMP. The coincidence of these two seemingly disconnected properties is best understood if the objects are primitive, with the cometary shape and the low metallicity reflexing dynamical and chemical youth, respectively.
All these results combined are consistent with the local tadpole galaxies being turbulent disks in early stages of assembling. Their star formation seems to be sustained by accretion of external metal-poor gas.
- Abazajian et al. (2009) Abazajian, K. N., Adelman-McCarthy, J. K., Agüeros, M. A., et al. 2009, ApJS, 182, 543
- Andersen (1985) Andersen, T. B. 1985, Optical specifications and performance of the Nordic 2.5 M Telescope
- Asplund et al. (2009) Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481
- Baldwin et al. (1982) Baldwin, J. A., Spinrad, H., & Terlevich, R. 1982, MNRAS, 198, 535
- Balthasar et al. (1982) Balthasar, H., Thiele, U., & Woehl, H. 1982, A&A, 114, 357
- Bosch et al. (2009) Bosch, G., Terlevich, E., & Terlevich, R. 2009, AJ, 137, 3437
- Boselli et al. (2008) Boselli, A., Boissier, S., Cortese, L., & Gavazzi, G. 2008, ApJ, 674, 742
- Bournaud & Elmegreen (2009) Bournaud, F. & Elmegreen, B. G. 2009, ApJ, 694, L158
- Brinchmann et al. (2008) Brinchmann, J., Kunth, D., & Durret, F. 2008, A&A, 485, 657
- Campos-Aguilar et al. (1993) Campos-Aguilar, A., Moles, M., & Masegosa, J. 1993, AJ, 106, 1784
- Ceverino et al. (2010) Ceverino, D., Dekel, A., & Bournaud, F. 2010, MNRAS, 404, 2151
- Ceverino et al. (2012) Ceverino, D., Dekel, A., Mandelker, N., et al. 2012, MNRAS, 420, 3490
- Chung et al. (2009) Chung, A., van Gorkom, J. H., Kenney, J. D. P., Crowl, H., & Vollmer, B. 2009, AJ, 138, 1741
- Corbin et al. (2005) Corbin, M. R., Vacca, W. D., Hibbard, J. E., Somerville, R. S., & Windhorst, R. A. 2005, ApJ, 629, L89
- Cresci et al. (2010) Cresci, G., Mannucci, F., Maiolino, R., et al. 2010, Nature, 467, 811
- Crowther (2007) Crowther, P. A. 2007, ARA&A, 45, 177
- Dalcanton & Stilp (2010) Dalcanton, J. J. & Stilp, A. M. 2010, ApJ, 721, 547
- de Avillez & Mac Low (2002) de Avillez, M. A. & Mac Low, M.-M. 2002, ApJ, 581, 1047
- Dekel et al. (2009) Dekel, A., Birnboim, Y., Engel, G., et al. 2009, Nature, 457, 451
- Denicoló et al. (2002) Denicoló, G., Terlevich, R., & Terlevich, E. 2002, MNRAS, 330, 69
- D’Onghia et al. (2010) D’Onghia, E., Vogelsberger, M., Faucher-Giguere, C.-A., & Hernquist, L. 2010, ApJ, 725, 353
- Elmegreen et al. (2008) Elmegreen, B. G., Bournaud, F., & Elmegreen, D. M. 2008, ApJ, 688, 67
- Elmegreen & Elmegreen (2010) Elmegreen, B. G. & Elmegreen, D. M. 2010, ApJ, 722, 1895
- Elmegreen et al. (2012a) Elmegreen, B. G., Zhang, H.-X., & Hunter, D. A. 2012a, ApJ, 747, 105
- Elmegreen et al. (2009) Elmegreen, D. M., Elmegreen, B. G., Marcus, M. T., et al. 2009, ApJ, 701, 306
- Elmegreen et al. (2007) Elmegreen, D. M., Elmegreen, B. G., Ravindranath, S., & Coe, D. A. 2007, ApJ, 658, 763
