DWARFS GOBBLING DWARFS: A STELLAR TIDAL STREAM AROUND NGC 4449 AND HIERARCHICAL GALAXY FORMATION ON SMALL SCALES

A fundamental characteristic of the modern cold dark matter (ΛCDM) cosmology (Mo et al. 2010) is that galaxies assemble hierarchically under the influence of gravity, continually accreting smaller DM halos up until the present day. If those halos contain stars, then they will be visible as satellites around their host galaxies, eventually disrupting through tidal forces into distinct streams and shells before phase mixing into obscurity (Cooper et al. 2010). This picture seems to explain the existence of satellites and substructures observed around massive galaxies (Arp 1966; Schweizer & Seitzer 1988; McConnachie et al. 2009; Martínez-Delgado et al. 2010). However, quantitative confirmation of this aspect of ΛCDM has been more elusive, with lingering doubts provoked by small-scale substructure observations (Lovell et al. 2012; Boylan-Kolchin et al. 2012; Ferrero et al. 2012).

There has been relatively little work on substructure and merging in the halos around low-mass, "dwarf" galaxies. Many cases of extended stellar halos around dwarfs have been identified observationally (Stinson et al. 2009), but it is not clear if these stars were accreted, or formed in situ. Star formation in dwarfs is thought to occur in stochastic episodes (Tolstoy et al. 2009; Weisz et al. 2011), which could be triggered by accretion events.

An iconic galaxy in this context is NGC 4449, a dwarf irregular in the field that has been studied intensively as one of the strongest galaxy-wide starbursts in the nearby universe. Its absolute magnitude of M_V  =  − 18.6 makes it a Large Magellanic Cloud (LMC) analogue (and not formally a dwarf by some definitions), but with a much higher star formation rate. It is strongly suspected to have recently interacted with another galaxy based on various signatures including peculiar kinematics in its cold gas and H ii regions (Hartmann et al. 1986; Hunter et al. 1998), but the nature of this interaction is unknown.

An elongated dwarf galaxy or stream candidate near NGC 4449 was first noticed by Karachentsev et al. (2007) from Digitized Sky Survey (POSS-II) plates (object d1228+4358), and is visible in the Sloan Digital Sky Survey (SDSS).¹⁴ Here we present deep, wide-field optical imaging that supplies the definitive detection of this ongoing accretion event involving a smaller galaxy, leading to interesting implications about the evolution of this system and of dwarf galaxies in general.

Our observations of NGC 4449 and its surroundings consist of two main components. The first is imaging from a small robotic telescope to confirm the presence of a low-surface-brightness substructure and provide its basic characteristics (similar techniques were used with larger galaxies in Martínez-Delgado et al. 2008, 2010). The second is follow-up imaging with the Subaru Telescope to map out the resolved stellar populations.

We obtained very deep images with the f/8.3 Ritchey–Chretien 0.5 m telescope of the BlackBird Remote Observatory (BBRO)¹⁵ during different dark-sky observing runs over the periods 2010 April 13 through June 10 and 2011 January 13 through 28 (UT). We used a 16 mega-pixel Apogee Imaging Systems U16M CCD camera, with 313 × 313 field-of-view and 0.46 arcsec pixel⁻¹ plate scale. We acquired 18 hr of imaging data in half-hour sub-exposures, using a non-infrared clear luminance (λ = 3500–8500 Å) Astrodon E-series filter. Each sub-exposure was reduced following standard procedures for dark-subtraction, bias-correction, and flat-fielding (Martínez-Delgado et al. 2009).

The resulting image was calibrated photometrically to SDSS using the brighter regions of NGC 4449 (see Martínez-Delgado et al. 2010). The final image has 5σ g-band surface-brightness detection limits from 26.4 to 27.5 mag arcsec⁻² for seeing-limited and large-scale diffuse features, respectively.

