MID-INFRARED COLORS OF DWARF GALAXIES: YOUNG STARBURSTS MIMICKING ACTIVE GALACTIC NUCLEI

There is now an overwhelming body of evidence suggesting that all massive galaxies host a central supermassive black hole that grows alongside the stellar population (Kormendy & Richstone 1995; Kormendy & Ho 2013). In addition, a relationship has been observed between the central black hole mass and the galaxy bulge stellar velocity dispersion that spans many orders of magnitude (Ferrarese & Merritt 2000; Gebhardt et al. 2000). The origin of this relationship is still not well understood, but current theories imply that galaxy mergers and interactions play a role, both in increasing the stellar mass of a galaxy and in driving gas toward the centers of galaxies, feeding black holes (Hopkins et al. 2008; Koss et al. 2010; Ellison et al. 2011; Bessiere et al. 2012; Sabater et al. 2013). However, this view of galaxy growth implies the existence of low-mass "seed" black holes that must have existed at high redshift. To understand these difficult-to-observe objects, researchers have turned to observations of nearby dwarf galaxies that may host analogous lower-mass black holes (for a review, see Reines & Comastri 2016). By assembling large samples of low-mass black holes, it may be possible to distinguish between the different proposed theoretical scenarios for their creation: these objects may be remnants of massive Population III stars (Bromm & Yoshida 2011), a result of direct collapse of primordial dense gas (Haehnelt & Rees 1993; Begelman et al. 2006; Lodato & Natarajan 2006; van Wassenhove et al. 2010), or perhaps they are the end product of very massive stars formed through stellar mergers in dense star clusters (Gürkan et al. 2004; Freitag et al. 2006; Goswami et al. 2012; Giersz et al. 2015; Lützgendorf et al. 2016).

Assembling these large samples of low-mass black holes is made difficult by the fact that resolving their gravitational sphere of influence is currently not feasible at distances larger than a few megaparsecs. However, active galactic nucleus (AGN) emission across the electromagnetic spectrum can be used to infer the existence of a black hole. Observations at optical wavelengths have been used to uncover AGNs in NGC 4395 (Filippenko & Sargent 1989; Filippenko & Ho 2003) and POX 52 (Barth et al. 2004), while data at X-ray and radio wavelengths have been used to find AGNs in both Henize 2–10 (Reines et al. 2011; Reines & Deller 2012) and the dwarf galaxy pair Mrk 709 (Reines et al. 2014). Larger samples of low-mass AGNs have been uncovered at optical (Greene & Ho 2004, 2007; Dong et al. 2012; Reines et al. 2013; Moran et al. 2014) and X-ray (Lemons et al. 2015; Mezcua et al. 2015; Pardo et al. 2016) wavelengths, which have been targeted with successful follow-up observations (Baldassare et al. 2016), including the discovery of a 5 × 10⁴ M_☉ BH in RGG 118 (Baldassare et al. 2015).

One important and often-used method for selecting luminous AGNs relies on observations made in the mid-IR, where dust, heated by the central accreting black hole, reprocesses the light and emits with a characteristic red IR power-law spectrum. Infrared emission only minimally suffers from nuclear and galaxy-scale obscuration, and so mid-IR observations have successfully uncovered large numbers of unobscured and obscured luminous AGNs and quasars (Lacy et al. 2004, 2013; Stern et al. 2005, 2012; Hickox et al. 2007; Donley et al. 2008; Ashby et al. 2009; Assef et al. 2010; Mendez et al. 2013; Hainline et al. 2014). The all-sky mid-IR coverage of the Wide-field Infrared Survey Explorer (WISE, Wright et al. 2010) has allowed for observations of large samples of objects, and multiple authors have proposed WISE color schemes that select for the red AGN power-law emission in the infrared (Jarrett et al. 2011; Mateos et al. 2012; Stern et al. 2012). These selection methods rely on the fact that AGNs are capable of heating dust to temperatures well above what is observed from stellar processes in moderate to high-mass galaxies, and have demonstrated high levels of reliability when applied to these objects.

Recently, Satyapal et al. (2014) and Sartori et al. (2015) used mid-IR selection methods to assemble large samples of low-mass galaxy AGN candidates from WISE data. Under the assumption that WISE selection targets optically obscured, "hidden" AGNs in these objects, these authors draw broad conclusions about the population of low-mass black holes. The Satyapal et al. (2014) study targets "bulgeless" galaxies without optical evidence for an AGN, and the authors conclude that star formation is not the primary source of the IR emission in their sample. In addition, they propose that the fraction of galaxies hosting IR-selected AGN activity increases at low masses. This puzzling trend was also seen (with lower significance), using a larger sample of dwarf galaxies, by Sartori et al. (2015), who compared multiple AGN selection methods and concluded that dwarf galaxies with WISE colors indicative of AGN activity are bluer and potentially may have more ongoing star formation and lower metallicities than those selected using optical emission lines.

