© ESO, 2011

1. Introduction

Nearby star-forming emission-line galaxies play an important role for our understanding of star-formation processes in low-metallicity environments, and they can be considered as local counterparts or “analogs” of high-redshift star-forming Lyman-break galaxies (LBGs). Recently, Heckman et al. (2005) identified nearby (z < 0.3) ultraviolet-luminous galaxies (UVLGs) selected from the Galaxy Evolution Explorer (GALEX). Eventually, compact UVLGs were called Lyman-break analogs (LBAs). They resemble LBGs in several respects. In particular, their metallicities are subsolar, and their star-formation rates (SFRs) of  ~4–25 M_⊙ yr^-1 are overlapping with those for LBGs. Recently, Cardamone et al. (2009) selected a sample of 251 compact strongly star-forming galaxies at z ~ 0.112–0.36 on the basis of their intense green colour on the Sloan Digital Sky Survey (SDSS) images (“green pea” galaxies), which again are similar to LBGs owing to their low metallicity and high SFRs. Izotov et al. (2011) extracted a sample of 803 star-forming luminous compact galaxies (LCGs) with hydrogen Hβ luminosities L(Hβ)  ≥  3 × 10⁴⁰ erg s^-1 and Hβ equivalent widths EW(Hβ)  ≥  50 Å from SDSS spectroscopic data. These galaxies have properties similar to “green pea” galaxies but are distributed over a wider range of redshifts z ~ 0.02–0.63. The SFRs of LCGs are high  ~0.7–60 M_⊙ yr^-1 and overlap with those of LBGs. Izotov et al. (2011, see also Guseva et al. 2009 showed that LBGs, LCGs, luminous metal-poor star-forming galaxies (Hoyos et al. 2005), extremely metal-poor emission-line galaxies at z < 1 (Kakazu et al. 2007), and low-redshift blue compact dwarf (BCD) galaxies with strong star-formation activity obey a common luminosity-metallicity relation. Therefore, it is important to study nearby star-forming galaxies over a wide range of luminosities and metallicities to shed light on physical conditions and star-formation history in high-redshift galaxies even though most metal-deficient and low-luminosity high-redshift galaxies are still awaiting their detection.

Fig. 1 a) Dependence of the W1 − W2 colour on the oxygen abundance 12+logO/H for a sample of  ~1000 galaxies. Galaxies with Hβ luminosity L(Hβ)  ≥  3 × 1040 erg s-1, corresponding to SFR(Hβ)  ≥  0.7 M⊙ yr-1, are shown by red filled circles while galaxies with L(Hβ)  <  3 × 1040 erg s-1 are shown by blue open circles. The NW and SE components of the low-metallicity BCD I Zw 18 are labelled and connected with a solid line. Newly identified galaxies with W1 − W2  >  2 mag are shown by large red filled circles, the three galaxies discussed by Griffith et al. (2011) are shown by large red filled squares. Typical error bars are shown in the lower right corner. b) Histograms of W1 − W2 distributions for galaxies with L(Hβ)  ≥  3 × 1040 erg s-1 (red dotted line) and with L(Hβ)  <  3 × 1040 erg s-1 (blue solid line). c) and d) same as a) and b), respectively, but the sample is split into galaxies with EW(Hβ)  ≥  50 Å (red symbols and lines) and EW(Hβ)  <  50 Å (blue symbols and lines).

The Infrared Space Observatory (ISO), Spitzer, and most recently the Wide-field Infrared Survey Explorer (WISE, Wright et al. 2010) open up the opportunity to probe properties of star-forming galaxies in the mid-infrared range (MIR)  ~  3.5–24 μm, the range of warm and hot dust. The WISE mission has an advantage because it is directed to produce an all-sky photometric survey at wavelengths 3.4 μm (W1), 4.6 μm (W2), 12 μm (W3) and 22 μm (W4) with a sensitivity at  ~12–24 μm that is  ~1000 times higher than that of the InfraRed Astronomical Satellite (IRAS) and has an angular resolution of  ~6′′ at 3.4 μm.

Thuan et al. (1999) first showed (from ISO spectroscopy) that one of the most metal-deficient BCDs, SBS 0335–052E (e.g. Izotov et al. 1990), is extraordinarily bright in the MIR range, implying a large amount of warm (~100–300 K) dust. Later, Houck et al. (2004) based on Spitzer spectra have confirmed the presence of warm dust in SBS 0335–052E and found that the dust emission peaks at a wavelength of  ~28 μm, much shorter than that for the bulk of star-forming galaxies. Ground-based infrared spectroscopy of SBS 0335–052E by Hunt et al. (2001) revealed that the continuum at shorter wavelengths in the range 3.4–4 μm is rising in the direction of longer wavelengths. Later, Griffith et al. (2011) found that the WISE 3.4–4.6 μm (W1 − W2) colour of SBS 0335–052E is very red,  > 2 mag. This longward rising of MIR emission implies the presence of hot (up to 1000 K) dust emission.