We subsequently obtained images from the 8.2 m Subaru Telescope and the Suprime-Cam wide-field imager (34' × 27' field-of-view, 0202 pixel scale; Miyazaki et al. 2002) on 2011 January 5 (UT). Conditions were photometric, and we took dithered exposures in r' and i' bands, with total exposure times of 225 s per filter. We reduced the data using a modified SDFRED pipeline (Ouchi et al. 2004), including bias subtraction, flat-fielding, and distortion correction. Each frame was re-projected to a common astrometric coordinate system followed by background rectification and image co-addition using Montage.¹⁶

The exquisite image quality (∼05 FWHM) allows us to resolve individual stars in the outer regions of NGC 4449. We carried out point-spread-function photometry using DAOPHOT II/ALLSTARS (Stetson 1987) and identified stars as objects with sharpness parameter |S| < 1.0. We calibrated the photometry based on two central images from the Hubble Space Telescope Advanced Camera for Surveys (HST/ACS; Annibali et al. 2008, hereafter A+08).

The ACS photometry was originally in F555W and F814W, and recalibrated to Johnson-Cousins VI (A+08). We used fairly bright, red stars in common between the data sets to derive linear transformation equations from r'i' instrumental magnitudes to VI, including foreground extinction corrections of E(B − V) = 0.019 (Schlegel et al. 1998). Our final star catalog has statistical internal errors in V − I color of ∼0.11 mag and ∼0.14 mag at V = 25 and V = 26, respectively.

Figure 1 (left) shows a BBRO image subsection, where with an adopted distance of 3.82 Mpc (A+08), 1' corresponds to 1.1 kpc. Clearly visible near the minor axis of NGC 4449, ∼10 kpc to the southeast, is a very elongated, S-shaped feature of dimensions ∼1.5 × 7 kpc, which we designate "the stream." The stream's position does not overlap with any of the complex H ii-gas features surrounding NGC 4449 (Hunter et al. 1998). Also, it is on the opposite side of the galaxy with respect to an interesting star cluster that may be linked to another past accretion event (Annibali et al. 2012). The main galaxy's existing stellar halo is also apparent, including shell-like features in the southwest previously noticed by Hunter et al. (1999).

Zoom In Zoom Out Reset image size

The right-hand panel shows a Suprime-Cam zoom-in of the stream. It is clearly resolved into stars and has the general appearance of a dwarf spheroidal (dSph) galaxy that is elongated by tides. This classification (defined by a spheroidal morphology, stellar mass M_⋆ < 10⁸ M_☉, and the absence of star formation and cold gas) is also supported by the H i non-detection of Huchtmeier et al. (2009).

To complement these integrated surface-brightness maps, we construct stellar-density maps around NGC 4449 using individual red giant branch (RGB) stars down to g' = 25.8 from Suprime-Cam. We discuss the RGB selection and modeling later, but in general these stars trace a population older than 1 Gyr. We define a grid across the image with 130 × 130 pc² bins and count the number of stars in each bin, subsequently applying a Gaussian smoothing kernel with σ_smooth = 130 pc. There are typically seven to nine stars per bin in the stream.

Figure 2(a) shows the resulting Subaru-based stellar-density map, where the overall stream morphology is indistinguishable from the BBRO integrated-light results. Since there is no evidence for resolved young stars (see next section), we infer that the visible light is dominated by RGB stars and hence by a population older than 1 Gyr.

Zoom In Zoom Out Reset image size

Panel (b) shows a higher-contrast version of the same map, demonstrating that the main galaxy's preexisting stellar halo extends out to at least ∼10 kpc. We can furthermore discern a very faint feature that seems to extend the stream's angular path in a loop-like structure. This loop is also apparent in the BBRO image, and we infer that the stream is stretched out over at least half its orbit, with the projected turning point at a galactocentric radius of 13 kpc.

