© ESO, 2014

1. Introduction

Extragalactic OH maser emission has been observed for 40 years or so, since the first detection in NGC 253 (λ ~ 18 cm, Whiteoak & Gardner 1974). Emission has been reported from 119 galaxies so far (e.g., Darling & Giovanelli 2002; Chen et al. 2007; Fernandez et al. 2010; Willett 2012). About 90% of all published OH maser sources (106/119) have an isotropic luminosity larger than 10 L_⊙, which is million times more luminous than typical Galactic OH masers (hereafter taken as megamasers). The OH megamaser (OHM) emission was mostly detected in the main line of 1667 MHz, and the main line of 1665 MHz was normally weak or absent. The satellite transitions at 1612 MHz and 1720 MHz were only detected in a few nearby galaxies (e.g., McBride & Heiles 2013) and one gravitational lens source, PMN J0134-0931 (z ~ 0.765, Kanekar et al. 2005). The spectral lines of OHM emission are normally broad, with individual components broader than 10 km s^-1 and total linewidths of ~100−1000 km s^-1 (e.g., Lockett & Elitzur 2008). These properties of OHM are different from those of Galactic OH masers, which have a linewidth that are typically narrower than 1 km s^-1 and an emission at 1665 MHz stronger than 1667 MHz. The differences of maser line properties should reflect differences in the environment in which the masing occurs and in the mechanism by which the maser inversion is produced (McBride et al. 2013).

The hosts of OHMs are almost all luminous infrared galaxies (LIRGs, L_FIR> 10¹¹L_⊙, 102/106), where about one third of them are ultra luminous infrared galaxies (ULIRGs, L_FIR> 10¹²L_⊙, 35/106). The OHM detection rate increases with the far infrared (FIR) luminosity of the maser host galaxy, up to $\frac{\mathrm{1}}{\mathrm{3}}$ for ULIRGs (Darling & Giovanelli 2002; Baan 1991). The high IR luminosity galaxies mostly show signs of interaction or merging from their optical imaging observations (Clements et al. 1996). Thus OHMs are believed to be related to galaxy interaction or merging. However, the majority of (U)LIRGs (~80%) was not detected with OHM emission at all (e.g., Staveley-Smith et al. 1992; Bann et al. 1992; Darling & Giovanelli 2002). Do OHM (U)LIRGs represent some kind of distinct population? Do they have some kind intrinsic properties? Comparisons of multi-band properties in detail between samples of OHM (U)LIRGs and non-OHM (U)LIRGs are, thus, important and helpful to answer above questions. Many related works have been done, and no significant difference could be found between OHM and non-OHM (U)LIRGs in the radio, optical, and X-ray regimes (e.g., Lonsdale et al. 1998; Baan et al. 1998; Vignali et al. 2005; Kandalian 1996; Darling & Giovanelli 2002). Since OHMs in (U)LIRGs were generally believed to be pumped by infrared radiation (e.g., Henkel & Wilson 1990; Randell et al. 1995; Lockett & Elitzur 2008; Willett et al. 2011b), more related studies thus focused on the infrared properties of OHM hosts, which is important for understanding the pumping mechanism of megamaser and its physical nature. Existing infrared studies to date are mainly based on IRAS photometry data, Spitzer spectra and photometry observations. Studies of IRAS photometry data show that OHM (U)LIRGs tend to have FIR color excess (at 25μm and 60μm, Henkel et al. 1986; Darling & Giovanelli 2002) and steep FIR spectral indices in the 25−60 μm region (Chen et al. 2007). Willett et al. (2011a, b) performed a comparison analysis of Spitzer middle infrared (MIR) spectra and photometry for 51 OHM and 15 non-OHM ULIRGs. With respect to non-OHM ULIRGs, OHM ULIRGs have warmer dust, a steeper continuum from 15 μm to 35 μm, and deeper silicate absorption associated with a smooth, thick dust shell surrounding the nucleus.

The Wide-field Infrared Survey Explorer (WISE) is the most sensitive infrared satellite to date, with a sensitivity more than one hundred times better than IRAS in the 12 μm band (Wright et al. 2010). And the Galactic absorption is negligible at all WISE bands, with respect to near infrared observations (e.g., J − H − K bands by 2MASS, Massaro et al. 2012). The ULIRGs is one of its main science goals and WISE will be sensitive enough to detect the most luminous ULIRGs out to a lookback time of when the Universe was only 3 billion years old. The WISE all-sky mapping data at 3.4, 4.6, 12, and 22μm (W1, W2, W3, and W4, with an angular resolution of 6.1′′, 6.4′′, 6.5′′ and 12.0′′, respectively) has been released (e.g., Wright et al. 2010). Here, WISE data are searched and analyzed for all published OHM host (U)LIRGs to investigate their mid-infrared properties. For comparison, a sample of non-OHM (U)LIRGs is also compiled.