The properties of SBS 0335–052E in the MIR range are very different from that in another extremely metal-deficient BCD with similar metallicity, I Zw 18. Hirashita & Hunt (2004) attributed these differences to different modes of star formation, an “active” mode in very compact regions of SBS 0335–052E and a “passive” mode in the more diffuse interstellar medium of I Zw 18.

Griffith et al. (2011) argued that WISE can be an efficient tool in searching for other star-forming galaxies with appreciable hot dust emission and demonstrated the truth of this statement by finding of three low-metallicity BCDs with W1 − W2 > 2 mag. However, they noted that these galaxies are rare. In the present paper we attempt to find new star-forming galaxies selected from the Data Release 7 (DR7) of the SDSS with red WISE W1 − W2 colours in the Preliminary Release Source Catalogue (PRSC), which covers 57% of the sky.

Fig. 2 a) Dependence of W1 − W2 colour on redshift z for the sample of  ~ 1000 galaxies. b) (W1 − W2) vs. (W2 − W3) colour–colour diagram. Symbols are the same as in Fig. 1a.

2. Selection criteria

We used the whole spectroscopic data base of the SDSS DR7 (Abazajian et al. 2009), which comprises  ~900 000 spectra of galaxies to select a sample of  ~16 000 spectra with strong emission lines, excluding those that show evidence for AGN activity. Thus, our sample includes the majority of SDSS star-forming galaxy spectra with EW(Hβ)  ≥  10 Å. A small fraction of these spectra (~2%) represents individual H ii regions in nearby spirals but the overwhelming majority is composed of integrated spectra of galaxies from farther distances. The coordinates of SDSS selected galaxies are used to identify sources in the WISE PRSC within a circular aperture of 10′′ in radius. In total,  ~5000 WISE sources were identified with SDSS galaxies out of the sample of  ~16 000 galaxies. Nearly 1000 galaxies out of  ~5000 galaxies detected by WISE have sufficiently strong [O iii] λ4363 emission lines in their SDSS spectra with a line flux error not exceeding 50%, allowing for a reliable oxygen abundance determination. We analyse the MIR properties of this sample of  ~1000 galaxies in Sect. 3.

3. Results

Several galaxy components can contribute to the emission in the 3.4–4.6 μm range: stars, ionised gas, polycyclic aromatic hydrocarbon (PAH) emission, and hot dust. Generally, the PAH emission is weak in a low-metallicity environment (Engelbracht et al. 2008; Hunt et al. 2010). Stellar (with an effective temperature  ≥ 3000 K) and interstellar ionised gas emission is characterised by W1 − W2 colours of  ~0–0.4 mag. In particular, the W1 − W2 ~ 0.5 mag of I Zw 18, one of the most metal-deficient BCDs, is consistent with that of stellar and ionised gas emission only. On the other hand, a colour excess above W1 − W2 ~ 0.4 mag could be an indication of hot dust with a temperature of several hundred Kelvin. In particular, W1 − W2 colours of 3, 2, and 1 mags correspond to black body temperatures of 350, 500, and 900 K, respectively. This colour excess also depends on the relative contribution of hot dust emission to the total emission. Therefore, while it clearly points at hot dust in extreme cases, the W1 − W2 colour alone is not sufficient for a precise determination of the dust temperature. The spectral energy distribution is needed in a wide wavelength range to fit and subtract stellar and gaseous emission.

Fig. 3 20′′ × 20′′ SDSS images of galaxies with W1 − W2  >  2 mag.

Fig. 4 Observed spectral energy distributions (SED) of galaxies with W1 − W2  >  2 mag, which include SDSS optical spectra (black solid lines) and WISE MIR photometric data in all four bands (red symbols and solid lines). For comparison in all panels are shown MIR SEDs for the “active” BCD SBS 0335–052E (red dotted lines) and for the “passive” BCD I Zw 18 (blue dashed lines).