These results are intriguing in light of the fact that no other tracer of AGN activity has thus far uncovered such large samples of AGNs in dwarf galaxies. In most AGN selection regimes, more luminous AGNs are easier to find both because massive BHs have a higher Eddington limit and because of increasing confusion due to star formation at low mass. Thus, the observed increase in AGN fraction at the lowest dwarf galaxy masses is puzzling. Low-mass SMBHs that exist in dwarf galaxies power AGNs that have such low luminosities that star formation in their hosts becomes a significant source of contamination. In particular, it has been shown that low-metallicity dwarf starburst galaxies are capable of heating dust to very high temperatures (Hirashita & Hunt 2004; Reines et al. 2008; Griffith et al. 2011; Izotov et al. 2011, 2014; Rémy-Ruyer et al. 2015), producing red mid-IR colors. This was recently explored in O'Connor et al. (2016), who found that galaxies with low stellar masses have predominantly red WISE colors, which the authors associate with higher specific star formation rates (sSFRs) in these galaxies. Thus, it may be that using common mid-IR AGN selection methods on dwarf galaxies results in the selection of a large number of star-forming galaxies that contaminate the samples, leading to erroneous conclusions about AGN fractions at these masses.

In this paper, we use WISE data to empirically examine the infrared colors of dwarf galaxies as a function of their properties as probed by Sloan Digital Sky Survey (SDSS) data. Our results suggest that star formation, and not AGN activity, is responsible for the red WISE colors for the majority of the dwarf galaxy population (also see Izotov et al. 2014). We propose a small sample of dwarf galaxy AGN candidates that require follow-up observations to confirm black hole activity.

In Section 2, we discuss our sample selection and describe how these objects were matched to WISE photometry. We start by exploring the infrared properties of dwarf galaxies that have optical spectroscopic evidence for an AGN in Section 3.1. We then investigate optically selected star-forming galaxies and propose a sample of IR-selected AGN candidates in Section 3.2. We expand our sample to include IR-selected AGN candidates without optical emission-line flux measurements in Section 3.3, after which we compare these objects to samples of IR-selected AGN candidates presented in the literature in Section 3.4. Finally, we explore our WISE detection limits in Section 3.5, and we discuss our results and draw conclusions in Section 4.

The dwarf galaxies we explore in this paper were selected from the NASA Sloan Atlas (NSA), a catalog of galaxies at z < 0.055 selected from the SDSS DR8 (York & SDSS Collaboration 2000; Aihara et al. 2011), where the observations were re-analyzed, resulting in improved photometry (Blanton et al. 2011) and spectroscopy (Yan 2011; Yan & Blanton 2012). In addition, stellar masses for each object were derived with the kcorrect code of Blanton & Roweis (2007), which uses the stellar population synthesis models of Bruzual & Charlot (2003) and the nebular emission-line models of Kewley et al. (2001). These masses are provided in the NSA in units of M_☉ h⁻², and in this paper we have assumed h = 0.73.

Following Reines et al. (2013), we selected all galaxies with M_* < 3 × 10⁹ M_☉, the approximate mass of the Large Magellanic Cloud (van der Marel et al. 2002). We matched those objects to the AllWISE data release, which provides increased sensitivity compared to the previous WISE All-sky data release.⁶ We note that because of the polar orbit of the WISE telescope, the total exposure time across the full sky is non-uniform and depends on both the sky position and the zodiacal foreground emission. As the $W1-W2$ color forms the basis for both of the primary AGN selection criteria that we will explore in this study (that of Jarrett et al. 2011 and Stern et al. 2012), and given the importance of deep data at λ < 5 μm for separating emission from dust heated by star formation and AGN activity, we chose to use the updated AllWISE data, with W1 (3.4 μm) and W2 (4.6 μm) 5σ sensitivities of 54 μJy and 71 μJy, respectively (compared to 68 μJy and 111 μJy in the All-sky data release). These sensitivities are the minimum depth of the survey estimated for low coverage sky away from the Galactic plane.

When cross-matching with the AllWISE catalog, we used a 5'' matching radius, and then chose the specific WISE photometry for each object based on the WISE ext_flg value. Following the AllWISE Explanatory Supplement,⁶ for most objects (ext_flg = 0 or 4), we chose the "profile-fitting" photometric magnitudes, which are optimized for objects that are unresolved in WISE. For those objects where ext_flg = 1, 2, or 3, we used the "standard" aperture magnitudes, which better traces the photometry for objects that are resolved. Finally, for a small subset of objects with 2MASS photometry, where ext_flg = 5, we used the photometry using apertures scaled from the 2MASS XSC shape values. Each object was only included in our sample if the signal-to-noise ratio (S/N) for the W1-, W2-, and W3-band photometry was greater than 3, yielding a sample of 18,482 dwarf galaxies with significant WISE detections. The majority of the objects were removed from our final sample because of their low S/N W3 photometry (∼58% of the total NSA sample, compared to ∼4% and ∼7% for W1 and W2, respectively). The median $W1-W2$ and $W2-W3$ color uncertainties are 0.07 and 0.19, respectively, which we show with error bars in our WISE color–color diagrams.

