Galaxy and Mass Assembly (GAMA): A WISE Study of the Activity of Emission-line Systems in G23

The optical line fluxes of the galaxies are extracted from the G23 GAMA II catalog ("GaussFitSimplev05" from within the SpecLineSFR DMU; Gordon et al. 2017) with redshifts >0.002 and a magnitude limit of i < 19.2 mag.

The spectra are fitted with the IDL code "mpfitfun" (Markwardt 2009), which uses a Levenberg–Marquardt nonlinear least-squares minimization to identify the best-fitting parameters for the model given the data and its associated uncertainties.

Among the lines available in the catalog, we use Hβ λ4861, [O iii] λ4959, [O iii] λ5007, [N ii] λ6548, Hα λ6563, [N ii] λ6583, [S ii] λ6716, and [S ii] λ6731.

For lines that are always expected to be narrow such as [S ii], the fitting can either be a straight line (in the case of nondetection) or a Gaussian. Fitting Hβ and Hα in combination with their associated lines [O iii] λ4959/λ5007 and [N ii] λ6548/λ6583, respectively, can have more levels of complexity above the continuum. For example, fitting Hα + [N ii] λ6548/λ6583 can increase in complexity as follows:

• 1.  
  no emission or absorption, just continuum,
• 2.  
  Hα in absorption and no [N ii] λ6548/λ6583 emission,
• 3.  
  [N ii] λ6548 + Hα + [N ii] λ6583 all with narrow emission,
• 4.  
  [N ii] λ6548 + [N ii] λ6583 in emission + Hα in absorption,
• 5.  
  [N ii] λ6548 + [N ii] λ6583 in emission + Hα in emission and absorption,
• 6.  
  [N ii] λ6548 +[N ii] λ6583 in emission + Hα in narrow plus broad emission.

For each of the above fits, three different model selection methods are used to give a single global score based on whether a more complicated model should be chosen versus a simpler one and vice versa. For models with the same number of fitted parameters, the model with the lowest χ² value is chosen.

The following limits are applied to the fitting model: the line position is maintained within 200 km s⁻¹ of the expected position. The width of the Gaussian (σ) for the narrow emission lines is constrained by $0.75{\sigma }_{\mathrm{inst}}\lt \sigma \lt \sqrt{{500}^{2}+\sigma {{}^{2}}_{\mathrm{inst}}}$, where σ_inst is the resolution of the spectrum in km s⁻¹. The precedent constraints are made larger for broad emission lines, $\sqrt{{500}^{2}+\sigma {{}^{2}}_{\mathrm{inst}}}\lt \sigma \lt \sqrt{{5000}^{2}+\sigma {{}^{2}}_{\mathrm{inst}}}$, and the position of the line is maintained within 400 km s⁻¹.

Only the AAOmega spectra are flux calibrated and can be used as individual line fluxes unlike in the line ratio used for both calibrated and uncalibrated spectra. For doublets such as [O iii] λ4959/λ5007, [N ii]λ6548/λ6583, and [S ii] λ6716/λ6731, the line with the longest wavelength is generally the strongest (e.g., [O iii] λ5007 can be up to three times stronger than [O iii] λ4959) and is used to estimate the Hα/[O iii] line ratio. Hα and Hβ must be corrected for stellar absorption as it was not accounted for in the fitting.

The whole process of extracting the WISE-resolved sources arises from the fact that the survey was optimized for the detection of point-source galaxies. Potentially resolved galaxies based on the W1 (3.4 μm) band are selected from the ALLWISE catalog using the PSF goodness-of-fit metric, w1rchi2. We search for values of w1rchi2 > 2, remeasure them with our special resolved source pipeline that characterizes the global and surface brightness emission, revealing sources that are extended beyond the PSF. These candidates are then characterized using the resolved-galaxy pipeline.

Using the reprocessed WISE imaging described above, the photometry can be remeasured as appropriate for a resolved source. The different steps are done via a custom software which adapted tools developed for 2MASS extended sources (Jarrett et al. 2000) and the WISE pipeline (Jarrett et al. 2011, 2013; Cutri et al. 2012). The process is semiautomated such that each level can be visually inspected and corrected if necessary. The details are as follows:

• (i)  
  The point sources are removed by a combination of PSF subtraction and pixel masking. The bright stars in the vicinity are also masked out for a better estimation of the background. The shape of the galaxy is determined based on the 3σ isophote (of the background rms), which is considered to be constant at all isophotes. The first set of photometry is derived at the 1σ isophote, which already encloses more than 90% of the light (Jarrett et al. 2019). A double-Sérsic fit (Graham & Driver 2005) is used to estimate the light beyond the 1σ isophote as well as computing the curve of growth asymptotic fluxes extending down to convergence. Extracted sources are then removed from the images using their smoothed light distribution.
• (ii)  
  The process is repeated until all sources are extracted.
• (iii)  
  The Rfuzzy parameter (see Cluver et al. 2014) is the second-order intensity-weighted moment that allows a separation between faint, resolved, and unresolved systems. Each type of object has a characteristic Rfuzzy value that is systematically smaller than that of the resolved galaxies. This step is important in order to get our final sample of resolved galaxies.

The final step is visual inspection to validate and identify blends and other complex scenarios. The whole area of study is divided into smaller areas centered on galaxies identified as resolved and processed by the pipeline. Each one of the galaxies is visualized (by the first author) to make sure that it is definitely not a star or unresolved galaxy, that the deblending is well performed (mostly for close proximity blend cases), and the apertures and position angles follow well the 2D light profile. The pipeline does well when the background is smooth and the signal-to-noise ratio (S/N) high (10 or higher). It needs human assistance in dense and low-S/N environments, and with closely blended pairs. The parameters and source contamination can be interactively adjusted (see Jarrett et al. 2017).