Both samples with their WISE data are presented in Sect. 2. In Sect. 3, we analyze their MIR properties, including distributions of luminosities at four bands, mutual spectral indices, color−color properties etc. Furthermore, possible connection between OHM emission and MIR emission of maser hosts are explored. Section 4 summarizes our main results.

2. Sample and data

All OH maser galaxies published so far were cross-correlated to the All-Sky data release¹ using the default search radius of 10′′. The retrieved WISE sources were found for all maser sources except IRAS 11257+5850 and IRAS 20550+1655. The position difference (in arcseconds) was checked from the source position in the 2MASS PSC reference frame to the WISE cataloged position of this source. The position difference of all cross-correlated sources is mostly less than 1′′ (Δ r in Table 4, 103/117), and over 96% (113/117) are less than 2′′, which ensure the reliability of our cross-identification. For all WISE detections of cross-correlated sources, the signal-to-noise ratio in all four WISE bands is mostly larger than 10 (i.e., the flux quality indicator ph_qual = A). Upper limits are listed for three sources at the W3 band (IRAS 14586+1432, 17161+2006, and 21077+3358), and four sources at the W4 band (IRAS 09513+1430, 17161+2006, 14586+1432, and 21077+3358), with its parameter Sigmpro (instrumental profile-fit photometry flux uncertainty in mag units) value of null. Table 4 lists all OH maser galaxies (about 10% are kilomasers in italics, i.e., L_OH < 10 L_⊙) with their WISE magnitudes and uncertainties at four bands (W1, W2, W3, and W4).

For the control sample, we chose all non-detection sources among the updated Arecibo survey sample (Darling & Giovanelli 2002). The survey sample was selected from the flux-limited IRAS catalog (f_{60 um} > 0.6 Jy, Saunders et al. 2000) and successfully detected 52 OHMs out of the selected 311 IRAS luminous galaxies (with a redshift range of 0.1−0.3). Similar to OHMs, the WISE counterparts of those 259 non-OHM (U)LIRGs were also checked (with the default radius of 10′′). Among them, there are 223 sources (86%) with WISE magnitude values (mostly at 10-σ level or above). Some sources with upper limits are listed with the parameter Sigmpro value of null. Sources without WISE data represent that they are nominally detected, but no useful brightness estimate could be made. The position difference for those matched sources was also checked. The position difference is less than 1′′ for 141 sources (62.9%) and less than 5′′ for 200 sources (~90%), which reflects the reliability of our cross-identification. The control sample with corresponding WISE data are shown in Table 5.

Tables 6 and 7 present derived physical parameters from WISE measurements for our samples, including luminosities at four bands and spectral indices. For the OH maser galaxy sample, the measured isotropic OH line luminosity (logarithmic scale, in L_⊙) of each source was listed in Table 6, and the upper limit of OH luminosity for the control sample was also listed in Table 7, which was derived from their non-detection spectra (Darling & Giovanelli 2000, 2001, 2002): $\begin{array}{ccc}{\mathit{L}}_{\mathrm{OH}}^{\max }\mathrm{=}\mathrm{4}\mathit{\pi }\hspace{0.17em}{\mathit{D}}_{\mathit{L}}^{\mathrm{2}}\mathrm{1.5}\hspace{0.17em}\mathit{\sigma }\hspace{0.17em}\frac{\mathrm{\Delta }\hspace{0.17em}\mathit{v}}{\mathit{c}}\hspace{0.17em}\frac{{\mathit{\nu }}_{\mathrm{0}}}{\mathrm{1}\mathrm{+}\mathit{z}}\mathrm{·}& & \end{array}$where D_L is the luminosity distance, σ is the rms noise value of the nondetection spectra, ν₀ is the rest frequency of the 1667 MHz transition line, and Δv is the assumed rest-frame width of 150 km s^-1, respectively. We note that upper limits of some sources are larger than maser luminosities of some maser sources, which implies that some masers among non-detections were missed due to the low sensitivity.