- Elmegreen et al. (2012b) Elmegreen, D. M., Elmegreen, B. G., Sánchez Almeida, J., et al. 2012b, ApJ, 750, 95
- Erb et al. (2006) Erb, D. K., Shapley, A. E., Pettini, M., et al. 2006, ApJ, 644, 813
- Förster Schreiber et al. (2006) Förster Schreiber, N. M., Genzel, R., Lehnert, M. D., et al. 2006, ApJ, 645, 1062
- Franx et al. (1997) Franx, M., Illingworth, G. D., Kelson, D. D., van Dokkum, P. G., & Tran, K.-V. 1997, ApJ, 486, L75
- Fuentes-Masip et al. (2000) Fuentes-Masip, O., Muñoz-Tuñón, C., Castañeda, H. O., & Tenorio-Tagle, G. 2000, AJ, 120, 752
- Garnett et al. (1997) Garnett, D. R., Shields, G. A., Skillman, E. D., Sagan, S. P., & Dufour, R. J. 1997, ApJ, 489, 63
- Genel et al. (2012) Genel, S., Naab, T., Genzel, R., et al. 2012, ApJ, 745, 11
- Genzel et al. (2008) Genzel, R., Burkert, A., Bouché, N., et al. 2008, ApJ, 687, 59
- Guseva et al. (2003) Guseva, N. G., Papaderos, P., Izotov, Y. I., et al. 2003, A&A, 407, 105
- Immeli et al. (2004) Immeli, A., Samland, M., Westera, P., & Gerhard, O. 2004, ApJ, 611, 20
- Izotov & Thuan (2004) Izotov, Y. I. & Thuan, T. X. 2004, ApJ, 616, 768
- Kehrig et al. (2011) Kehrig, C., Oey, M. S., Crowther, P. A., et al. 2011, A&A, 526, A128
- Kennicutt (1998) Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189
- Kereš et al. (2009) Kereš, D., Katz, N., Fardal, M., Davé, R., & Weinberg, D. H. 2009, MNRAS, 395, 160
- Kereš et al. (2005) Kereš, D., Katz, N., Weinberg, D. H., & Davé, R. 2005, MNRAS, 363, 2
- Kewley & Ellison (2008) Kewley, L. J. & Ellison, S. L. 2008, ApJ, 681, 1183
- Kunth & Östlin (2000) Kunth, D. & Östlin, G. 2000, A&A Rev., 10, 1
- Laing & Jones (1985) Laing, R. & Jones, D. 1985, Vistas in Astronomy, 28, 483
- Leitherer et al. (1999) Leitherer, C., Schaerer, D., Goldader, J. D., et al. 1999, ApJS, 123, 3
- Lin & Faber (1983) Lin, D. N. C. & Faber, S. M. 1983, ApJ, 266, L21
- López-Sánchez et al. (2012) López-Sánchez, Á. R., Koribalski, B. S., van Eymeren, J., et al. 2012, MNRAS, 419, 1051
- Mac Low & Ferrara (1999) Mac Low, M.-M. & Ferrara, A. 1999, ApJ, 513, 142
- MacAlpine & Lewis (1978) MacAlpine, G. M. & Lewis, D. W. 1978, ApJS, 36, 587
- Martin (1971) Martin, B. R. 1971, Statistics for Physicists (London: Academic Press)
- Mayer et al. (2006) Mayer, L., Mastropietro, C., Wadsley, J., Stadel, J., & Moore, B. 2006, MNRAS, 369, 1021
- Melnick et al. (1999) Melnick, J., Tenorio-Tagle, G., & Terlevich, R. 1999, MNRAS, 302, 677
- Miyauchi-Isobe et al. (2010) Miyauchi-Isobe, N., Maehara, H., & Nakajima, K. 2010, Publications of the National Astronomical Observatory of Japan, 13, 9
- Morales-Luis et al. (2011) Morales-Luis, A. B., Sánchez Almeida, J., Aguerri, J. A. L., & Muñoz-Tuñón, C. 2011, ApJ, 743, 77
- Moran et al. (2012) Moran, S. M., Heckman, T. M., Kauffmann, G., et al. 2012, ApJ, 745, 66