While we have attempted to account for resolution effects with our usage of the WISE ext_flg value to choose appropriate photometry for each object, we also wish to highlight some systematic effects that could affect our results. First, the W1, W2, and W3 bands have angular resolutions of 61, 64, and 65, respectively (Wright et al. 2010). As a result, for any object that is slightly resolved in W1 but not in W2, the resulting W1 flux would be underestimated, resulting in a redder $W1-W2$ color. This effect could potentially move objects into the WISE color space spanned by AGNs. In addition, for objects with existing 2MASS photometry, which are largely well-resolved by WISE, colors estimated using apertures scaled from the 2MASS XSC shape value might be artificially blue as these apertures may contain foreground stars. Furthermore, these apertures may not contain all of the infrared emission from the object, leading to a systematic underprediction of the fluxes (see Cluver et al. 2014 for an examination of this effect). Because of these photometric effects, we have carefully visually examined the objects that have red WISE colors indicative of potential AGN activity, although objects with ext_flg = 5 represent a small percentage of the total sample.

3.1. Optically Selected AGNs and Composite Galaxies

We begin our analysis of the IR properties of dwarf galaxies by exploring the IR colors of the AGNs and composite dwarf galaxies selected from Reines et al. (2013). These 136 dwarf galaxies were initially chosen based on their optical spectroscopic properties and specifically their position on the BPT diagram (Baldwin et al. 1981), a common optical emission-line diagnostic diagram that plots the flux ratio of [O iii]λ5007/Hβ against [N ii]λ6583/Hα.⁷ This diagram separates AGNs from star-forming galaxies based on commonly used dividing lines, an empirical line given in Kauffmann et al. (2003), and a theoretical "maximum starburst line" from Kewley et al. (2001). Objects above the Kewley et al. (2001) maximum starburst line on the diagram are often categorized as AGNs, while objects between the Kauffmann et al. (2003) and Kewley et al. (2001) lines are thought to have contributions to their emission-line flux from both star formation and AGN activity, and are known as "composite" objects.

Of the 136 BPT AGNs and composites from Reines et al. (2013), 104 objects have significant WISE detections as described in Section 2 (four objects did not have a match to the AllWISE catalog, and the other 28 objects do not have significant detections in the W1, W2, or W3 bands). We plot these 104 dwarf galaxies on the infrared AGN diagnostic WISE color–color plot in Figure 1, which compares the $W1-W2$ color to the $W2-W3$ color. We color the points based on whether they were classified as an optical AGN (red) or composite galaxy (purple). We also plot common infrared AGN selection criteria with colored lines. In blue, we plot the selection box from Jarrett et al. (2011), and in orange, we plot the $W1-W2\gt 0.8$ color criteria from Stern et al. (2012). These selection boxes select for luminous quasars by the presence of power-law emission that extends from the W1 through W3 bands due to reprocessed emission from the dust torus around the accretion disk feeding the central supermassive black hole. As can be seen from Figure 1, the majority of the dwarf galaxies selected as AGNs and composite objects by their optical emission would not be selected as an IR AGN on this diagram. Only six of the Reines et al. (2013) BPT-AGN dwarf galaxies would be selected as AGN by the Stern et al. (2012) criterion. Five BPT-AGN and one BPT-composite object would be selected as an AGN by the Jarrett et al. (2011) criteria. We plot the SDSS and WISE thumbnail images for these six objects in Figure 2. From this figure, it can be seen that these optically selected AGNs have significant WISE detections out to the longest wavelengths probed by WISE, evidence of an underlying infrared power-law from AGN-heated dust. As discussed in the introduction, more luminous AGNs are more likely to be found with WISE selection, as their infrared emission can be observed above the emission from dust heated by host galaxy star formation. This helps to explain the general trends observed in Figure 1, where five out of six of the objects in the Jarrett et al. (2011) selection box are BPT-AGNs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The majority of both the BPT-AGNs and BPT-composite dwarf galaxies lie below the AGN selection regions in the area occupied by star-forming galaxies. These objects form a sequence that moves from blue $W1-W2$ and $W2-W3$ colors to the right and toward redder $W2-W3$ colors. For these objects, the AGN IR luminosity must be low enough that the reprocessed infrared emission is overwhelmed by the emission from both the galaxy's stellar continuum, the continuum emission from dust heated by star formation throughout host galaxy, and polycyclic aromatic hydrocarbon (PAH) emission. We will explore this in more detail in the next section.