We next wish to locate the disrupting satellite galaxy's original center or nucleus. In both the BBRO and Subaru images there is a density enhancement near the mid-point of the "S," as would be expected if the "arms" are leading and trailing tidal tails around a still marginally bound, or just disrupting, main body. We construct a stellar-density contour map for the stream region, using RGB stars and σ_smooth = 270 pc. The central stellar clump is visible in Figure 2(d), with a position slightly offset from the stream's main ridgeline (α_J2000 = 122843.32, δ_J2000 = +435830.0; positional uncertainty ∼6''). An off-center nucleus is also seen in a tidally disrupting Milky Way dSph, Ursa Major (Martínez-Delgado et al. 2001).

Figure 3 shows the color–magnitude diagram (CMD) for point sources in the stream region. The RGB stars are the dominant feature, along with a few brighter, redder stars that may be oxygen- or carbon-rich thermally pulsating asymptotic giant branch stars from an intermediate-age or old population (Marigo et al. 2003). We find no blue stars that would trace recent star formation.

Zoom In Zoom Out Reset image size

The detection of the tip of the RGB (TRGB) permits a distance estimate. Using techniques from A+08 and Cioni et al. (2000), we find a TRGB magnitude of I_TRGB = 24.06 ± 0.04 (random) ±0.08 (systematic). The random error was estimated using bootstrapping techniques; the systematic error is dominated by the magnitude-transformation uncertainties. The main body of NGC 4449 was found by A+08 to have I_TRGB = 24.00 ± 0.01 (random) ±0.04 (systematic). Thus, the stream is at the same distance as the main body, to within ∼180 kpc, and we conclude that there is a physical association rather than a chance superposition.

Although the RGB is affected by the well-known age–metallicity degeneracy, this feature can still be used to constrain the properties of stars older than ∼1 Gyr. In Figure 3, we overplot the Padua isochrones (Girardi et al. 2002) for ages of 2, 4, and 10 Gyr, for both Z = 0.004 and Z = 0.001. For Z = 0.004 the 2–4 Gyr isochrones trace the data reasonably well. The Z = 0.001 models appear significantly bluer than the mean RGB color (by ≳0.13 mag: our systematic color uncertainties are ≲0.1 mag), although a Z = 0.001, 10 Gyr model is consistent with the blue edge of the observed RGB. Presumably a 10 Gyr model with Z  ∼  0.002 would also be consistent with the data, while perhaps providing a better match to the observed CMD slope. We conclude that, if the bulk of the RGB stars are old (>10 Gyr), the metallicity range is roughly Z = 0.001–0.004, while for younger ages the metallicity range shifts to higher values.

These results are comparable to the RGB analysis of the main body of NGC 4449 by A+08 (their Figure 17; see also Ryś et al. 2011). Therefore, both the main galaxy and the stream contain similar old, intermediate-metallicity populations, although the main galaxy also contains very young stars, as well as more metal-poor stars as inferred from its globular clusters (Strader et al. 2012).

We provide a preliminary overview of spatial stellar population variations by splitting the RGBs into color-based subpopulations, using the 4 Gyr, Z = 0.004 model as a boundary. We then create stellar-density maps as before, for the two subpopulations separately. Using blue and green color coding to represent the subpopulations left and right of the model boundary, we show the results in Figure 2(c). The stream's bright parts have no visible RGB color gradient and have an overall color similar to the main galaxy's halo at radii of ∼3–5 kpc. Both the stream's faint-loop continuation and the halo at ∼5–10 kpc are redder, implying older or more metal-rich stars.

We now proceed to estimate the luminosities and masses of NGC 4449 and its stream. For NGC 4449, it is straightforward to add up the SDSS pixel fluxes inside the "optical radius" (μ_r = 25 mag arcsec⁻²). We find an extinction-corrected M_r = −17.8. For the stream, we use the SDSS-calibrated BBRO image, integrating the flux within the faintest isophote that closes without including the main galaxy, which corresponds to μ_g = 26.75 mag arcsec⁻² and a stream semi-major axis distance of 3 kpc. We find a stream magnitude of M_r = −13.5, which is comparable to the brightest Local Group dSphs, Fornax, and And VII. Such galaxies have typical projected half-light radii of ∼0.4–1.0 kpc (Brodie et al. 2011), which is consistent with the stream's ∼0.8 kpc half-width.