To account for the redshifted emission across the optical and infrared bands, we applied a "k-correction" to the magnitudes based on spectral energy distribution (SED) fitting before the derivation of any physical value. The SEDs are built by combining flux densities spanning the optical, the near-IR, and WISE mid-IR; for the optical and near-infrared data, we have used preliminary photometry available in G23 (generated by ProFound; Robotham et al. 2018). They are fitted to empirical templates of well-studied galaxies from Brown et al. (2014a, 2019) and Spitzer-SWIRE/GRASIL (Silva et al. 1998; Polletta et al. 2006, 2007). From the rest-frame-corrected catalog, key parameters such as stellar mass were derived using the W1 in-band luminosity and the mass-to-light ratio (M/L; Cluver et al. 2014; Querejeta et al. 2015; Kettlety et al. 2018; Meidt et al. 2012). The SFR (SFR_12μm) and the specific star formation rate (sSFR) are estimated using νLν luminosities and the mid-IR to total-IR calibration from Cluver et al. (2017). The SEDs allow, to some extent, the differentiation of the AGNs from normal SF galaxies. Examples of SEDs are presented in the case studies in the Appendix.

For the comparative study of WISE colors and optical BPT diagrams, we have applied a magnitude and redshift cut of W1 < 15.5 mag (in Vega) and z < 0.3, respectively, to the WISE catalog of galaxies in G23 having redshifts, hereafter referred to as "the sample" representing (B) in Table 1. For the resolved galaxies, this corresponds to an average S/N ≈ 30 in W1, and corresponding lower values for the W2 and W3 bands. An S/N limit of 3 is imposed on all the optical spectral lines used in the BPT. Table 1 shows how the number of galaxies decreases from the initial WISE-GAMA G23 catalog to the final number of galaxies used for the optical and infrared diagnostics. (A), (B), (C), and (D) represent each step. As an example, "A" with W1 < 15.5 mag, z < 0.3 means that in the second row, a magnitude limit of W1 < 15.5 mag and redshift z < 0.3 are applied to the total WISE-GAMA G23 catalog.

In Figure 2, the sample (dashed gray) is presented along with the WISE-GAMA galaxy catalog, with more than 70% of galaxies found at redshifts less than 0.2. The magnitude cut in W1 (the most sensitive band) ensures detections with high S/N, but also selects well-deblended sources (based on WISE images), hence, good photometry quality in general.

The magnitude limit (in W1) of 15.5 mag eliminates most of the low-mass galaxies, log (M_stellar/M_⊙) < 9, which are often misclassified using only the WISE color–color diagram (Hainline et al. 2016). Using a sample 18,000 nearby dwarf galaxies (M_* < 3 × 10⁹ M_⊙), selected from the Sloan Digital Sky Survey, they found that while the WISE color–color diagram is reliable in classifying moderate- to high-mass galaxies, the W1 – W2 color alone should not be used for dwarf galaxies.

As can be seen in Figure 3, a cut of 15.5 mag offers the best of our carefully measured WISE-resolved galaxies (2383, ∼90% of the WISE-resolved sample) and also the brightest unresolved sources (7426 point-like or compact galaxies). A brighter magnitude cut will lead to fewer resolved galaxies, and a fainter cut will include more faint point-source galaxies with lower S/N. In this way, 15.5 mag is the best balance between completeness and mid-infrared (W1, W2, and W3 bands) photometric quality, giving a final sample of 9809 galaxies.

The W1 – W2 versus W2 – W3 colors (color–color diagram) of the sample, with an S/N cut of 5, 5, and 2 in W1, W2, and W3, respectively, are presented in Figure 4. The lower S/N for W3 is to account for the lower sensitivity in this band relative to the W1 and W2 bands, as well as taking advantage of the large dynamic range (five orders of magnitude) seen in the W2 – W3 color. The W1 – W2 color is the more critical of the two, as its lower dynamic range requires higher accuracy. The AGN box (Jarrett et al. 2011) is the expected location of luminous AGNs, while the W1 – W2 = 0.8 mag is a conservative limit for AGNs; galaxies above are classified as WISE (mid-infrared) AGNs and are frequently in the QSO class. The dashed blue line is the fit to the sequence (Equation (1)) seen in the color–color diagram for all catalogs of WISE-resolved sources (Cluver et al. 2017; Jarrett et al. 2019). The relation is described by Jarrett et al. (2019) as

$$\begin{eqnarray}&&({\rm{W}}1-{\rm{W}}2)=\left[0.015\times \ {{\rm{e}}}^{\tfrac{({\rm{W}}2-{\rm{W}}3)}{1.38}}\right]-0.08.\end{eqnarray} \tag{ 1 }$$

In Figure 5, fewer than <2% of the galaxies in our sample have a stellar mass log (M_stellar/M_⊙) ≤ 9. We therefore do not expect the result in this study to be significantly affected by misclassifications related to dwarf galaxies.