In Fig. 1a we show the dependence of W1 − W2 on the oxygen abundance 12+logO/H for the sample of  ~1000 galaxies with the best determined oxygen abundances (see Sect. 2). The sample is split into two subsamples of objects with the Hβ luminosity L(Hβ)  ≥  3 × 10⁴⁰ erg s^-1 (small red filled circles) corresponding to a star-formation rate SFR  ≥ 0.7   M_⊙ yr^-1 (according to prescriptions of Kennicutt 1998) and L(Hβ)  <  3 × 10⁴⁰ erg s^-1 (small blue open circles). Figure 1b shows the histogram of colour distributions for both samples. It is seen from Figs. 1a,b that a major fraction of galaxies has W1 − W2  ≤  0.5 mag, implying that emission in these galaxies is dominated by stars and ionised gas. On the other hand, there are 65 galaxies with W1 − W2 ~ 1–2 mag where the contribution of hot dust emission is significant. Most of these galaxies are characterised by high SFRs  ≥  0.7 M_⊙ yr^-1 and can therefore be classified as LCGs (Izotov et al. 2011) or “active” galaxies following the nomenclature of Hirashita & Hunt (2004). No colour dependence on metallicity is present, at variance with the conclusion made by Griffith et al. (2011) on the base of a smaller BCD sample.

In Fig. 1c,d the same sample is split into sources with low and high Hβ equivalent widths EW(Hβ), which is a measure of star formation burst age (high values relate to very recent bursts). Clearly, red W1 − W2 colours are observed mainly in galaxies with very young (<3–4 Myr) starbursts.

Figures 2a,b present the dependence of W1 − W2 colours on the redshift z and the (W1 − W2)–(W2 − W3) colour–colour diagram, respectively, for the same sample as in Fig. 1 split into two parts with L(Hβ)  ≥  3 × 10⁴⁰ erg s^-1 (red symbols) and L(Hβ)  <  3 × 10⁴⁰ erg s^-1 (blue symbols). No evident trend of the W1 − W2 colour on z is present for galaxies with W1 − W2    ≥ 1 mag, while for galaxies with bluer W1 − W2 colours there is a tendency to be redder with increasing z. The colour–colour diagram (Fig. 2b) shows that galaxies with low SFRs (blue symbols) occupy the region of spiral galaxies, according to Lake et al. (2011). On the other hand, galaxies with high SFRs (red symbols) are mixed with QSOs, Seyferts, ultraluminous infrared galaxies (ULIRGs), LINERs and starbursts. Physical properties of our galaxies are very different from those of other types of galaxies, excluding starbursts. Evidently, the colour–colour diagram in Fig. 2b fails to separate galaxies of different types.

Coming back to Figs. 1a,c, it can also be seen that galaxies with very red colours W1 − W2  >  2 mag are rare. Griffith et al. (2011) found three galaxies (red filled squares) that are not present in our sample, because no SDSS spectra are available for them. We find only four more galaxies like these (large red filled circles) out of  ~ 5000 galaxies from our SDSS sample with the available WISE data. In the colour–colour diagram our four objects together with three galaxies from Griffith et al. (2011); (large symbols in Fig. 2b) are located in the region of ULIRGs, LINERs and obscured AGNs, but their other properties are very different. General characteristics of the newly identified four galaxies are shown in Table 1, SDSS images and observed spectral energy distributions (SEDs) are shown in Figs. 3 and 4. All four galaxies are very compact (~1–2′′ in diameter, corresponding to linear scales  ~ 3–6 kpc at their redshifts of 0.15–0.33) and are almost unresolved. The SDSS spectra resemble those with strong emission lines, which is suggestive of active star-formation in young bursts with rates SFR ~ 6−27   M_⊙ yr^-1. Their MIR SEDs (red symbols and solid lines in Fig. 4) indicate a flux excess similar to that in the “active” BCD SBS 0335 − 052E (red dotted lines) in contrast to no flux excess in the “passive” BCD I Zw 18 (blue dashed lines). Their low stellar masses of  ~10⁸–10⁹   M_⊙ (Table 1) derived from the fitting of SDSS spectra are characteristic of dwarf galaxies according to Izotov et al. (2011).

Certainly, these galaxies can be classified as “green pea” galaxies (Cardamone et al. 2009) or LCGs (Izotov et al. 2011). The characteristics of the newly identified four galaxies with very red W1 − W2 colours are similar to those in high-redshift LBGs. On the other hand, oxygen abundances in these galaxies are higher than in the three BCDs discussed by Griffith et al. (2011), indicating that there is no apparent dependence on metallicity. In the sample of 803 LCGs by Izotov et al. (2011) there are many other galaxies with optical characteristics similar to the LCGs in Table 1 but with more moderate characteristics in the MIR range. Probably, high Hβ luminosity or young age of the burst is not the only factor for heating the dust to high temperatures. Other factors may play a role, such as the morphology and compactness of the galaxy, the presence of super-star clusters (SSCs), distribution of the interstellar gas and dust in the vicinity of young clusters. Therefore, Hubble Space Telescope (HST) high-angular resolution imaging, spectroscopic observations in a wide wavelength range including the MIR range and interferometric observations in the H i λ21 cm line and in the radio-continuum will be useful for a better understanding of the origin of hot dust and the determination of its properties.

