Census of the Local Universe (CLU) Narrowband Survey. I. Galaxy Catalogs from Preliminary Fields

CLU-Hα searches for emission-line galaxies using four wavelength-adjacent, narrowband filters with a combined wavelength range of 6525–6878 Å. Each emission-line galaxy can be identified via a flux excess in one filter (the "On" filter), signifying the presence of an emission line compared to a filter that covers only the adjacent continuum (the "Off" filter). Thus CLU-Hα provides a distance constrained by the width of our narrowband filters.

The transmission curves for our narrowband filters are shown in Figure 1, where the solid lines represent the transmission from the manufacturer, the dotted line represents the CCD quantum efficiency of the camera, and the dashed-dotted line represents the atmospheric transmission at Palomar (Ofek et al. 2012).

Zoom In Zoom Out Reset image size

The filter widths and central wavelengths are calculated via

$$\begin{eqnarray}&&{\rm{\Delta }}\lambda \equiv \int T(\lambda )d\lambda ,\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&\bar{\lambda }\equiv \displaystyle \frac{\int \lambda T(\lambda )d\lambda }{\int T(\lambda )d\lambda },\end{eqnarray} \tag{ 2 }$$

where T is final transmittance of the filter curves with the peak normalized to unity. Table 1 presents the properties of our four narrowband filters.

The first filter is centered near the z = 0 Hα emission line (λ = 6563 Å), and the wavelength range of the last filter extends to the z = 0.047 (i.e., ∼200 Mpc) Hα emission line. The first two filters (Hα1 and Hα2) make up the first filter pair, and the third and fourth filters (Hα3 and Hα4) make up the second filter pair. We note that the small overlap between the filter pairs can result in decreased efficiency of emission line detection for emission lines whose wavelengths fall in this gap; however, the overlap is only 10 Å and is not likely to affect a large number of sources.

Observations were taken on the Oschin 48 inch telescope at the Palomar Observatory with a 7.92 deg² mosaic imager composed of 12 CCD detectors (or chips) and a 1'' per pixel image scale. However, one of the chips (chip #3) is nonfunctional, resulting in an 11-chip imager with a field of view 7.26 deg² (for details, see Law et al. 2009; Rau et al. 2009). The survey uses an observational strategy of three spatially staggered, overlapping grids that are limited to a decl. of greater than −20°, where each grid contains N = 3626 iPTF fields (26,470 deg² of the sky). Observations using the Hα1 and Hα2 filters cover the entire sky above −20° decl. Observations using the Hα3 and Hα4 filters also cover the sky above −20° decl., but avoid the Galactic plane ($| b| \gtrsim 3^\circ$). A single 60 s exposure is taken for each field in each grid, resulting in three images for the majority of positions on the sky. The multiple images facilitate cosmic ray rejection, filling in chip gaps and the nonfunctional chip, and deeper final co-added images.

Data acquisition ended in 2017 March (due to decommissioning of the iPTF camera), with 98.3% of fields observed in Hα1 and Hα2 and 91.3% observed in Hα3 and Hα4. However, the second filter pair is 95% completed at a Galactic latitude limit of 3° (95% at $| b| \gtrsim 3^\circ$) and 99% completed at latitudes above 11° (99% at $| b| \gtrsim 11^\circ$). Figures 2 and 3 present the sky coverage maps in equatorial coordinates (where 0° R.A. is represented as the center, vertical red line) of the CLU-Hα survey for the first filter pair (Hα1 and Hα2) and the second filter pair (Hα3 and Hα4), respectively. The colors for each iPTF pointing box in these figures represent the number of observations, where lighter/brighter colors indicate more observations. The Galactic plane is easily identified in Figure 3 as a dark strip of fields with little to no observations in the second filter pair. The majority of the fields within our targeted sky coverage have three observations. In addition to our survey, there were multiple iPTF projects that observed specific regions of the sky using the CLU narrowband filters and can provide even deeper Hα imaging for these fields. The locations that these programs targeted are apparent in Figures 2 and 3 as the brightest regions, with ∼100–200 observations.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Although the ultimate goal of the CLU-Hα survey is to coadd the images from the 3 staggered, overlapping grid patterns, the preliminary analysis here uses the single exposures from only 1 of these grids in 14  fields (see Section 3.1).