Zoom In Zoom Out Reset image size

In this section, we investigate the optical emission-line properties of the sample derived from our WISE-GAMA catalog. Special care has been taken to derive the photometry of the resolved sources, which makes our sample a high-quality data set for a comparative study between the optical and mid-infrared diagnostics of the activity. The constraints put on the magnitude and the redshift allow us to take advantage of the best photometry from WISE and avoid the required optical lines being shifted out of the wavelength range of the AAOmega spectrograph used in GAMA. All of the optical spectra have been visually inspected to detect any systematic fitting anomalies. To this end, GAMA has a flag for the quality of each fit corresponding to each of the optical lines of interest, but sometimes human supervision is needed for the final decision, notably with broad-line systems. The selection criteria are that we detect the emission lines of Hα, Hβ, and [O iii] as well as [N ii] or [S ii] depending on the diagnostic diagram used. Hα and Hβ are corrected for both reddening and stellar absorption using Equation (2), and all the lines have S/N > 3 (note that this strict criterion eliminates a large fraction of sources from our sample). The Balmer line absorption correction is applied as follows:

$$\begin{eqnarray}&&{F}_{\mathrm{corr}}=\left[\displaystyle \frac{\mathrm{EW}+2.5}{\mathrm{EW}}\right]\times {F}_{\mathrm{obs}},\end{eqnarray} \tag{ 2 }$$

where EW is the equivalent width of the line of interest, F_obs is the observed flux, and 2.5 is the approximate correction required for GAMA spectra (see Gunawardhana et al. 2011; Hopkins et al. 2013).

Zoom In Zoom Out Reset image size

The BPT classification on the plane [O iii]5007/Hβ versus [S ii]6716,6731/Hα is shown in Figure 6. A count of 1037 out of the 9809 initial sample satisfy the conditions mentioned above. Seventy-five galaxies are classified as Seyferts and five as LINERs.

Zoom In Zoom Out Reset image size

The redshift (z) cut of 0.3 is only for GAMA-derived spectra. But the catalog also contains 6dF and 2dF spectra, where z = 0.13 and z = 0.2, respectively, should be used as upper limits when using the [S ii] line flux.

The second BPT diagnostic diagram, [O iii]5007/Hβ versus [N ii]6583/Hα, is shown in Figure 7. Similar constraints have been applied. In total, 1173 galaxies are selected, 122 are classified as AGNs, 170 as composites, and 881 as SF.

Zoom In Zoom Out Reset image size

The [N ii]6583/Hα line varies as a function of metallicity, which has a correlation with stellar mass (Wu et al. 2016). Therefore, the low-metallicity AGNs could be found below the extreme starburst line presented in Kewley et al. (2019). The study by Kauffmann et al. (2003) shifted the Kewley et al. (2006) line to create a new limit for pure star-forming galaxies. The mixed region in between the two lines supposedly made up by galaxies exhibiting both AGN and star-formation activities is referred to as composites (Kewley et al. 2019). The current optical separation lines derived based on local galaxies are subject to variations with redshift and can be reliably used only at redshift z < 1 (Kewley et al. 2013).

The redshift limit used in our study is 0.3 (local universe), where low-metallicity AGNs are extremely rare (Groves et al. 2006).

This is further confirmed by comparing both optical methods in Figure 8. The AGNs seen as outliers in the figure have extremely broad Hα lines that are difficult to be accurately measured as they are deeply entangled with the blended [N ii] lines. The broad-line (BL) AGNs are specially addressed in the current work. One more noticeable feature is the quasi-presence of the composites in the SF region, Figure 8, which appears to confirm indeed the rarity of low-metallicity AGNs in our sample, susceptible of being found in the composite region. Except for particular cases of BLAGNs mentioned above, the two diagrams largely agree on classifying AGN and SF galaxies. So, we do not expect our result to be significantly affected by the misclassification of AGNs as SF caused by the variation in metallicity.

Zoom In Zoom Out Reset image size

The [N ii] emission lines are stronger than the [S ii] lines, and using the [N ii] rather than the [S ii] line has the added benefit of including composite galaxies in the analysis. Based on these facts, we will be using [O iii]5007/Hβ versus [N ii]6583/Hα as the primary optical AGN diagnostic plane which will be compared to the W1 – W2 versus W2 – W3 color plane throughout this work, except where explicitly stated otherwise.

For the comparison between WISE and the BPT, the detection S/N in W1, W2, and W3 is also required to be greater than 5, 5, and 2, respectively, giving a total of 1154 galaxies. We use the results from Figure 9, which plots the optical classification from Figure 7 displayed in the WISE color–color diagram, keeping the same color-coding scheme. The violet dots represent the optically selected AGNs (designated as oAGN hereafter), the blue points are the optical star-forming galaxies (oSF), and the green points are the composites (referred to as "composite"), which are in between the two. The crosses represent galaxies identified as broad-line AGNs (BLAGNs). The BLAGN selection is based on the method described in Gordon et al. (2017). However, we only keep the galaxies clearly seen as BLAGNs without ambiguity after visual checking. We overplot in magenta and gold circles the galaxies classified as Seyferts and LINERs (where available in the current diagram) from Figure 6.

Zoom In Zoom Out Reset image size

We establish a 2σ offset from the SF color–color sequence (Equation (1)) delineated by the green dashed curve in Figure 9. The color–color diagram is divided into three zones represented by the circled numbers: 1, 2, and 3. Zone 1 (pink shade) is where the WISE obscured AGNs, QSOs, LINERs, and ULIRGs are located. Zone 2 (gray shade), just below, is where the low-power Seyferts and LINERs reside—we call it the "mWarm" zone signifying warmer W1 – W2 color due to greater nuclear activity and corresponding accretion disk emission. Zone 3 (light blue shade) contains the sequence of WISE star-forming galaxies, mostly a mixture of intermediates and star-forming disk galaxies having low (blue) W1 – W2 colors.