4. Summary

We carried out a search for star-forming Sloan Digital Sky Survey (SDSS) galaxies with strong emission lines in the Wide-field Infrared Survey Explorer (WISE) Preliminary Release Source Catalogue (PRSC) aiming to find galaxies with hot dust emission at wavelengths λ3.4–4.6 μm (W1 and W2 bands). We find that  ~5000 galaxies out of the total sample of  ~16 000 SDSS galaxies are present in the PRSC, which covers only  ~57% of the sky. About 1000 galaxies out of the galaxies detected with WISE have a sufficiently strong [O iii] λ4363 emission line in their SDSS spectra to allow a reliable abundance determination. The comparison of optical and mid-infrared properties for the sample of galaxies with reliably derived oxygen abundances led us to following conclusions:

A major fraction of galaxies has W1 − W2 of  ~0.0–0.4 mag, consistent with the colour for the emission from stars and the ionised interstellar medium. The contribution of hot dust emission is small. On the other hand, there are 65 galaxies with redder W1 − W2 colours,  ~1–2 mag. Most of these galaxies are luminous compact galaxies (LCGs) with star-formation rates SFR > 0.7 M_⊙ yr^-1 and high Hβ equivalent widths EW(Hβ)  >  50 Å, implying a very recent starburst, that can efficiently heat interstellar dust to high temperatures.

Star-forming galaxies with very red W1 − W2 colours  > 2 mag are rare. In addition to three galaxies previously studied by Griffith et al. (2011), which are not present in our sample because of the lack of SDSS spectra for them, we find only four galaxies like these from the sample of  ~5000 galaxies detected with WISE.

Acknowledgments

We thank the referee S. Bianchi for valuable comments. Y.I.I., N.G.G. and K.J.F. are grateful to the staff of the Max Planck Institute for Radioastronomy (MPIfR) for their warm hospitality. Y.I.I. and N.G.G. acknowledge a financial support of the MPIfR. 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, funded by the National Aeronautics and Space Administration. Funding for the Sloan Digital Sky Survey (SDSS) and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, and the Max Planck Society, and the Higher Education Funding Council for England.

References

All Tables

All Figures

Fig. 1 a) Dependence of the W1 − W2 colour on the oxygen abundance 12+logO/H for a sample of  ~1000 galaxies. Galaxies with Hβ luminosity L(Hβ)  ≥  3 × 1040 erg s-1, corresponding to SFR(Hβ)  ≥  0.7 M⊙ yr-1, are shown by red filled circles while galaxies with L(Hβ)  <  3 × 1040 erg s-1 are shown by blue open circles. The NW and SE components of the low-metallicity BCD I Zw 18 are labelled and connected with a solid line. Newly identified galaxies with W1 − W2  >  2 mag are shown by large red filled circles, the three galaxies discussed by Griffith et al. (2011) are shown by large red filled squares. Typical error bars are shown in the lower right corner. b) Histograms of W1 − W2 distributions for galaxies with L(Hβ)  ≥  3 × 1040 erg s-1 (red dotted line) and with L(Hβ)  <  3 × 1040 erg s-1 (blue solid line). c) and d) same as a) and b), respectively, but the sample is split into galaxies with EW(Hβ)  ≥  50 Å (red symbols and lines) and EW(Hβ)  <  50 Å (blue symbols and lines).

In the text

	Fig. 2 a) Dependence of W1 − W2 colour on redshift z for the sample of  ~ 1000 galaxies. b) (W1 − W2) vs. (W2 − W3) colour–colour diagram. Symbols are the same as in Fig. 1a.
In the text

	Fig. 3 20′′ × 20′′ SDSS images of galaxies with W1 − W2  >  2 mag.
In the text

	Fig. 4 Observed spectral energy distributions (SED) of galaxies with W1 − W2  >  2 mag, which include SDSS optical spectra (black solid lines) and WISE MIR photometric data in all four bands (red symbols and solid lines). For comparison in all panels are shown MIR SEDs for the “active” BCD SBS 0335–052E (red dotted lines) and for the “passive” BCD I Zw 18 (blue dashed lines).
In the text