3.2. Optically Selected Star-forming Dwarf Galaxies

We now look at the dwarf galaxies in the NSA with line ratios dominated by star formation. We use a subsample (N = 14,013) of our WISE-selected dwarf galaxies that have significant (S/N > 3.0) SDSS detections of the strong optical emission lines (Hβ, [O iii]λ5007, Hα, and [N ii]λ6583) and fall in the star-forming part of the BPT diagram. We use the wealth of SDSS optical data to understand how dwarf galaxy SF properties relate to WISE infrared colors and explore the potential origin of the red WISE colors observed in some dwarf galaxies. In Figure 3, we show the star-forming subsample on the BPT diagram with points colored by $W1-W2$ infrared color. We plot those dwarf galaxies with $W1-W2\gt 0.8$, the selection criteria used to select AGNs from Stern et al. (2012), using black stars. For these star-forming dwarf galaxies, we can see that, generally, the $W1-W2$ color becomes redder as you move upwards and to the left on the sequence. Star-forming galaxies on the BPT diagram form a sequence in metallicity. Objects with the reddest WISE colors are among the lowest metallicity galaxies in our sample and also have the largest ionization parameters (see Kewley et al. 2013 for a review of how objects move on the star-forming region of the BPT diagram).

Zoom In Zoom Out Reset image size

3.2.1. Physical Properties of Star-forming Dwarf Galaxies

We can use the SDSS spectroscopic data to derive properties of the stellar populations in the dwarf galaxies for comparison to their WISE colors. Some properties, such as galaxy size and optical color, are taken directly from the NSA, but in this section, we outline the methods we use to estimate the host galaxy SFR, stellar age, and ionizing luminosity.

Star Formation Rates. The luminosity of Hα can be used to measure the SFR for a galaxy, following Kennicutt & Evans (2012):

$$\begin{eqnarray}&&\mathrm{log}(\mathrm{SFR}[{M}_{\odot }\,{\mathrm{yr}}^{-1}])=\mathrm{log}(L({\rm{H}}\alpha )\,[\mathrm{erg}\,{{\rm{s}}}^{-1}])-41.27.\end{eqnarray} \tag{ 1 }$$

We have corrected the Hα luminosity for the effects of extinction, using an intrinsic Balmer ratio of ${F}_{{\rm{H}}\alpha }/{F}_{{\rm{H}}\beta }=2.86$, R_V = 3.1, and the O'Donnell (1994) extinction curve. This method of estimating extinction is more valid in regions of low obscuration, and is not reliable for the most obscured regions. We also calculate the sSFR, the star formation rate divided by the stellar mass of a galaxy. Since the Hα emission-line fluxes were measured using 3'' diameter fibers,⁸ we calculated the sSFR for each galaxy using masses estimated from the fiberflux values given in the NSA. These fluxes, which were measured within a 3'' aperture, were converted to stellar mass using the method discussed in Reines & Volonteri (2015), where we used the color-dependent mass-to-light ratio from Zibetti et al. (2009), adopting a solar absolute i-band magnitude of 4.56 mag (Bell et al. 2003). While most of the sources have sSFR < 1.0 Gyr⁻¹, the majority of the red WISE sources have sSFR > 1.3 Gyr⁻¹. In addition, we also use the Petrosian half-light radius from the NSA to estimate the SFR surface density (Σ_SFR), the SFR divided by the area of the galaxy in kpc².

Stellar Age. The equivalent width (EW) of the hydrogen Balmer lines anticorrelates with the age of a stellar population (Stasińska & Leitherer 1996). We estimate stellar ages by comparing measured Hα values to Starburst99 models (Leitherer et al. 1999), assuming a Salpeter (1955) IMF with power-law α = 2.35 and a metallicity of Z = 0.004 (where Z = 0.02 represents solar metallicity). The majority of the galaxies with red WISE colors have high Hα EWs and inferred stellar ages less than 7 Myr.

Ionizing Luminosity. We use the extinction-corrected Hα luminosity to estimate the rate of ionizing photons (Q_Lyc) for these objects following the procedure from Reines et al. (2008):

$$\begin{eqnarray}&&\left(\displaystyle \frac{{Q}_{\mathrm{Lyc}}}{{{\rm{s}}}^{-1}}\right)\geqslant 2.25\times {10}^{12}{\left(\displaystyle \frac{{T}_{e}}{{10}^{4}{\rm{K}}}\right)}^{0.07}\left(\displaystyle \frac{{L}_{{\rm{H}}\alpha }/2.86}{\mathrm{erg}\ {{\rm{s}}}^{-1}}\right).\end{eqnarray} \tag{ 2 }$$

Here, we start from Equations (2) and (3) from Condon (1992), and then assume Case B recombination, where F_Hα/F_Hβ = 2.86. We further assume an electron temperature of 10,000 K (Izotov et al. 1997), and, as seen from the above equation, our estimated ionizing photon rate is only a lower limit due to the fraction of ionizing radiation that is absorbed by dust or escapes the star-forming regions before ionizing hydrogen atoms. The dwarf galaxies in our sample with the reddest $W1-W2$ colors have large Hα luminosities and values of Q_Lyc equivalent to greater than several thousand O-type stars.