- Muñoz-Tuñon (1994) Muñoz-Tuñon, C. 1994, Supersonic motions in giant H II regions., ed. G. Tenorio-Tagle, 25
- Noguchi (1999) Noguchi, M. 1999, ApJ, 514, 77
- O’dell & Townsley (1988) O’dell, C. R. & Townsley, L. K. 1988, A&A, 198, 283
- Osterbrock (1989) Osterbrock, D. E. 1989, Astrophysics of gaseous nebulae and active galactic nuclei
- Pagel & Edmunds (1981) Pagel, B. E. J. & Edmunds, M. G. 1981, ARA&A, 19, 77
- Pagel et al. (1992) Pagel, B. E. J., Simonson, E. A., Terlevich, R. J., & Edmunds, M. G. 1992, MNRAS, 255, 325
- Papaderos et al. (2008) Papaderos, P., Guseva, N. G., Izotov, Y. I., & Fricke, K. J. 2008, A&A, 491, 113
- Pérez-Montero & Contini (2009) Pérez-Montero, E. & Contini, T. 2009, MNRAS, 398, 949
- Pettini & Pagel (2004) Pettini, M. & Pagel, B. E. J. 2004, MNRAS, 348, L59
- Press et al. (1986) Press, W. H., Flannery, B. P., & Teukolsky, S. A. 1986, Numerical recipes. The art of scientific computing
- Recchi et al. (2004) Recchi, S., Matteucci, F., D’Ercole, A., & Tosi, M. 2004, A&A, 426, 37
- Roy et al. (1986) Roy, J.-R., Arsenault, R., & Joncas, G. 1986, ApJ, 300, 624
- Rozas et al. (2006) Rozas, M., Richer, M. G., López, J. A., Relaño, M., & Beckman, J. E. 2006, A&A, 455, 539
- Salucci et al. (2007) Salucci, P., Lapi, A., Tonini, C., et al. 2007, MNRAS, 378, 41
- Salucci & Persic (1997) Salucci, P. & Persic, M. 1997, in Astronomical Society of the Pacific Conference Series, Vol. 117, Dark and Visible Matter in Galaxies and Cosmological Implications, ed. M. Persic & P. Salucci, 1
- Sánchez Almeida et al. (2009) Sánchez Almeida, J., Aguerri, J. A. L., Muñoz-Tuñón, C., & Vazdekis, A. 2009, ApJ, 698, 1497
- Schaerer et al. (1999) Schaerer, D., Contini, T., & Pindao, M. 1999, A&AS, 136, 35
- Shi et al. (2005) Shi, F., Kong, X., Li, C., & Cheng, F. Z. 2005, A&A, 437, 849
- Shirazi & Brinchmann (2012) Shirazi, M. & Brinchmann, J. 2012, MNRAS, 421, 1043
- Stoughton et al. (2002) Stoughton, C., Lupton, R. H., Bernardi, M., et al. 2002, AJ, 123, 485
- Straughn et al. (2006) Straughn, A. N., Cohen, S. H., Ryan, R. E., et al. 2006, ApJ, 639, 724
- Tenorio-Tagle (1996) Tenorio-Tagle, G. 1996, AJ, 111, 1641
- Terlevich & Melnick (1981) Terlevich, R. & Melnick, J. 1981, MNRAS, 195, 839
- Tremonti et al. (2004) Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ, 613, 898
- van den Bergh (1994) van den Bergh, S. 1994, ApJ, 428, 617
- van Zee et al. (2004) van Zee, L., Skillman, E. D., & Haynes, M. P. 2004, AJ, 128, 121
- Vilchez et al. (1988) Vilchez, J. M., Pagel, B. E. J., Diaz, A. I., Terlevich, E., & Edmunds, M. G. 1988, MNRAS, 235, 633
- Windhorst et al. (2006) Windhorst, R. A., Cohen, S. H., Straughn, A. N., et al. 2006, New A Rev., 50, 821
- Wisnioski et al. (2012) Wisnioski, E., Glazebrook, K., Blake, C., et al. 2012, MNRAS, 422, 3339