The diagram in Figure 10 summarizes the galaxy classification considering both their optical emission-line properties and their infrared WISE colors. The galaxies displayed in Figure 9 are divided into different groups, taking into account both their optical line properties and their mid-IR colors in WISE. Galaxies classified as AGNs in Figure 7 (BPT), which are in color–color Zone 1 (see Figure 9(b)), are labeled as oAGN (mAGN). The optically classified SF or composites in Zone 1 are labeled non-oAGN (mAGN). The same applies to Zone 2 where the two groups are called respectively oAGN (mWarm) and non-oAGN (mWarm). The mWarm are the galaxies warmer than the typical SF galaxies based on the W1 – W2 color. Finally, in Zone 3, all of the galaxies classified as AGNs in the optical which reside in Zone 3 are called oAGN (mSF). The optical SF and composites in Zone 3 (see Figure 9(a)) are called "SF" and "Composite," respectively.

Zoom In Zoom Out Reset image size

There is a clear presence of oAGNs in the WISE star-forming zone. Some non-oAGN objects are located in both the mid-infrared warm (mWarm) and the obscured-AGN zone of which 25% in the mAGN zone are composites and 50% are broad-line AGNs misclassified by the BPT (Figure 9(a)). Having some optically classified composite galaxies in the mid-infrared warm or AGN zone is expected as they are a mixture of both AGN and star formation. However, the vast majority lie in the mid-infrared star-forming zone, meaning that their host galaxies generally dominate over the AGN activity. The Seyferts are found in the mid-infrared AGN zone as well as in the SF and are generally classified as oAGNs while in most cases the LINERs (classification based on [S ii] line) designate a non-oAGN (classification based on [N ii] line) and are classified as SF galaxies in the WISE color–color diagram. The nature of LINERs is not well understood, and while they may be associated with SF and shocks thereof, they may also harbor low-luminosity AGNs (Flohic et al. 2006). Our data and analysis are not able to properly address this ongoing issue, and we provide no further analysis on LINERs.

Another trend is the presence of oAGNs in the WISE star-forming zone. There are even a few BLAGNs with SF colors, which is intriguing because WISE seems to be generally very sensitive to broad Hα lines. They constitute very good specimens to be followed up in radio and X-rays.

We now reproject back into the BPT plane the new optical and mid-IR classification codes, as shown in Figure 11(a). The figure shows a clear separation of the oAGN (mAGN) into two subsamples. The lower-right corner above the green curve are broad-line AGNs whose line ratios have been overestimated by the line-fitting algorithm in GAMA. Here it is striking to see all the galaxies with [N ii]/Hα >1 being broad lines with generally overestimated [N ii] fluxes. A larger sample would likely have shown a few non-BLAGNs for which [N ii]/Hα flux ratios are truly >1, but these cases are rare. It is sometimes extremely challenging, if not impossible, to separate Hα and [N ii] when the Hα line becomes broad. Even a reassessment of the flux, which will probably give a more reasonable ratio, lower than 1, will shift the galaxy (BLAGN) into the SF zone unless the [O iii] line's flux is highly elevated. Those galaxies with an underestimated flux ratio constitute 50% of our non-oAGN (mAGN) sample. The GAMA pipeline does not handle the broad lines well, notably the blending of Hα and [N ii]. Therefore, the BPT is not appropriate using the automated pipeline data from GAMA. We have displayed them in this diagram to show where they might correctly or incorrectly be located in the BPT diagram.

Zoom In Zoom Out Reset image size

The oAGN (mSF) are generally located on the upper-left side of the AGN zone along with the oAGN (mAGN). Although classified as AGNs both in WISE and optical, some of the latter galaxies are labeled as either peculiars or late-type spirals by their SED template fits. This could be explained by the fact that they are sharing similar properties with some of the oAGN (mSF) , which are also classified as non-AGN based on the SED. Among the non-oAGNs classified as mAGNs, only six are oSFs. The trend is that most of the WISE AGNs are also optical AGNs or at least composite galaxies that need further attention using other diagnostic methods based on radio or X-ray data. Zone 2 (mWarm) seems to be a transition zone where we find the optical star-forming galaxies, the composites, and the BLAGNs in similar proportions.

Figure 11(b) shows the host stellar mass as a function of [O iii] line luminosity. In this figure, which requires the use of individual fluxes unlike the BPT based on line ratios, only the calibrated measurements are used (see Hopkins et al. 2013 for the data description). The number of galaxies, therefore, is reduced. The distribution can be separated into two regions, with optical AGNs generally having luminosity greater than 10⁴¹ erg s⁻¹ and the SF galaxies with lower values. The BLAGNs appear to have also strong [O iii] lines. Most of the oAGN (mSF) present a strong [O iii] luminosity confirming the presence of nuclear activity (if we assume that [O iii] lines are always related to the AGN or nuclear shocks). As expected, the non-oAGN (mAGN) usually have weak [O iii] luminosity, like the SF and composites, with a few exceptions. WISE classifying them as AGNs probably has to do with other processes beyond that of a central AGN activity. Some interesting cases such as oAGN (mSF) with high [O iii] luminosity are discussed in Section 3.3.

Table 3 contains the detailed measurements used in this study, organized by their BPT-IR classification. W1, W2, and W3_PAH are rest-frame fluxes (using SED fitting; see Jarrett et al. 2017 for more details). The stellar emission has been subtracted in W3. W1 – W2 and W2 – W3 are the rest-frame colors used for the color–color diagram. [O iii]/Hβ and [N ii]/Hα represent the optical line ratios used for the BPT diagnostic. Stellar masses and SFRs (SFR_12μm) are derived using the calibration by Cluver et al. (2014) and Cluver et al. (2017), respectively (see Table 4). The specific SFR is the ratio between the SFR and the stellar mass. The galaxies have been classified according to their different AGN groups.