3. Analysis and discussion

3.1. WISE luminosity distributions

For both the OH maser galaxy sample and the control sample, the WISE magnitudes were converted into flux densities using the formula, F = F₀ × 10^{− W/ 2.5}, where F₀ is the published flux density at zero-magnitude(309.54, 171.787, 31.674, and 8.363 Jy at 3.4, 4.6, 12, and 22μm, respectively²). Then the corresponding luminosities at four bands were calculated from derived flux densities and the luminosity distance (Darling & Giovanelli 2000, 2001, 2002: $\log \hspace{0.17em}\mathrm{(}\mathit{\nu }\hspace{0.17em}{\mathit{L}}_{\mathit{\nu }}\mathrm{)}\mathrm{=}\log \hspace{0.17em}\mathrm{(}\mathrm{4}\mathit{\pi }\hspace{0.17em}{\mathit{D}}_{\mathit{L}}^{\mathrm{2}}\mathit{\nu }\hspace{0.17em}{\mathit{F}}_{\mathit{\nu }}\mathrm{)}$, hereafter luminosity in units of L_⊙). The upper limits of luminosity were given for those sources with limits of magnitudes (see details in Sect. 2 and also Tables 6 and 7). For all OHMs with WISE detection (hereafter our statistics excluding those ~10% kilomasers, since they should not be powered by the same physical processes that dominate OHMs), the distributions of the luminosities at four WISE bands were plotted in Fig. 1 (in histograms with filled slash lines). Among them, the distributions for those OHMs from the Arecibo survey were shown in a gray color. For comparison, the distributions for the non-OHM sample (i.e., the control sample) were plotted in empty histograms. Since upper limits were given for the luminosity of some sources, a survival analysis thus should be applied. The logrank test was used here to check the significance of difference on the luminosity distributions. For the entire OHM sample and the non-OHM sample, the distributions of luminosity at WISE bands (except W4 band) are significantly different (with a chance probability less than 0.05, see details in Table 1). However, comparison of the entire OHM sample and the non-OHM sample should introduce the selection bias due to the Arecibo survey detection limits (z> 0.1). Thus, comparison should be more appropriate between the Arecibo OHM and the non-OHM sample, which were selected and observed with uniform criteria. For the Arecibo OHM and the non-OHM sample, the logrank test results show that the difference of the luminosity distribution at short wavelengths (i.e., 3.4 μm and 4.6 μm) is significant with a chance probability less than 0.01. The difference at W3 and W4 bands are statistically non-significant. The mean luminosities with its error (throughout the paper, given errors are the standard deviation of the mean) at four bands were also listed in Table 1 for the entire OHM sample, the Arecibo OHM and the non-OHM sample. With regard to non-OHM sources, OHMs tend to have lower luminosities, especially at short wavelengths, which can be supported by t-Test results (see details in Table 1). The t-Test probability that the difference of luminosity means at W1 and W2 band between the Arecibo OHM and the non-OHM sample is not significant is 0.012 and 0.033 (less than 0.05), respectively. Figure 2 presents the luminosity at four bands of our samples as a function of redshift. The Arecibo survey targeted sources with redshift of 0.1−0.3, while other OHM sources include many nearby luminous source with redshift less than 0.1. It shows clearly that the luminosity difference between OHMs and non-OHM sources should come from low OHM fraction at large 3.4 μm and 4.6 μm luminosity. This important point is discussed further in Sect. 3.4.

Fig. 1 Distributions of WISE luminosities (logarithmic scale, in L⊙) for OHM and non-OHM sources. Histograms with slash lines represent the distributions for the whole OHM sample. Among them, those filled in gray show the distributions for OHMs in the Arecibo survey. The distributions of non-OHM sources in the survey are presented in empty histograms.

Fig. 2 WISE luminosities of our samples as a function of redshift. Filled and empty circles represent OHM detections and non-detections from the Arecibo survey, respectively (e.g., Darling et al. 2002). Filled triangles show other OHM sources.

3.2. Color−color properties

Based on WISE magnitude values (see Tables 4 and 5), we obtained colors ([ W1 ] − [ W2 ], [W2 ] − [ W3 ], and [ W3 ] − [ W4 ]) for all sources in our samples. The mean WISE color values of OHMs (both Arecibo OHMs and other OHMs) appear to be slightly larger than those of non-masing sources. A t-Test was used to check if the difference is significant between OHM and non-OHM sources. For the Arecibo OHM and the non-OHM sample, the difference on the mean color W2 − W3 and W3 − W4 is significant with a chance probability of ~0.01. The difference on W1 − W2 mean values between them is not supported by the statistical test results (Table 2).