3.2.2. Mid-IR Colors of Star-forming Dwarf Galaxies

To investigate the origin of the infrared emission from the star-forming dwarfs, we plot the optical g − r and infrared $W1-W2$ colors for these objects in Figure 4. Here, we can see that the galaxies with the bluest optical colors have the reddest infrared colors. We have colored the points by the EW of the Hα emission line taken from the NSA, and we see that those dwarf galaxies with the reddest WISE colors have the highest Hα EW. We plot WISE W1 − W2 color as a function of Σ_SFR for each galaxy in Figure 5, and we see that the majority of the objects with the reddest W1 − W2 colors also have the highest Σ_SFR, further supporting the idea that we are observing compact young starbursts in these dwarf galaxies with red WISE colors. We note that our usage of aperture fluxes to derive SFRs would lead to values of Σ_SFR for objects with Petrosian half-light radii larger than the SDSS fiber to be underestimated, although this effect does not significantly change the results seen in Figure 5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We show WISE color–color diagrams for the dwarf star-forming galaxies in Figure 6. We color the points based on the EW (left panel) and luminosity (right panel) of the Hα emission line. In both panels of Figure 6, we also plot the same AGN selection boxes shown in Figure 1. We can see that the majority of the objects lie outside the Jarrett et al. (2011) selection box. At the reddest W2 − W3 colors, the star-forming sequence extends upwards to redder W1 − W2 colors, wrapping around the Jarrett et al. (2011) AGN selection box and above the Stern et al. (2012) selection line. In the left panel of Figure 6, we see that those objects with the reddest WISE infrared colors have the largest Hα EW values and youngest stellar ages, also indicated in Figure 4. Similarly, in the right panel of Figure 6, we can see an even stronger trend for the objects with the reddest WISE colors to have the largest Hα luminosities and rate of ionizing photons.

Zoom In Zoom Out Reset image size

We explore the dwarf galaxy sequence on the WISE color–color diagram as a function of dust temperature in Figure 7. Here, on the left, we plot the same dwarf galaxies and AGN selection boxes as in Figure 6, but now the objects are colored by their sSFR. We see that the sequence of BPT-selected star-forming dwarf galaxies drives to the right and upward, and is correlated with the sSFR. We also plot extreme WISE dwarf galaxies from Griffith et al. (2011) and Izotov et al. (2011), which have W1 − W2 ∼ 2.0. These rare objects are at the extreme end of the sequence traced by the star-forming dwarf galaxies in our sample. We also plot the WISE colors for the infrared-bright star-forming galaxies Arp220 and M82, as measured from their SWIRE SED templates (Polletta et al. 2007). Finally, in this panel we plot color tracks that we generated by linearly combining an elliptical galaxy template from Assef et al. (2010) with a simple single-temperature blackbody (BB) at z = 0. For these tracks, we chose an elliptical galaxy template as this template does not show evidence for dust in the infrared. We plot three tracks at three different dust temperatures, 400 K (blue), 330 K (olive), and 300 K (magenta). The position along the track (to the right and upwards) reflects the relative normalization of the BB and the galaxy template, which we parameterize using ${f}_{\mathrm{BB},10\mu {\rm{m}}}$, the fraction of the total flux emitted from the BB at 10 μm. In the right panels we show example SEDs for both the 300 K (top) and 400 K (bottom) examples, with the galaxy template in blue, the BB in red, and the total flux in green. We overplot the W1, W2, and W3 fluxes, and provide the W1 − W2 and W2 − W3 colors in each panel.

Zoom In Zoom Out Reset image size

From the tracks and SEDs in Figure 7, we can see how the sequence of dwarf galaxies arises in WISE color space. At a fixed BB normalization, as the temperature of the hottest dust in a given galaxy increases, galaxies move upward and to the left, satisfying the Stern et al. (2012), and potentially the Jarrett et al. (2011), AGN selection criteria. Alternately, at a fixed BB temperature, as the normalization of the BB template increases, objects move upward and to the right. While a BB is a simplistic approximation of the dust emission observed in the infrared, our model does replicate the observed range in the W1 − W2and W2 − W3 colors for the dwarf galaxies. We emphasize that our model lacks PAH emission, which may significantly change the observed colors, although lower PAH abundance has been found in low-metallicity, high-sSFR dwarf galaxies (Wu et al. 2005; Madden et al. 2006; Engelbracht et al. 2008; Galliano et al. 2008; Rémy-Ruyer et al. 2015). The BB temperatures from our models are consistent with those estimated for dust heated from star formation in dwarf galaxies in the literature (Johnson et al. 2004; Engelbracht et al. 2005; Jackson et al. 2006; Reines et al. 2008). In addition, as the WISE wavelength coverage does not extend to the far-infrared, we cannot quantify cooler dust in these galaxies. We further discuss the implications of these results in Section 4.