Table 2 gives the number of galaxies and their average parameter values in the subsets for different conditions applied to the sample. The number of BLAGNs included in each subset is also shown. "Condition A" presents only the infrared classification, while in "B" the optical classification is presented. Both diagnostics are combined in "C," giving a total number of 1154 classified galaxies. "Condition D" is where we reclassify BLAGNs based on the fact that they are clearly optical AGNs that were misclassified by the BPT diagram. Notably, the 2 BLAGNs found in the composite group were reclassified as oAGN (mSF) , the BLAGNs (3 galaxies) found in the non-oAGN (mWarm) group become oAGN (mWarm), and finally 11 non-oAGN (mAGN) become oAGN (mAGN). We consider the galaxies in the mWarm region to have AGN activity; therefore, in "D," both methods agree on the final classification of 84.4% and 8.2% of the galaxies as non-AGNs (SF + composite) and AGNs, respectively, and disagree on the classification of 7.4%. The redshift distribution of the different groups of galaxy is presented in the next section Figure 13.

Table 2.  Subgroups and Their Different Constraints Applied to the Sample

	Number	Broad-line AGN	Redshift	$\mathrm{Log}\,{M}_{\mathrm{stellar}}$	SFR12μm	Log (sSFR)
				(${M}_{\odot }$)	(M⊙yr−1)	(yr−1)
Groups	No. (%)	BLAGN	mean(med)	mean(med)	mean(med)	mean(med)
WISE/G23	Condition Aa
Mid-IR SF (mSF)	6133 (94.5%)	N/A	0.14 (0.13)	10.52 ± 0.003 (10.55)	5.44 ± 0.041 (3.57)	−10.0 ± 0.004 (−9.95)
Mid-IR"warm" (mWarm)	181 (2.8%)	N/A	0.19 (0.2)	10.52 ± 0.014 (10.6)	11.35 ± 0.425 (9.23)	−9.63 ± 0.019 (−9.61)
Mid-IR AGN (mAGN)	179 (2.8%)	N/A	0.19 (0.2)	10.6 ± 0.012 (10.71)	23.98 ± 1.126 (13.84)	−9.49 ± 0.017 (−9.52)
Optical/G23	Condition Bb
Optical SF (oSF)	881 (75.11%)	1	0.11 (0.09)	N/A	N/A	N/A
Optical (composite)	170 (14.49%)	15	0.15 (0.14)	N/A	N/A	N/A
Optical AGN (oAGN)	122 (10.4%)	56	0.18 (0.19)	N/A	N/A	N/A
WISE/Optical/G23	Condition Cc
SF	838 (72.6%)	0	0.10 (0.09)	10.11 ± 0.01 (10.13)	06.23 ± 0.13 (03.48)	−9.58 ± 0.01 (−9.52)
Composites	138 (12.0%)	2	0.14 (0.12)	10.39 ± 0.02 (10.41)	10.43 ± 0.48(07.28)	−9.56 ± 0.02 (−9.48)
oAGN (mAGN)	49 (4.2%)	30	0.20 (0.21)	10.90 ± 0.02 (10.93)	32.89 ± 2.67 (24.92)	−9.56 ± 0.03 (−9.58)
oAGN (mWarm)	33 (2.9%)	18	0.17 (0.18)	10.61 ± 0.03 (10.62)	14.84 ± 1.25 (10.86)	−9.58 ± 0.04 (−9.55)
oAGN (mSF)	37 (3.2%)	7	0.17 (0.18)	10.61 ± 0.03 (10.60)	09.33 ± 0.76 (08.82)	−9.78 ± 0.04 (−9.66)
Non-oAGN (mAGN)	23 (2.0%)	11	0.21 (0.21)	10.79 ± 0.03 (10.78)	28.14 ± 2.53 (22.92)	−9.43 ± 0.04 (−9.43)
Non-oAGN (mWarm)	36 (3.1%)	3	0.18 (0.19)	10.46 ± 0.03 (10.51)	16.79 ± 1.43 (13.11)	−9.41 ± 0.04 (−9.37)
WISE/Optical/G23	Condition Dd
SF	838 (72.6%)	0	0.10 (0.09)	10.11 ± 0.01 (10.13)	06.23 ± 0.13 (03.48)	−9.58 ± 0.01 (−9.52)
Composites	136 (11.8%)	0	0.14 (0.12)	10.39 ± 0.02 (10.41)	10.42 ± 0.49 (07.02)	−9.56 0.02 (−9.49)
oAGN (mAGN)	59 (5.1%)	41	0.20 (0.21)	10.89 ± 0.02 (10.91)	30.75 ± 2.23 (23.72)	−9.57 ± 0.03 (−9.58)
oAGN (mWarm)	36 (3.1%)	21	0.18 (0.18)	10.64 ± 0.03 (10.64)	15.82 ± 1.26 (11.81)	−9.58 ± 0.04 (−9.55)
oAGN (mSF)	39 (3.4%)	9	0.16 (0.18)	10.60 ± 0.03 (10.59)	09.41 ± 0.73 (09.17)	−9.76 ± 0.04 (−9.66)
Non-oAGN (mAGN)	13 (1.1%)	0	0.20 (0.20)	10.74 ± 0.04 (10.77)	34.18 ± 4.23 (24.68)	−9.29 ± 0.06 (−9.23)
Non-oAGN (mWarm)	33 (2.9%)	0	0.17 (0.19)	10.41 ± 0.03 (10.46)	15.90 ± 1.45 (12.64)	−9.39 ± 0.04 (−9.36)

Notes. The mean and median values of some properties have been added. Condition C was applied to Figure 9.