Figure 3 presents the W1 − W2 vs. W2 − W3 and W1 − W2 vs. W3 − W4 color−color diagrams for our samples. It is not clear yet whether the central engine of OHM host (U)LIRGs is dominated by an active galactic nucleus (AGN) or star formation activity, which is the key issue to understand production and physics of OHM in (U)LIRGs. The AGN fraction of OHM hosts is still open from previous studies at different wavelengths, such as optical spectroscopy (Baan et al. 1998; Darling & Giovanelli 2006), radio, FIR (Kandalyan 2005; Baan & Klöckner 2006), MIR properties (Willett et al. 2011a,b) etc. An existence of the AGN fraction difference is still uncertain between OHM and non-OHM (U)LIRGs (e.g., Darling & Giovanelli 2006; Willett et al. 2011b). To discuss possible radiation mechanisms of the MIR emission of OHM and non-OHM (U)LIRGs, we derived theoretical WISE colors from the blackbody model and the power-law model. According to the formula, F = F₀ × 10^{− W/ 2.5} (F₀: the zero-magnitude flux density, see Sect. 2), we can get equations for WISE colors as follows, $\begin{array}{ccc}& & \mathrm{\left[}\mathit{W}\mathrm{1}\mathrm{\right]}\mathrm{-}\mathrm{\left[}\mathit{W}\mathrm{2}\mathrm{\right]}\mathrm{=}\mathrm{2.5}\log \hspace{0.17em}\mathrm{(}{\mathit{F}}_{\mathrm{2}}\mathit{/}{\mathit{F}}_{\mathrm{1}}\mathrm{)}\mathrm{+}\mathrm{2.5}\log \hspace{0.17em}\mathrm{(}{\mathit{F}}_{\mathrm{01}}\mathit{/}{\mathit{F}}_{\mathrm{02}}\mathrm{)}\\ & & \mathrm{\left[}\mathit{W}\mathrm{2}\mathrm{\right]}\mathrm{-}\mathrm{\left[}\mathit{W}\mathrm{3}\mathrm{\right]}\mathrm{=}\mathrm{2.5}\log \hspace{0.17em}\mathrm{(}{\mathit{F}}_{\mathrm{3}}\mathit{/}{\mathit{F}}_{\mathrm{2}}\mathrm{)}\mathrm{+}\mathrm{2.5}\log \hspace{0.17em}\mathrm{(}{\mathit{F}}_{\mathrm{02}}\mathit{/}{\mathit{F}}_{\mathrm{03}}\mathrm{)}\\ & & \mathrm{\left[}\mathit{W}\mathrm{3}\mathrm{\right]}\mathrm{-}\mathrm{\left[}\mathit{W}\mathrm{4}\mathrm{\right]}\mathrm{=}\mathrm{2.5}\log \hspace{0.17em}\mathrm{(}{\mathit{F}}_{\mathrm{4}}\mathit{/}{\mathit{F}}_{\mathrm{3}}\mathrm{)}\mathrm{+}\mathrm{2.5}\log \hspace{0.17em}\mathrm{(}{\mathit{F}}_{\mathrm{03}}\mathit{/}{\mathit{F}}_{\mathrm{04}}\mathrm{)}\mathit{.}\end{array}$If the origin of radiation is the thermal emission described by the blackbody model ($\mathit{I}\mathrm{\left(}\mathit{\nu ,T}\mathrm{\right)}\mathrm{=}\frac{\mathrm{2}\mathit{h}{\mathit{\nu }}^{\mathrm{3}}}{{\mathit{c}}^{\mathrm{2}}}\frac{\mathrm{1}}{{\mathrm{e}}^{\frac{\mathit{h\nu }}{\mathit{kT}}}\mathrm{-}\mathrm{1}}$), we can obtain the flux density ratio at different frequencies as a function of temperature. Then the color values from the blackbody model at different temperatures can be derived from above equations. Similarly, if the radiation originate from the non-thermal emission by the power-law model (F_ν ∝ ν^{− α}), the flux density ratio at different frequency can be obtained as a function of the spectral index. Then the colors from the power-law model can be derived. In both color−color diagrams, color results from the blackbody model (marked with two typical temperature values) and the power-law model (marked with four typical spectral index of 0, 1, 2, 3) were also plotted. There is one common striking feature in both color-color diagrams (which also appears in other color−color diagrams in which two examples among them, W1 − W3 vs. W1 − W4 and W2 − W3 vs. W2 − W4 were shown at the bottom of Fig. 3) that both OHM and non-OHM host (U)LIRGs are far away from the blackbody model line and many of them can follow the path described by the power-law model.