For both galaxies, these luminosity estimates are lower limits since they do not include potential extended low-surface-brightness features. The implied luminosity ratio is ∼1:50.

To calculate stellar masses M_⋆, we use two different approaches, adopting a Chabrier (2003) initial mass function (IMF; final mass, including stellar remnants). The first is based on the integrated optical colors, following the relations between color and stellar mass-to-light ratio (ϒ_⋆) from Zibetti et al. (2009, Table B1). This Letter also introduced a technique for mapping out local ϒ_⋆ and M_⋆ variations pixel by pixel, which we apply to NGC 4449, and after integrating, find a total M_⋆ = 7.46 × 10⁸ M_☉. For the stream, we assume a global color corresponding to the central SDSS measurement, g − r = 0.45 ± 0.1. We find a stream mass of M_⋆ = 1.5^+0.8 _{− 0.6} × 10⁷ M_☉, implying a stellar-mass ratio between stream and host of ∼1:50, for any uniform IMF.¹⁷

The second mass-estimation approach uses the CMD, comparing observed star counts to predicted numbers from a stellar populations model. For the stream, we use the I-band stellar luminosity function near the TRGB and normalize it to Monte Carlo simulations drawn assuming Z = 0.001–0.004, and ages 2–10 Gyr. We obtain M_⋆ ∼ (2–5.5) × 10⁷ M_☉ for the stream (a lower limit because of incompleteness), which agrees with the color-based results. For the main galaxy, a similar approach was followed by McQuinn et al. (2010), whose results imply M_⋆ = (1.2 ± 0.2) × 10⁹ M_☉. This mass is somewhat higher than by using colors, but the CMD-based stellar-mass ratio turns out to be similar, ∼1:40.

Remarkably, the mass of NGC 4449's stellar halo is similar to the stream's mass: based on the RGB counts of Ryś et al. (2011) and their normalization to K-band surface-brightness photometry, we estimate M_⋆ ∼ 2 × 10⁷ M_☉ for the halo between projected radii of 5–10 kpc. This halo could have therefore been built up directly by one or a few past accretion events similar to the present-day stream.

We next consider the dynamical masses of the host galaxy and its stream, including DM. The quantity that is arguably the most relevant to the current stream-galaxy interaction is the dynamical mass ratio within the interaction region: the ∼14 kpc galactocentric radius. Based on the H i gas kinematics, we estimate an inclination-corrected circular velocity of v_c ≃ 62 km s⁻¹ at this radius (Bajaja et al. 1994; Hunter et al. 2002), which means a dynamical mass for the main galaxy of M_dyn(r  <  15 kpc) ≃ 1.1 × 10¹⁰ M_☉. The H i gas mass is ∼10⁹ M_☉ (Hunter et al. 1998), so this region is DM dominated. Note that the v_c and M_⋆ values together suggest that NGC 4449 is intermediate in mass to the LMC and Small Magellanic Cloud (SMC; cf. Besla et al. 2010).

For the stream mass, we have no direct measurements, and instead turn to a plausibility argument based on Local Group dSphs, where the brightest cases have estimated circular velocities of v_c ∼ 15–20 km s⁻¹ on ∼1–3 kpc scales (Walker & Peñarrubia 2011; Boylan-Kolchin et al. 2012). Then if we assume the v_c values for both stream and main galaxy are fairly constant with radius, the ratio of v²_c yields the dynamical mass ratio. This ratio is ∼1:20–1:10, and thus the stream may be significantly perturbing the main galaxy.