Data reduction and source extraction are carried out in an automated pipeline built by the Infrared Processing and Analysis Center (IPAC)⁷ specifically for the iPTF survey. The full description of this pipeline can be found in Laher et al. (2014), but we provide a brief overview here.

The IPAC reduction pipeline consists of both "off-the-shelf" and custom software that have been extensively tested on millions of images from iPTF broadband images. After each night of data is acquired, the IPAC pipeline performs a bias subtraction and applies a flat-field correction. The astrometric solution for each processed image is computed via SCAMP (Bertin 2006) on one of three stellar catalogs: SDSS-DR7 (Abazajian et al. 2009), UCAC3 (Zacharias et al. 2010), or USNO-B1 (Monet et al. 2003).

After data reduction is completed, Source Extractor (Bertin & Arnouts 1996) is used to generate a source catalog for every chip and filter image in each field. The fluxes are reported using many aperture definitions; however, we utilize the fluxes from aperture photometry at 5 pixels (∼5'') in diameter for galaxy candidate selection (see Section 3) and the point-spread function (PSF) fitted fluxes of stars for calibration purposes (discussed later). The 5-pixel aperture is used to select narrowband excess sources, as the Hα colors showed smaller scatter in this aperture size compared to larger 8 and 10 pixel diameter apertures. The median and standard deviation 5σ detection limits of a point source with a 5'' diameter aperture for each chip image in all 14 preliminary fields is 18.6 ± 0.30, 18.7 ± 0.28, 18.8 ± 0.28, 18.8 ± 0.25 AB mag for filters Hα1, Hα2, Hα3, and Hα4, respectively.

Calibration is carried out for each chip and filter image combination, where Pan-STARRS DR1 (PS1; Chambers et al. 2016) stars (N > 100) are matched with CLU-Hα sources. We match all CLU-Hα sources that (1) are relatively isolated (i.e., no source within 10''), (2) have a photometric error less than 0.1 mag, and (3) have no Source Extractor photometry flags to PS1 stars with those that (1) have a g − i color between 0 and 1.7 mag, (2) have a photometric error less than 0.1 mag, and (3) have a PSF minus Kron magnitude in the i band of less than 0.05 mag to select stars.⁸ The photometric zero points and color correction terms are determined by fitting a linear relationship between Hα − r (CLU −PS1) and g − i (PS1 − PS1) colors, where both CLU-Hα and PS1 magnitudes are PSF-fitted magnitudes.

Figure 4 graphically presents our calibration method, where the gray dots represent all CLU-Hα sources matched to PS1 sources, the blue dots represent matched "good" stars that meet our criteria, the dashed line represents the zero-point, and the solid cyan line represents the fit to "good" stars. The error in the fit is used the zero-point error and is added in quadrature to the photometric error of each source. Typical errors in the zero points are ∼0.05 mag. The fitted line in Figure 4 is used to perform a color correction to the magnitudes of the Hα sources.

Zoom In Zoom Out Reset image size

Due to the large field of view of the iPTF instrument, variations in the sensitivity across the imaging array exist after the images are processed with the IPAC image reduction pipeline and need to be taken into account when deriving the final calibrated magnitudes of extracted sources. This variation can be quantified in the zero-point of stars located across the entire field of view. The procedure for deriving a sensitivity variation map is described in detail by Ofek et al. (2012), but we provide a brief overview here. Using a minimum of 200 images for each chip and filter image, we measure the zero points for all stars that meet our calibration star criteria (see the previous paragraph), then combine them into bins of 32 pixels on a side. The median zero-point residual for each bin, normalized by the median zero-point across the entire image, is used as the zero-point variation correction. This correction is applied to the final calibrated magnitude for each source, given the position on each chip and the filter used in the observation. The zero-point variation across a single chip has a typical range of 0.01 to 0.05 mag, and the standard deviation of the zero-point variation is ∼0.02 mag for all filters.

We supplement our CLU-Hα fluxes with information from the SDSS DR12 (Alam et al. 2015), Galaxy Evolution Explorer (GALEX) all-sky (Martin et al. 2005), and Wide-field Infrared Survey Explorer (WISE) all-sky surveys (Wright et al. 2010), in addition to PS1 PSF and Kron magnitudes. We utilize the model ugriz magnitudes from SDSS DR12. In addition, we cross-match the CLU-Hα sources against entries in the SDSS "galSpecLine" table and extract the redshift, Hα line flux ("h_alpha_flux"), and EW ("h_alpha_eqw"). We also extract FUV and NUV kron fluxes from the GALEX all-sky imaging survey (Bianchi et al. 2014), and the instrumental profile-fit photometry of the first and fourth WISE bands ("w1mpro" and "w4mpro"). These ancillary data are used to cull contaminants and measure several physical properties of galaxies.