^aSignal-to-noise ratio (S/N): W1 > 5, W2 > 5, and W3 > 2; redshift <0.3; and magnitude (W1) < 15.5 mag (in Vega). ^bS/N (optical lines) > 3, redshift < 0.3, Hα > 0, Hβ > 0, [O iii]5007 > 0, [N ii]6583 > 0, and magnitude (W1) < 15.5 mag (in Vega). ^cS/N (optical lines) > 3, redshift < 0.3, Hα > 0, Hβ > 0, [O iii]5007 > 0, [N ii]6583 > 0. S/N: W1 > 5, W2 > 5, and W3 > 2; and magnitude (W1) < 15.5 mag (in Vega). ^dWhen in "Condition c," all broad-line (BL) non-oAGNs are considered to be BL oAGNs and BL composites to be BL oAGN(mSF). Recall that the position of the BLAGNs are highly uncertain on the BPT due to errors related to the [N ii]/Hα flux ratio.

Download table as:  ASCIITypeset image

In Figure 12, we plot SFR_12μm as a function of stellar mass and include the main-sequence relation of the local galaxies from Grootes et al. (2013), Cluver et al. (2020), Parkash et al. (2018), and Jarrett et al. (2017).

Zoom In Zoom Out Reset image size

The upper limits trace the galaxies with the lowest SFR corresponding to the passively evolving galaxies (also presented in Figure 4). All the groups share the same mass range from 10^9.8 M_⊙ < M_stellar < 10^11.5 M_⊙. The low-mass end (M_stellar < 10^9.8 M_⊙) is exclusively populated by SF galaxies having low star formation along with a few non-oAGN (mWarm). The high-mass end (M_stellar > 10¹¹ M_⊙) is populated by galaxies with broad Balmer lines (panel (b)), which are classified as AGNs, both in the optical and the mid-infrared. The linear fit to the selection of Jarrett et al. (2017), which is similar to our sample (except that here the AGNs have been separated from SFs), follows our sequence, especially the high-mass end, which is generally populated by AGNs. The new sequence fit by Cluver et al. (2020), based on a sample of isolated SF galaxies, also traces our distribution with a slightly flatter slope.

Grootes et al. (2013) UV-selected a sample of nearby (z < 0.13) spirals in GAMA that gives a flatter SFR–M_* relation. The relation given in Parkash et al. (2018) traces best our low-mass galaxies and also seems to follow the background distribution of galaxies (the gray points), which likely includes absorption-line features. Our sample is limited to only bright galaxies (W1 < 15.5 mag); moreover, it requires at least a detection in W3 (dusty), which might preferably select more massive, star-bursting systems, compared to UV or optically selected samples, as discussed in Jarrett et al. (2017). Figure 12(c) shows a flat distribution, with the selected galaxies having generally log sSFR > −11.4 yr⁻¹. The BPT based on emission-line galaxies might be preferentially selecting galaxies that are still building up their disk and not probing enough intermediate (so-called "green valley") galaxies. We note that we do expect a mass (and SFR_12μm) overestimation in galaxies having an AGN due to the additional mid-infrared flux coming from the active nucleus, rather than related to the star formation activity itself; this section is therefore intended to be illustrative.

The SF galaxies with the highest SFR_12μm (log SFR_12μm > 1.6 M_⊙ yr⁻¹) are all located at W2 – W3 > 4.6 mag and 0.22 mag < W1 – W2 < 0.7 mag. The majority of them (7/9, 77%) have W1 – W2 well above 0.43 mag, indicating extreme activity. Also, six in a total of nine have a peculiar morphological type (by the SED template and confirmed through visual inspection). This could be a hint that the star formation has been triggered by external processes such as mergers or the tidal influence of the environment. Their masses range from log (M_stellar/M_⊙) = 10.6 to 11. But on the other hand, the SF galaxies with the highest stellar mass (log M_stellar > 11 M_⊙) are SF disk galaxies in the WISE color–color diagram with 2.6 < W2 – W3 < 3.9 mag. They also have very low W1 – W2 color (<0.13 mag), except one galaxy (CATAID: 5256068; see Table 3), which is warmer (W1 – W2 ∼ 0.54 mag and W2 – W3 ∼ 4.5 mag). They are among the lowest sSFR with a mean log sSFR of −10.2 yr⁻¹.

We show the specific SFR_12μm (SFR/stellar mass) as a function of stellar mass in Figures 12(c) and (d). The non-oAGN (mAGN) and non-oAGN (mWarm) have the highest log sSFR, or building rate, on average, ∼ −9 yr⁻¹. In this case, the environmental influences such as tidal interactions or mergers could have triggered additional star formation. They are therefore warm enough to be classified as warm-AGNs in WISE.

The non-oAGN (mAGN) group with average log (sSFR) = −9.29 yr⁻¹ can be divided into two subsamples. The first is composed of galaxies having a very strong Hα line compared to [N ii] (≈60% of galaxies in this group), where the [O iii] and Hβ are almost nonexistent. The galaxy 5200866, which is in the starburst zone, has one of the highest sSFR (see Table 4). The second subgroup concerns BLAGNs that have been misclassified by the BPT diagnostic. Indeed, in some cases of BLAGNs, it is impossible to disentangle Hα and [N ii], such that their ratio is either under- or overestimated. The non-oAGN (mWarm) share similar properties to the first group. The third highest mean sSFR group is oAGN (mAGN), and the lowest sSFR are seen among the oAGN (mSF). It looks like the AGNs classified by WISE with almost no/weak [O iii] line exhibit higher sSFR compared to the galaxies classified as AGNs in the optical and SF with very strong [O iii] lines (Table 2 gives a brief summary of the mean values). The fact that the oAGN (mSF) galaxies have the lowest sSFR might be a hint of a quenching activity (AGN feedback) occurring inside the galaxies. Once more, the trend seen could be just related to the overestimation of the parameters due to the presence of the AGN itself. An alternative way of deriving the parameters (mass, SFR_12μm, sSFR), is with the AGN modeled and removed from the host emission (Assef et al. 2010; Hainline et al. 2014), which should provide a more robust separation of AGN and ISM dust emission.