However, polycyclic aromatic hydrocarbon (PAH) emission and silicate absorption are critical in the MIR for star-forming galaxies (Mateos et al. 2012). (U)LIRGs are complex interacting or merging system with intense starburst activity and thus star-forming template with more than one blackbody component is likely more appropriate.

Further, various WISE AGN criteria were also presented in Fig. 3. Consistent with previous results (e.g., Lake et al. 2011), part of both OHM and non-OHM host (U)LIRGs match these AGN criteria. For example, for the AGN criteria W1 − W2 > 0.8 defined by Stern et el. (2012), about 45% (48/106) of the entire OHMs sample and 42% (22/53) of the Arecibo OHM subsample match this AGN criteria. This AGN fraction is consistent with previous results from optical and radio studies (Baan et al. 1998; Baan & Klöckner 2006; Darling & Giovanelli 2006). The fraction is similar for our non-OHM (U)LIRGs sample (92/219, ~42%), which supports no difference for the AGN fraction between OHMs and non-OHM (U)LIRGs (Darling & Giovanelli 2006). Our results also support the existence of different types of ULIRGs, that is, ULIRGs can be classified into H II-like ULIRGs, AGN-like ULIRGs (with larger W1 − W2 value), and composite ULIRGs, based on their optical emission line properties (Su et al. 2013).

Fig. 3 WISE color−color diagrams (upper left: [ W1 ] − [ W2 ] vs. [ W2 ] − [ W3 ],upper right: [ W1 ] − [ W2 ] vs. [ W3 ] − [ W4 ]; lower left: [ W1 ] − [ W4 ] vs. [ W1 ] − [ W3 ], lower right: [ W2 ] − [ W4 ] vs. [ W2 ] − [ W3 ]) for OHMs and non-OHM sources. All symbols are the same as Fig. 2. The dashed lines show results from the blackbody model at different temperatures (squares for typical temperature values) and the solid lines present results from the power-law model with different spectral index (F ∝ ν− α, four typical spectral index of 0, 1, 2, 3 are shown). Various AGN criteria were also shown in upper panels represented by dotted line (defined by Stern et al. 2012), a dashed line wedge (Mateos et al. 2012), and a dash-dotted line wedge (Jarrett et al. 2011), respectively.

3.3. WISE spectral indices

Given that the MIR radiation of our samples was described by the power-law model (F ∝ ν^{− α}), the spectral index α can be thus derived as $\begin{array}{ccc}{\mathit{\alpha }}_{\mathrm{21}}\mathrm{=}\log \mathrm{(}{\mathit{F}}_{{\mathit{\lambda }}_{\mathrm{2}}}\mathit{/}{\mathit{F}}_{{\mathit{\lambda }}_{\mathrm{1}}}\mathrm{)}\mathit{/}\log \mathrm{(}{\mathit{\lambda }}_{\mathrm{2}}\mathit{/}{\mathit{\lambda }}_{\mathrm{1}}\mathrm{)}\mathit{.}& & \end{array}$Figure 4 presents the distributions of the spectral indices between four WISE bands. Unlike the distributions of α_{4.6−3.4 μm} and α_{12−4.6 μm} (Fig. 4), different distributions of α_{22−12 μm} can be obviously found between OHM (both the entire OHM sample and the Arecibo OHM subsample) and non-OHM sample. This is supported by the Kolmogorov-Smirnov test (K-S test) results (see details in Table 3) where the difference on the distribution of α_{22−12 μm} is statistically distinguishable. Both the Arecibo OHM and the non-OHM sample come from the same parent population at a probability of 5 × 10^-3. Their mean values of α_{22−12 μm} are 2.16 ± 0.09 and 1.90 ± 0.04 for the Arecibo OHM and the non-OHM sample, respectively. The difference is statistically significant with a chance probability of 0.008. This is consistent with previous results from the Spitzer spectra and photometry, where OHMs show steeper continuum from 15 μm to 35 μm (Willett et al. 2011b).

This difference can also be found in the distribution of α_{22−3.4 μm} (Fig. 4). The distribution of α_{22−3.4 μm} peaks at the ~2.5 for OHM sample (both megamaser samples) and at the 1.5−2 bin for non-OHM sample. The K-S test probability that both the Arecibo OHM and the non-OHM sample come from the same parent population is 7 × 10^-3. The mean value is ⟨ α_{22−3.4 μm} ⟩ = 1.96 ± 0.06 and 1.77 ± 0.03 for the Arecibo OHM and the non-OHM sample, respectively. The difference can be supported by the t-Test results, and the probability that the difference on α_{22−12 μm} means between the Arecibo OHM and the non-OHM sample is not significant is 0.004.