The preliminary fields were chosen based on the following criteria: (1) the CLU-Hα images must contain Hα filter pairs taken on the same night and have observations in all four filters to ensure a complete analysis, (2) the fields must have SDSS coverage to provide a list of galaxies with known redshifts, and (3) the decl. must be close to 30^◦ to facilitate spectroscopic follow-up from Palomar Observatory. The basic properties of the 14  preliminary fields are listed in Table 5.

In addition, we have chosen to include one field that contains a galaxy cluster. The field labeled "p3967" in Table 5 is spatially coincident with the Coma cluster whose redshift falls in the wavelength range of our third narrowband filter (Hα3). We include this field to robustly test the limits of our selection methods, as this cluster contains hundreds of galaxies whose Hα EWs span a large range from ellipticals with no emission lines to star bursts with strong emission lines (Mahajan et al. 2010). However, the majority of the preliminary fields occupy relatively sparse sections of the sky (i.e., outside the Galactic plane and not coincident with any galaxy clusters). Each pointing in our Hα survey covers 7.26 deg² on the sky; thus our 14  preliminary fields cover a total of 101.6 deg².

The identification of emission-line candidates follows the general methods of previous emission-line galaxy surveys using narrowband filters (e.g., Bunker et al. 1995; Pascual et al. 2001; Fujita et al. 2003; Geach et al. 2008; Sobral et al. 2009, 2012; Ly et al. 2010; Lee et al. 2012; Stroe & Sobral 2015). The photometry used for the selection process are taken from the Source Extractor catalogs. We use the 5-pixel diameter aperture photometry for all selection criteria. In addition, we require a candidate to have a 5σ detection in the "On" band and set non-detections to the 5σ upper limit (see Table 5).

Emission-line candidates are initially selected based on the significance of the excess in "On-Off" color quantified by the parameter Σ, where the flux is greater in one filter (the "On" filter) due to the presence of an emission line compared to the corresponding continuum filter (the "Off" filter). Each filter pair (Hα1/Hα2 and Hα3/Hα4) provides both the "On" and "Off" photometry. For example, Hα2 is used as the "Off" filter for Hα1 selection, while Hα1 is used as the "Off" filter for Hα2 selection.

The significance of the narrowband colors (Σ) is defined as the number of standard deviations between the color excess ("On-Off") in counts and the random scatter of counts expected for a source with zero color (Bunker et al. 1995). This can be expressed as

$$\begin{eqnarray}&&{\rm{\Sigma }}\,=\,\displaystyle \frac{{c}_{\mathrm{on}}-{c}_{\mathrm{off}}}{\delta },\end{eqnarray} \tag{ 3 }$$

where c_on,off are the counts for the "On, Off" filters and δ is the sky fluctuations in both images combined in quadrature. The combined photometric uncertainty is expressed as

$$\begin{eqnarray}&&\delta =\sqrt{\pi {r}^{2}({\sigma }_{\mathrm{on}}^{2}+{\sigma }_{\mathrm{off}}^{2})},\end{eqnarray} \tag{ 4 }$$

where σ_on/off represent the median sky count fluctuations in 100 sky regions for the "On" and "Off" images, and r is the radius of the photometric aperture. This criteria is similar to a standard signal-to-noise selection and can be expressed in terms of measured colors and magnitudes via

$$\begin{eqnarray}&&{m}_{\mathrm{off}}-{m}_{\mathrm{on}}=-2.5\mathrm{log}(1-{\rm{\Sigma }}\delta {10}^{-0.4(\mathrm{ZP}-{m}_{\mathrm{on}})}),\end{eqnarray} \tag{ 5 }$$

where m_on/off are the calibrated magnitudes in the "On" and "Off" filters and ZP is the photometric zero-point for the "On" filter.