Figure 13 has been added to show the repartition of the different groups of galaxies classified in the current study as a function of redshift. Our sample shows a high concentration of SF galaxies at low redshift that decreases significantly toward higher redshift. This is probably due to an observation bias where fewer galaxies are detected at higher redshift as they get fainter. It is important to recognize that the oAGN (mSF) and the oAGN (mWarm) are equally distributed from low to high redshifts as opposed to the non-oAGN (mAGN) generally found at higher redshift. Recall that the non-oAGN (mAGN) and the oAGN (mSF) groups represent galaxies for which the infrared and optical classification are contradictory.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Here we present the optical SF galaxies and composites both classified as SF in the mid-infrared (mid-IR). They are by far the most abundant type with a total (SF+composites) of ∼84% of the classification (see Table 2). They have the lowest averages of the stellar mass, partly due to the fact that they are young galaxies which are still in the growing process, but on the other hand, unlike the galaxies hosting AGNs whose fluxes are overestimated (as explained above), they have the most unaffected, reliable parameters (mass, SFR_12μm, and sSFR).

The galaxies 5241095 and 5240983 (Figures A1 and A2) are an interacting SF pair located at redshift z = 0.027. Both galaxies are quite similar in their spectra and location in the BPT and WISE color–color diagram. This galaxy pair is atypical in the sense that each of the galaxies shows the presence of a very high [O iii] line, which is stronger than the Hα and a quasi absence of the [N ii] line. The best-fit SED template classifies 5241095 as "Irr" and 5240983 as a normal spiral "Sc" galaxy. The KiDS image shows a clear interaction between the two galaxies. It appears as though 5240983 is pulling material from 5241095, which is being ripped apart. Although the parameters here have much smaller values than the averages in their group (SF group, Table 2), the high values of stellar mass, SFR_12μm, and sSFR for 5240983 (9.14, 0.49, and −9.45) in comparison to 5241095 (9.05,0.2, and −9.75) seem to corroborate this scenario. It could also be viewed as follows: galaxies get bigger with time through a series of mergers, which in turn trigger occasional AGN activity (looking at the unusually high [O iii] in this pair).

The galaxy 5306682 (Figure A3) is the perfect candidate (typical) to represent the SF group for having a very high Hα line flux followed by [N ii], [O iii], and Hβ with significantly smaller relative proportions. It is a good example that illustrates how the deep KiDS image helps to decipher and reveal the host and its environment. Figure A3 shows one of our closest specimens in the total sample (redshift z = 0.005). While in the WISE image the magenta circle seems to enclose the entire galaxy, KiDS reveals a far larger galaxy that goes beyond the dusty part presented by WISE (recall that the two images have the same angular size).

The galaxy 5205726 in Figure A4 is presented in the KiDS image as a galaxy with clear spiral arms, a bar, and a nuclear ring. It is a large galaxy ($\mathrm{log}\ {M}_{\mathrm{stellar}}$ = 10.51 M_⊙) with an sSFR (sSFR = 10^−10.22 yr⁻¹). It is classified as a composite in optical with an [O iii] luminosity of 10^40.29 erg s⁻¹.

In this category, we have galaxies classified as AGNs both in the optical (BPT) and the WISE color–color diagram (5.1% of the entire study sample).

The galaxy with GAMA ID 5339805 (Figure A5) is a typical case in the oAGN (mAGN) group. Its spectrum visually shows a broad Hα line with a greenish color in the WISE three-band image characteristic of mAGNs. The KiDS image reveals a spiral galaxy with a bulge (or pseudo bulge) and a thin disk. It is also confirmed as an AGN by the SED fit.

Although close to ∼70% of this sample is made up of BLAGNs, they are mostly located at higher redshifts with a mean value of 0.2 (see Figure 13) as opposed to the current galaxy, which is one of the nearest of its kind (redshift z = 0.089). Generally, [N ii] gets swallowed up by the broad Hα, which leads either to an underestimation/overestimation of [N ii]/Hα line ratios. Indeed, the BPT shows a [N ii]/Hα line ratio of >2, when in reality the [N ii] line is barely visible next to the Hα.

In the same group, GAMA ID 5151978 presented in Figure A6 is located at higher redshift with the same typical WISE AGN color. We can see a very strong [O iii] line compared to Hβ, which places it well above the AGN dividing line in the BPT. It is located at the edge of the obscured-AGN box. The [O iii] line luminosity >10⁴² erg s⁻¹ makes it a potential obscured quasar according to the criteria by Jarvis et al. (2019). Intriguingly, the SED template suggests a normal late-type galaxy, not an AGN. The KiDS image provides a hint about the issue. Although the nearby source does not seem to interact with our main galaxy, it is close enough for WISE not to be able to disentangle both sources; they are instead treated as a single source due to the large beam of WISE. The main galaxy is surely an AGN as shown by the spectrum and also the WISE color, but the presence of the second might add extra features that somehow affect the SED fitting. Overall, we classify this galaxy as an AGN as demonstrated by the diagnostics presented in the figure.