Fig. 4 Distributions of the WISE spectral indices of OHM and non-OHM host (U)LIRGs. All pattern types are the same as Fig. 1.

3.4. WISE infrared luminosities versus colors

Fig. 5 WISE luminosity at 3.4 μm (left panel) and 22 μm (right panel) versus color ([3.4 μm−22 μm]) for OHMs and non-OHM sources. All symbols are the same as Fig. 2. The solid line shows linear fits for the Arecibo OHMs, and its error ranges were presented by two dashed lines. The Arecibo OHM sources mostly locate within the region limited by these two dashed lines.

Previous statistical studies indicated that the OHM fraction in (U)LIRGs tends to rise with increasing FIR luminosity and warmer far-IR color (larger 60 μm/100 μm, e.g., Baan et al. 1992; Darling & Giovanelli 2002). Here, possible relations were investigated between OHM fraction and WISE MIR luminosity, which are colors of megamaser host (U)LIRGs. As analyzed in Sect. 3.1, for OHM and non-OHM samples, it shows significant different distributions of the luminosity at W1 band and similar distributions of their W4 luminosities. Figure 5 plots the WISE luminosity (left panel: 3.4 μm, right panel: 22 μm) versus color [ W1 ] − [ W4 ] ([3.4 μm]−[22 μm]) for our samples. The fraction of OHMs tends to increase with cooler MIR colors (larger color values, or, larger F_{22 μm}/F_{3.4 μm}). For the Arecibo survey sample, the fraction is about 12%, 20%, 32%, and 50% in [ W1 ] − [ W4 ] bins of <7, 7–8, 8–9, and >9, respectively.

The OHM fraction is much lower in large MIR luminosity at 3.4 μm, which causes the different distribution of the 3.4 μm luminosity between OHMs and non-OHM sources (see Sect. 3.1). Among the Arecibo sample, there is only one OHM source among 18 sources with a 3.4 μm luminosity (log  ν₁L_ν1) that is larger than 10.5. It may hint that OHM emission is not related to MIR emission at short wavelengths, which is far from the FIR emission and the pumping energy of OHMs (e.g., radiation at 35 μm, Skinner et al. 1997; or 53 μm, Lockett & Elitzur 2008). In addition, high 3.4 μm luminosity may mean a harder radiation environment, and its strong UV radiation can dissociate OH even in the ice phase (Andersson & van Dishoeck 2008). If so, samples without extreme luminous 3.4 μm sources may improve detection rate of OHMs in the future survey.

No significant difference on the OHM fraction can be found in each bin of the 22 μm luminosity, which is consistent with our previous result (Sect. 3.1) where both the Arecibo OHM and the non-OHM sample have similar distributions of the 22 μm luminosity. Among the Arecibo sample, the OHM fraction is about 25%, 20%, and 19% in the 22 μm luminosity bins of 9−10, 10−11, and 11−12, respectively. However, one trend of a rising 22 μm luminosity with increasing MIR color value ([ W1 ] − [ W4 ]) is apparent, especially for the subsample of Arecibo OHM alone. The linear fit gives log L_{22 μm} = (0.27 ± 0.06) ([ W1 ] − [ W4 ]) + (8.68 ± 0.46) for the Arecibo OHM subsample with a Spearman’s rank correlation coefficient R = 0.59. The candidates of the successful Arecibo OHM survey were (U)LIRGs, which were selected from the IRAS Point Source Catalog Redshift Survey (15 000 IRAS galaxies with f_{60 μm}> 0.6 Jy, Saunders et al. 2000) within Arecibo sky coverage (declination range: 0°–37°). This correlation of MIR luminosity (at 22 μm) and color for OHMs should be good to place further constraints on OHM candidates and, thus, enhance detection rate by future advanced telescopes. The OHM-hosted (U)LIRGs mostly lie in a distinct region of luminosity-color diagram (limited by two dashed lines in Fig. 5, which were plotted from the linear fitting line with intercept error). Although many of the non-detection sources also located in the limited region, there are about 20% of non-detections that are located outside this region. It implies the detection rate could be increased much (~25%), if those non-detections outside the limited region among the Arecibo survey candidates are excluded. With respect to Arecibo, the future FAST (the Five-hundred meter Aperture Spherical radio Telescope) has a better sensitivity and a larger sky coverage (declination range of −15°–65° with a zenith angle of 40 degrees, Nan et al. 2011; Li et al. 2013), which provides enormous potential for searching and studying OHMs (Zhang et al. 2012). Providing this correlation exists, WISE (U)LIRGs sources located in the limited region (Fig. 5) will be OHM candidates for future surveys.