3.2.3. Mid-IR-selected AGN Candidates

While our results indicate that the majority of star-forming dwarf galaxies lie outside of the commonly used WISE AGN selection boxes, there are potential IR-selected AGN candidates that we want to highlight. We find only 10 dwarf star-forming galaxies with WISE colors that put them in the Jarrett et al. (2011) AGN selection box. We present SDSS thumbnails for these objects in Figure 8. These objects span a range in optical color, but are on average bluer than the full sample of dwarf galaxies (e.g., see Figure 4). Three out of the ten objects do not have significant (S/N > 3) W4 photometry (NSA ID 93798, 109919, and 117162), as would be expected for an AGN. Two of these objects, NSA ID 109919 and 93798, have low surface brightness, and do not show strong photometric evidence for a central nuclear source. NSA 93978 was presented by Secrest et al. (2015) who demonstrated that this galaxy was detected by XMM-Newton with a luminosity of ${L}_{2-10\mathrm{keV}}=2.4\times {10}^{40}$ erg s⁻¹. We further discuss this object in Section 3.4. For the remainder of the objects, their morphologies are either compact blue nuggets (i.e., 98135), objects with blue cores and extended features (4610, and 118025), or objects with evidence for disks and nuclear cores (6205, 57649, and 151888).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.3. Galaxies Without Optical Emission-line Flux Measurements

There are a number of dwarf galaxies in our sample (4469 objects) with low signal-to-noise ratio (S/N) detections of the emission lines in existing SDSS spectra or no spectroscopic information in the NSA. From this subsample, we initially found 68 objects that fall into the Jarrett et al. (2011) selection box. We examined these objects to explore whether they had any alternate spectroscopic information: many have SDSS DR12 spectra that allowed us to determine that the NSA redshifts (and therefore masses) were incorrect for these objects, and they are likely unobscured quasars at z ∼ 1–2. There are two objects, NSA IDs 110100 and 172349, that appear to be a chance alignment of a dwarf galaxy and a background quasar. After removing any quasar contaminants, we are left with only 25 potential IR-selected AGN dwarf galaxy candidates, and we show their SDSS and WISE thumbnails in Figures 10–12 in the Appendix. We note that two of these objects, NSA IDs 135305 and 150927, have poor W3 photometric fits (for these objects, the WISE w3rchi2 value, the χ² of the W3 profile fitting, is W3rchi2 ≥ 2), and their W3 flux might be underpredicted, leading to these objects being bluer in the W2 − W3 color. Many of the dwarf galaxies do not have significant W4 detections. We would expect W4 detections from the power-law emission typically associated with an AGN, as seen for the BPT-AGNs in Figure 2. While the sizes and g − r colors for these objects span a similar range as what was measured for the full sample of BPT star-forming objects in Figures 4 and 5, many are either optically redder or physically larger than the star-forming dwarf galaxies with red W1 − W2 colors. Optical spectroscopy is necessary to better understand these systems.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We list the NSA ID, SDSS ID, redshift, optical color, and WISE colors for all of the AGN candidates (using the Jarrett et al. 2011 selection box) in Table 1. Furthermore, we plot all the NSA dwarf galaxies that have WISE colors placing them inside the Jarrett et al. (2011) selection box in Figure 9, with the points colored by their optical properties. We note that these objects do not span the entire Jarrett et al. (2011) selection space, but cluster near the right and bottom of the box.