This category is made up of galaxies that are classified as AGN based on the optical BPT, but seen as normal SF or as just getting warmer in WISE color. We call galaxies in WISE for which W1 – W2 is greater than a certain threshold, but below the AGN zone (recall Figure 9 for the different zones), warm. One has to be careful during the classification for most of the misclassified cases are found in this category. Indeed, taking into account the data quality flag in GAMA and also visually inspecting the spectra reduce the number of galaxies here by 25% not to mention that without a k-correction and a correction for stellar absorption, the number of false positives would have been much greater. The oAGN(mWarm) and oAGN(mSF) groups seem to belong to the same family which evolves with redshift. Both groups have average redshifts of z = 0.16 and 0.18, respectively.

The oAGN(mWarm) group appears to be dominated by broad lines (21/36), which, unlike the general trend, are not able to raise W1 – W2 above the threshold required to be classified as infrared AGNs or QSOs. Indeed, WISE is very sensitive to the broad-line systems, and in most cases, such galaxies are classified as AGNs. Maybe a parallel process is at work in the host making it difficult for WISE to have a clear view of the ongoing AGN activity. Galaxy 5347780 in Figure A7 reveals broad Hα and Hβ lines, associated with a prominent [O iii] line, all strong indications of AGN activity. However, W1 – W2 is barely close to 0.6 mag, which falls in the "warm" AGN zone. The explanation of this behavior might lie in the SED characteristic of an SF galaxy (i.e., the host dominates the observed emission).

Galaxy 5294374 in Figure A8 tells a slightly different story. Here, in addition to the spectrum, the SED also presents the galaxy as an AGN, while the WISE color stills fall below the strong AGN threshold. The change, in this case, clearly a late-stage merger, WISE colors could be dominated by an SF galaxy (maybe the larger of the three in the merger) while AGNs could come from any one of these sources.

Galaxies 5249547 and 5155115 in Figures A9 and A10, respectively, are two examples of a clear disagreement between optical and mid-IR classifications. While the optical lines present them as AGNs without ambiguity, the two galaxies sit exactly on the WISE SF main sequence, Equation (1). Furthermore, the SED confirms star formation as the dominant activity. Generally, the galaxies in this category are relatively nearby, with the optical and the mid-IR giving evidence of AGN and SF activity, respectively. We could argue that the galaxies here have a strong-enough AGN to be picked up by the optical, but with continuum (and mid-IR) emission dominated by the star formation of the host. The narrow optical fibers may see only the center, while WISE integrates the flux over the entire galaxy. It could also be that the AGN has recently turned off or shut down. Generally, once the AGN turns off, the BLR, X-rays, and mid-IR shut down within a few decades, while the narrow-line region (NLR) could last thousands of years (Sartori et al. 2018). This might be a plausible justification for some narrow-line optical AGNs being classified as WISE star-forming galaxies. Note that as it can be seen in Figure A9 (5249547, log L[O iii] = 41.19 erg s⁻¹), a substantial number of the oAGN (mSF) present high luminosities (see Figure 11(b)) generally indicative of the presence of an AGN. The [O iii] lines in these cases could originate from other processes different from the AGN activity. Possible explanations for this scenario are proposed in the discussion, below.

We address in this subsection galaxies with no characteristic AGN emission lines in the optical, but classified as either an AGN or warm in the mid-infrared (the orange circles and squares in Figure 11), named non-oAGN(mAGN) and non-oAGN(mWarm), respectively.

In our study, the tension between the optical versus infrared classification is mainly caused by two factors. The first is due to tightly blended galaxies for which the WISE image alone cannot separate them but are clearly distinguishable in the deep KiDS r-band image. By comparing the KiDS versus WISE images in Figures A11, A12, and A15, we can see that multiple sources are mistakenly being considered as single (in WISE) owing to their angular proximity at the given redshift with mean redshift z = 0.2 (among the most distant galaxies of the study sample), while the optical (GAMA), which uses the KiDS galaxies location, is only targeting the central galaxy. The outcome of the classification in the mid-infrared will depend on the components of the merging system. A composition of several AGNs will probably lead to a higher W1 – W2 color while the opposite could instead lead to a decrease. The system presented in Figure A15 (CATAID: 5204947) is an extreme case where apparently more than three galaxies seem to be merging. As we can see from these three examples (i.e., Figures A11, A12, and A15), 5204947, which lies below the AGN region, is still the only one presented as an AGN by the SED.

The second cause of misclassification is related to the broad-line measurement. It has been noted several times that for broad Hα lines, it is sometimes challenging to distinguish between [N ii] and Hα to get the correct line ratios. In any case, broad Hα and Hβ lines are already enough for the galaxy to be classified as an AGN. Using the line ratio here leads to an underestimation of [N ii]/Hα, therefore classifying the galaxy as non-oAGN. This picture is seen in Figures A13 and A14 where the broad-line galaxies are correctly classified as AGNs both using the WISE color–color diagram and the SED fitting, though they lie in the composite area of the optical BPT.

The last case study, 5135645 (Figure A16), represents a blended system classified as an optical SF and infrared SF (SF), but located at the extreme limit between the AGN and SF zones. We suspect the second galaxy to be an AGN, justifying the high W1 – W2 (0.74 mag, very warm) or related to the fact that the interaction is triggering AGN activity. The extreme W2 – W3 > 5.04 mag is not unlike the class of HyperLirgs that WISE has recently uncovered (Tsai et al. 2015). The SED template fit using the famous starbust–AGN hybrid system NGC 3690 gives a glimpse of the probable internal processes taking place. These cases will be part of our next study, which will bring in radio data from SKA pathfinders ASKAP/EMU and MeerKAT already in the reduction phase. In order to better see the central regions, a zoomed KiDS r-band image of each one of the galaxies is presented in Figure A17.