3.5. Maser emission and MIR luminosities of OHM hosts

Fig. 6 Maser luminosity versus the WISE MIR luminosity (upper panels for 3.4μm and 4.6 μm, lower panels for 12μm and 22 μm luminosity) of maser hosts. The Arecibo OHMs are shown in filled circles and other OHMs in filled triangles. The upper limits on OH line luminosity of the non-OHM sample were also presented (cross-triangles). The solid and the dashed line show linear fits without and with considering the Malmquist effect, respectively.

Infrared radiation pumping of OHMs were supported by both observations (e.g., Skinner et al. 1997; Baan et al. 1998; Kegel et al. 1999; He & Chen 2004; Mcbride et al. 2013) and theoretical models (Baan 1985; Henkel et al. 1987; Henkel & Wilson 1990; Burdyuzha & Vikolov 1990; Randell et al. 1995; Lockett & Elitzur 2008). The statistical analysis showed that the OHM luminosity was found to be increasing with the FIR luminosity of maser hosts as ${\mathit{L}}_{\mathrm{OH}}\mathrm{\propto }{\mathit{L}}_{\mathrm{FIR}}^{\mathit{\gamma }}$ (1 < γ ≤ 2, Martin et al. 1988; Baan 1989; Kandalian 1996; Diamond et al. 1999; Darling & Giovanelli 2002). In the MIR, Willett et al. (2011b) combined both theory and Spitzer observation to explore the fundamental causes of OHM activity and found that OHMs prefer deeper 9.7 μm silicon absorption and warmer dust. Here, the relation between maser emission and the WISE MIR luminosity of maser hosts was investigated.

In Fig. 6, the OH maser luminosity was plotted against the MIR luminosity (the WISE luminosity at four bands was presented, respectively) for both the Arecibo OHMs and other OHMs. For the non-OHM sample, the upper limits on OH line luminosity were also presented. Although there are a few overlap between the detections and non-detections in the Arecibo survey, many known OHMs mix with non-OHM sources, even with some lower than the non-OHM sources. It implies that some low-level OHMs (e.g., L_OH < 100 L_⊙) should be undetected in the non-OHM sample (Darling & Giovanelli 2002), and it should be pointed out that this may affect statistical relation of maser luminosity and the MIR luminosity of its host. All panels show no significant correlation between luminosities for the Arecibo OHM sample alone. Since the Arecibo survey sample is limited to the upper end of the LIRG population, the entire megamaser sample is a more appropriate sample to resolve the relationship (Darling & Giovanelli 2002). For the entire megamaser sample, it appears that the maser luminosity increases along with the increases of the MIR luminosity of maser host galaxies. For example, for the panel of 22 μm luminosity versus the OH maser luminosity, a linear least-square fit gives Spearman’s rank correlation coefficient R = 0.42. However, the Malmquist bias should be considered, since both luminosities are correlated with the luminosity distance. After taking the Malmquist effect into account (e.g., Kandalyan & Al-Zyout 2010; Darling & Giovanelli 2002), the maser-MIR luminosity correlation becomes nonsignificant (R ~ 0.25). Unlike the strong correlation of FIR and maser luminosity, the marginal correlation of MIR and maser luminosity may support indirectly that the OHM emission is pumped by FIR radiation with values, such as 35 μm (e.g., Skinner et al. 1997) or 53 μm (Lockett & Elitzur 2008).

4. Summary

To investigate MIR properties of OHM host (U)LIRGs, the entire OHM sample and one (U)LIRGs sample without detected maser emission were compiled, and cross-identifications of these (U)LIRGs with WISE catalog were made. Our results show the following:

• 1)

  Based on their color-color diagrams, both OHM- andnon-OHM-hosted (U)LIRGs are far away from the singleblackbody model line and many of them follow the path describedby the power-law model well. According to one AGN WISEcriteria, the AGN fraction is similar for both OHM and non-OHMsamples with a value of ~40%.