Table 1.  WISE-selected/Jarrett et al. (2011) AGN Candidates

NSA ID	SDSS ID	z	g − r	W1	W1 − W2	W2 − W3
BPT-selected AGNsa
68765	J032224.64+401119.8	0.026	0.618	14.012 ± 0.020	0.842	3.674
10779	J090613.75+561015.5	0.047	0.426	13.400 ± 0.024	1.301	3.659
125318	J095418.15+471725.1	0.033	0.448	13.859 ± 0.027	1.058	3.971
113566	J114359.58+244251.7	0.050	0.683	14.275 ± 0.028	0.915	2.788
104527	J133245.62+263449.3	0.047	0.281	13.514 ± 0.025	1.033	3.721
BPT-selected Composite Galaxiesa
79874	J152637.36+065941.6	0.0384	0.299	14.121 ± 0.026	0.756	2.901
BPT-selected Star-forming Galaxiesb
6205	J005904.10+010004.2	0.018	0.240	13.308 ± 0.024	0.932	3.891
109919	J085233.75+135028.3	0.005	0.305	15.761 ± 0.052	0.779	3.086
117162	J102345.04+144604.7	0.047	0.218	14.794 ± 0.033	0.683	2.944
118523	J122822.81+202043.6	0.049	−0.053	16.432 ± 0.074	0.918	3.747
118025	J131253.76+171231.1	0.052	−0.260	16.583 ± 0.086	0.958	3.722
93798	J132932.42+323416.9	0.016	0.252	15.297 ± 0.036	0.727	2.586
98135	J154748.99+220303.2	0.031	−0.390	15.583 ± 0.041	1.388	4.177
57649	J160135.95+311353.7	0.031	0.136	14.526 ± 0.028	1.243	3.543
4610	J173501.25+570308.8	0.047	−1.003	13.041 ± 0.029	1.267	4.051
151888	J233244.60−005847.9	0.024	0.382	13.581 ± 0.011	1.119	3.721
Galaxies Without Optical Emission-Line Fluxesc
127223	J003441.39−212928.9	0.023	0.027	15.975 ± 0.055	0.793	3.901
127548	J004206.29−093335.0	0.054	0.515	15.659 ± 0.049	1.189	4.078
128988	J011600.59+063812.9	0.008	0.395	14.991 ± 0.035	0.737	2.919
130346	J014554.39+241555.0	0.035	−0.467	16.165 ± 0.056	1.107	4.098
130679	J015446.09+014326.0	0.039	0.208	16.544 ± 0.074	0.813	3.485
170240	J023850.49+272159.0	0.005	0.062	14.688 ± 0.030	0.712	2.550
64260	J075824.77+202352.3	0.035	−0.631	16.317 ± 0.074	1.519	3.894
135305	J083538.40−011407.0	0.044	0.812	12.038 ± 0.006	1.015	3.515
10311	J084624.06+540915.9	0.031	0.434	15.606 ± 0.042	0.951	3.219
135720	J091108.59+541051.9	0.027	−0.596	15.816 ± 0.042	0.768	3.852
135754	J091305.00+284845.0	0.027	−0.285	16.998 ± 0.119	1.079	4.042
136250	J095643.10−030409.0	0.047	0.700	14.258 ± 0.028	0.670	2.857
138274	J105301.99+365811.9	0.026	−0.268	16.522 ± 0.077	0.852	3.956
141379	J122655.39+424208.0	0.009	1.058	14.945 ± 0.032	0.858	4.197
171667	J123948.06−171753.9	0.028	0.788	13.990 ± 0.026	0.867	4.112
142435	J125037.90+434519.9	0.039	0.418	15.316 ± 0.036	1.464	3.222
142555	J125615.79−032723.0	0.045	0.146	16.044 ± 0.054	0.773	3.421
143373	J132119.70+320824.9	0.018	−0.019	16.823 ± 0.088	1.068	3.845
144502	J140609.20−032224.0	0.055	0.267	15.951 ± 0.050	0.917	3.547
53934d	J155320.20+420735.6	0.022	0.443	15.757 ± 0.037	1.092	3.097
171765	J204609.92−160058.2	0.033	0.240	14.964 ± 0.037	1.123	3.162
44938d	J211012.75+002428.2	0.020	0.638	16.368 ± 0.076	1.446	3.154
150295	J225515.30+180835.0	0.043	−0.192	13.158 ± 0.025	1.368	3.947
150927	J231544.60+065439.0	0.008	0.043	12.430 ± 0.007	0.791	3.704
152731	J235005.99+163505.9	0.049	−0.575	16.725 ± 0.099	1.062	3.601

Notes.

^aObjects from Reines et al. (2013). The SDSS and WISE thumbnails for these objects are shown in Figure 2. ^bThe SDSS and WISE thumbnails for these objects are shown in Figure 8. ^cThe SDSS and WISE thumbnails for these objects are shown in Figures 10–12 in the Appendix. ^dThese objects have SDSS spectra, but one or more of the emission lines used for BPT selection lack sufficient S/N for our optical classification.

Download table as:  ASCIITypeset image

3.4. Comparison to Recent Searches for AGNs in Dwarf Galaxies Using Mid-IR Colors

3.5. The Feasibility of Detecting Supermassive Black Holes in Dwarf Galaxies with WISE

To better understand the birth and growth of supermassive black holes, it is fundamentally important to find evidence for AGN activity in low-mass dwarf galaxies. Here, we looked at the mid-IR properties of a sample of dwarf galaxies at z < 0.055 to explore the use of WISE colors to select for AGN activity. We used the most up-to-date AllWISE photometry and individually inspected each candidate AGN to select a sample that is significantly smaller than previous samples of IR-selected AGNs in dwarf galaxies.

Our main conclusions are:

• 1.  
  The majority of optically selected AGNs in dwarf galaxies have IR colors that are dominated by their host galaxies.
• 2.  
  Dwarf galaxies with the reddest WISE colors are compact, blue galaxies with young stellar ages and high sSFRs. Dwarf galaxies with extreme star formation are capable of heating dust to temperatures producing W1 − W2 > 0.8, (e.g., Stern et al. 2012), and this single color cut alone should not be used to select AGNs in dwarf galaxies.
• 3.  
  We provide a sample of 41 dwarf galaxies in the NSA that have WISE colors in the Jarrett et al. (2011) AGN selection box, 6 of which have optical spectroscopic evidence for an AGN (Reines et al. 2013). While the majority of the objects in our sample have been included in previous samples of WISE-selected dwarf galaxy AGN candidates, our sample is much smaller due to the updated WISE photometry, more conservative selection criteria and S/N thresholds, and the removal of spurious candidates. We caution that follow-up observations are necessary to confirm the presence of active massive black holes in the other 35 objects.