A final metric is the ratio of total (virial) halo masses M_vir, which are not directly measurable but may be inferred on a statistical basis, assuming a ΛCDM framework. In this context, it is well established that the total mass-to-light ratios of dwarf galaxies increase dramatically at lower luminosities. Current estimates of M_⋆–M_vir and luminosity–M_vir relations (Moster et al. 2010; Tollerud et al. 2011) would imply M_vir ∼ (1–5) × 10¹¹ M_☉ for NGC 4449, and a pre-infall mass of ∼(1–10) × 10¹⁰ M_☉ for the stream progenitor—which although very uncertain, plausibly implies an initial virial mass ratio of ∼1:10–1:5.

Thus what appears to be a very minor merger in visible light may actually be closer to a major merger when including DM. Such an extreme circumstance could be compared with models of satellite disruption and potentially discriminate between ΛCDM and alternative theories (McGaugh & Wolf 2010).

We have detected and analyzed a stellar tidal stream in the halo of NGC 4449, which we interpret as the ongoing disruption of a dSph galaxy by a larger dwarf (an LMC/SMC analogue¹⁸). This appears to be the lowest-mass primary galaxy with a verified stellar stream.

We suggest some implications for galaxy evolution. It has been proposed that dSph's orbiting massive galaxies such as the Milky Way were "pre-processed" from gas-rich dwarfs by tidal effects within dwarf-galaxy groups (D'Onghia et al. 2009). We may be witnessing such a transformation in-action, with the H i streams surrounding NGC 4449 representing additional tidal debris.

We also suspect it is not just a coincidence that such a novel stream was found first around one of the most intensely star-forming nearby galaxies. The accretion event may well be the starburst trigger. The period of elevated star formation appears to have started ∼0.5 Gyr ago (McQuinn et al. 2010), which is suggestively similar to the stream's ∼1–2 Gyr orbital period¹⁹ (and to any process that is linked to the dynamical time on ∼30 kpc scales).

Are such accretion events frequent among other dwarf galaxies in recent epochs? We suspect that exact analogues to this stream are not very common, or they would have been noticed already in DSS/SDSS images. However, if the stream had been only a bit fainter, more diffuse, or at a larger radius, it could have been missed, and thus there may be many more dwarf-hosted stellar streams awaiting detection.

In theory, the history of DM halo assembly should be fairly scale free, and ∼1:10 mergers are expected to be the most generally dominant contributors to mass growth (Stewart et al. 2008). It is also increasingly recognized that such relatively minor mergers can have important effects on the larger galaxies, such as inciting global disk instabilities (Purcell et al. 2011).

If streams as in NGC 4449 are common in dwarfs, they re-ignite classic ideas about galaxy interactions triggering starbursts. Given the high rates of star formation in dwarf galaxies, it is natural to ask if satellites are responsible. Surveys along these lines have produced mixed results (Noeske et al. 2001; Brosch et al. 2004; Li et al. 2008), but until now, low-surface-brightness objects such as dSphs would have been missed.

Regardless of the implications for starbursts, dSph accretion appears to be an increasingly viable avenue for direct assembly of dwarf galaxies' stellar halos—as witnessed by NGC 4449, and by Fornax, which shows traces of swallowing an even smaller dSph (Coleman et al. 2005). Future observational determinations of dwarf stream frequency in combination with theoretical models may provide clues to the general substructure problem.

We thank Jay Strader for a preview of his paper, and James Bullock, Pavel Kroupa, Jorge Peñarrubia, Monica Tosi, and the referee for comments. Based on data collected at the Subaru Telescope (operated by the National Astronomical Observatory of Japan), via Gemini Observatory time exchange (GN-2010B-C-204). F.A. received partial financial support from ASI, through contracts COFIS ASI-INAF I/016/07/0 and I/009/10/0. The Dark Cosmology Centre is funded by the Danish National Research Foundation. Work was supported by the National Science Foundation (Grants AST-0808099, AST-0909237, AST-1109878, Graduate Research Fellowship), by NASA/Spitzer grant JPL-1310512, and by the UCSC-UARC Aligned Research Program.