• 2)

  The spectral indices are derived for both the OHM sample and the non-OHM sample devoid of detected OH emission. For the Arecibo OHM sample and the non-OHM sample, the distribution of the spectral index α_{22−12 μm} is significantly different, where OHMs tend to have larger spectral indices α_{22−12 μm} than non-OHM sources. The mean values of ⟨ α_{22−12 μm} ⟩ are 2.16 ± 0.09 and 1.90 ± 0.04 for the Arecibo OHM sample and the non-OHM sample, respectively.

• 3)

  The Arecibo OHMs tend to have a lower luminosity at short MIR wavelengths than non-OHM sources, which should come from the low OHM fraction among the Arecibo sample with large 3.4 μm and 4.6 μm luminosity, and OHM fraction tends to increase with cooler MIR colors (larger F_{22 μm}/F_{3.4 μm}). These clues should be helpful for guiding a future OHM survey, such as choosing samples that exclude extreme luminous sources at short MIR wavelength (e.g., at 3.4 μm) and choosing sources with cooler MIR colors.

• 4)

  For the Arecibo OHMs alone, the MIR luminosity at 22μm is found to be correlated with the MIR color [ W1 ] − [ W4 ]. A linear fit gives log L_{22 μm} = (0.27 ± 0.06) ([ W1 ] − [ W4 ]) + (8.68 ± 0.46) with a Spearman’s rank correlation coefficient R = 0.59. This possibly provides suitable constraints on sample selections for OHM surveys by future advanced telescopes (e.g., FAST).

• 5)

  Unlike the strong correlation of the FIR and the maser luminosity, the correlation of the MIR luminosity and the maser luminosity tends to be marginal. It suggests that the pumping of OHM emission is dominated by the FIR radiation, instead of the MIR radiation.

Online material

Acknowledgments

This work is supported by China Ministry of Science and Technology under State Key Development Program for Basic Research (2012CB821800) and the Natural Science Foundation of China (No. 11178009, 11473007). We made 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. And the NASA Astrophysics Data System Bibliographic Services (ADS) and the NASA/IPAC extragalactic Database (NED) were also used, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA.

References

All Tables

All Figures

	Fig. 1 Distributions of WISE luminosities (logarithmic scale, in L⊙) for OHM and non-OHM sources. Histograms with slash lines represent the distributions for the whole OHM sample. Among them, those filled in gray show the distributions for OHMs in the Arecibo survey. The distributions of non-OHM sources in the survey are presented in empty histograms.
In the text

	Fig. 2 WISE luminosities of our samples as a function of redshift. Filled and empty circles represent OHM detections and non-detections from the Arecibo survey, respectively (e.g., Darling et al. 2002). Filled triangles show other OHM sources.
In the text

Fig. 3 WISE color−color diagrams (upper left: [ W1 ] − [ W2 ] vs. [ W2 ] − [ W3 ],upper right: [ W1 ] − [ W2 ] vs. [ W3 ] − [ W4 ]; lower left: [ W1 ] − [ W4 ] vs. [ W1 ] − [ W3 ], lower right: [ W2 ] − [ W4 ] vs. [ W2 ] − [ W3 ]) for OHMs and non-OHM sources. All symbols are the same as Fig. 2. The dashed lines show results from the blackbody model at different temperatures (squares for typical temperature values) and the solid lines present results from the power-law model with different spectral index (F ∝ ν− α, four typical spectral index of 0, 1, 2, 3 are shown). Various AGN criteria were also shown in upper panels represented by dotted line (defined by Stern et al. 2012), a dashed line wedge (Mateos et al. 2012), and a dash-dotted line wedge (Jarrett et al. 2011), respectively.

In the text

	Fig. 4 Distributions of the WISE spectral indices of OHM and non-OHM host (U)LIRGs. All pattern types are the same as Fig. 1.
In the text

	Fig. 5 WISE luminosity at 3.4 μm (left panel) and 22 μm (right panel) versus color ([3.4 μm−22 μm]) for OHMs and non-OHM sources. All symbols are the same as Fig. 2. The solid line shows linear fits for the Arecibo OHMs, and its error ranges were presented by two dashed lines. The Arecibo OHM sources mostly locate within the region limited by these two dashed lines.
In the text

Fig. 6 Maser luminosity versus the WISE MIR luminosity (upper panels for 3.4μm and 4.6 μm, lower panels for 12μm and 22 μm luminosity) of maser hosts. The Arecibo OHMs are shown in filled circles and other OHMs in filled triangles. The upper limits on OH line luminosity of the non-OHM sample were also presented (cross-triangles). The solid and the dashed line show linear fits without and with considering the Malmquist effect, respectively.

In the text