From our analysis, optically blue, high-sSFR dwarfs with young starbursts and associated high ionizing fluxes can have red mid-IR colors that could be mistaken for AGN activity, particularly if using a simple W1 − W2 color cut. While these objects have a population of young stars that can heat dust to temperatures that result in red W1 − W2 colors (W1 − W2 > 0.5 − 0.6), they primarily have very red W2 − W3 colors (W2 − W3 > 4.2), which is less extreme, but similar to the W2 − W3 colors seen for the dwarf star-forming galaxies from Izotov et al. (2011).

AGNs, which can heat dust to even higher temperatures, are predominantly found in a different region of WISE color–color space, and we only find 10 optically selected star-forming dwarf galaxies that would be classified as an AGN by the Jarrett et al. (2011) selection criteria. Of those objects, only 5 have strong W3 and W4 fluxes as would be expected for an AGN. In addition, these objects are consistent with the overall trend of star-forming galaxies in WISE color space, and these objects could represent an extreme star-forming population that has scattered into the Jarrett et al. (2011) selection box. We also found 25 additional dwarf galaxies that fell into the Jarrett et al. (2011) box that do not have optical spectroscopy, or their SDSS emission lines were not strong enough to classify them on the BPT diagram. We include them as candidates, but we caution against using the mid-IR colors of these objects alone to classify them as AGNs; follow-up observations at optical or X-ray wavelengths would be helpful to understand the nature of their IR emission.

Overall, it is important to use both W1 − W2 and W2 − W3 colors when selecting candidate AGNs in dwarf galaxies, as any selection of AGNs in dwarf galaxies that only uses W1 − W2 color (for instance, the selection method of Stern et al. 2012, which has been demonstrated to be reliable at higher masses) will include a large amount of contamination from dwarf star-forming galaxies. Starting with the BPT star-forming galaxies discussed in Section 3.2, we find that 42 dwarf galaxies have W1 − W2 > 0.8 (compared to only 10 in the Jarrett et al. 2011 selection box), while 183 objects have W1 − W2 > 0.5. This contamination from star-forming dwarf galaxies is likely the cause of the observed rise in the fraction of AGN candidates at lower galaxy masses in Satyapal et al. (2014) and Sartori et al. (2015). While we certainly cannot rule out the presence of AGNs in the optically selected star-forming dwarf galaxies with red WISE colors, the systematic correlations between star formation properties and infrared colors lead us to conclude that the infrared emission is unlikely to be powered by AGNs. Furthermore, the majority of known optically selected AGNs in dwarf galaxies do not dominate the WISE colors.

Our results are consistent with evidence in the literature that the ionizing UV radiation from young stars is one of the primary sources of dust heating in low-metallicity dwarf galaxies (Izotov et al. 2014). Both IRAS and Spitzer observations demonstrated that dwarf galaxies have evidence for large quantities of hot dust (Helou 1986; Hunter et al. 1989; Melisse & Israel 1994; Cannon et al. 2006; Rosenberg et al. 2006). In addition, it has been shown that the IR SED peak of dwarf galaxies is broader, which is often explained as resulting from dust at higher temperatures than what is observed in more massive galaxies (Boselli et al. 2012; Smith et al. 2012; Rémy-Ruyer et al. 2013, 2015; Ciesla et al. 2014). The temperatures from the tracks in Figure 7 for the hot dust that would be necessary to produce the observed WISE colors has been observed in local low-metallicity galaxies from Spitzer 8 μm observations (Engelbracht et al. 2005; Jackson et al. 2006), and even hotter dust has been invoked to explain the near-IR excesses observed in the dwarf galaxies SBS 0335-052 (Reines et al. 2008) and Haro 3 (Johnson et al. 2004). Recently, in an analysis of the infrared properties of a large sample of low-metallicity dwarf galaxies, Rémy-Ruyer et al. (2015) demonstrated that the dust SED for these galaxies peaks at shorter wavelengths as compared to higher-metallicity systems, which they attribute to a clumpy interstellar medium that allows for a wider range of dust temperatures. In addition, the lower-metallicity dust will attenuate the light from young stars less, and the dust can then be heated deeper within individual molecular clouds. Cormier et al. (2015) used Herschel PACS spectroscopy of low-metallicity dwarf galaxies to provide evidence that the interstellar medium in these galaxies is more porous than in metal-rich galaxies, leading to a larger fraction of the stellar UV radiation heating dust. These results are supported by modeling by Hirashita & Hunt (2004), who found that dust temperature and dust luminosity is higher in dense, compact, low-metallicity star-forming regions.

Support for A.E.R. was provided by NASA through Hubble Fellowship grant HST-HF2-51347.001-A awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., for NASA, under contract NAS 5-26555. The work of D.S. was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA. We are grateful to the entire SDSS collaboration for providing the data that made this work possible, and to Michael Blanton and all those involved in creating the NASA-Sloan Atlas. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, and NEOWISE, which is a project of the Jet Propulsion Laboratory/California Institute of Technology. WISE and NEOWISE are funded by the National Aeronautics and Space Administration.

The SDSS and WISE thumbnails for the 25 potential IR-selected AGN dwarf galaxy candidates without optical emission-line flux measurements are shown in Figures 10–12.