The relationship between Σ, color, and magnitude is represented as curved lines in a color–magnitude diagram and graphically presented in Figures 5–8 for each Hα filter in one field ("p3967"). While all four of these figures are similar in form, we enlarge the first figure for clarity. In each of these figures, the black points represent all sources in the field, the red dots represent galaxies with a spectroscopic redshift (from SDSS) greater than the redshift coverage of our filter set (z > 0.047), the green Xs represent galaxies with a spectroscopic redshift (from SDSS) that fall in the wavelength range of the appropriate filter, and the green circles represent the galaxy candidates selected in each filter.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Inspection of Figures 5–8 also shows a non-zero color scatter, even at brighter magnitudes. Thus we impose a second condition based on the standard deviation of colors for bright continuum sources (i.e., stars) with magnitudes between 12 and 15 mag. The second criteria is represented as horizontal lines in Figures 5–8, illustrating the minimum color below which we cannot accurately infer the presence of an emission line. Furthermore, as the EW of an emission line is simply the ratio of the line flux and the continuum flux density, we can calculate an EW limit based on the "On-Off" color scatter in bright continuum sources. Following the prescription of (Stroe & Sobral 2015; see Equation (7)), the EW can be calculated via

$$\begin{eqnarray}&&\mathrm{EW}={\rm{\Delta }}{\lambda }_{\mathrm{on}}\left({10}^{-0.4{\rm{\Delta }}m}-1\right),\end{eqnarray} \tag{ 6 }$$

where Δλ_on is the FWHM of the "On" filter and Δm is the "On-Off" color. The EW cuts based on the "On-Off" colors for each filter in Figures 5–8 are labeled on the horizontal lines.

Both of the selection criteria adopted here (see Bunker et al. 1995; Fukugita et al. 2007; Ly et al. 2010; Lee et al. 2012; Sobral et al. 2009, 2012; Stroe & Sobral 2015) represent the number of standard deviations away from the expected random scatter and can be combined into a single Σ value. For each source, the number of standard deviations is computed above each criteria (bright stars and the noise in the images), and the final Σ is defined as the smaller of the two. For example, the Σ of a bright source (e.g., ${m}_{\mathrm{on}}$ ≲ 17 mag) will be calculated relative to the scatter in continuum sources, and the Σ of a faint source (e.g., m_on ≳ 17 mag) will be calculated relative the signal-to-noise curve.

We utilize two Σ cuts to define the extremes of our candidate selection methods. The first is a cut of Σ = 2.5, which will contain the majority of all star-forming galaxies but will contain a relatively higher contamination from galaxies with no emission lines in our narrowband filters. The second is a cut of Σ = 5, which will contain a reduced fraction of all star-forming galaxies but will contain a low contamination fraction (see Section 3.5). The median EW cut and standard deviation for all filters in the 14  preliminary fields is 7.3 ± 1.1 Å and 15.2 ± 2.4 Å for Σ = 2.5 and Σ = 5, respectively.

We provide an electronic version of the CLU candidates with Σ values above 2.5 (Table 2). In the future, we will generate source catalogs in larger areas of the sky, where Σ values will be calculated for each source and users can apply their own cuts based upon their science requirements. For a better understanding of the success and contamination rates at different Σ values, see Section 3.5.

The two color excess cuts introduced in the last section will effectively select any source that has a significant Hα "On-Off" color. However, the resulting galaxy candidates can still be contaminated by continuum sources with steep blue or red continuum slopes, high-redshift galaxies with an emission or absorption line whose wavelength has been redshifted into the wavelength range spanned by one of our filters, and by cosmic rays or chip defects (e.g., hot pixels, column defects, etc.).

Point sources with a steep blue or red continuum can be reduced by requiring that our candidates be spatially extended. As our survey is only sensitive to a distance of 200 Mpc and our angular resolution is limited to a FWHM ∼ 2'', we estimate that galaxies larger than 2 kpc at 200 Mpc will be extended in our data. Comparison to a statistically complete sample of star-forming galaxies in the local universe where the sample is dominated by dwarf galaxies (LVL; Kennicutt et al. 2008; Dale et al. 2009; Lee et al. 2011; Cook et al. 2014a) reveals that the semimajor axis histogram peaks between 2 and 4 kpc, suggesting that we will likely detect the majority of galaxies even at our furthest distance by requiring our galaxy candidates to be extended. Similar results are found for the size of Hα disks in galaxies out to z = 0.1 (Dale et al. 1999). We exclude point sources based upon the recommended PS1 star/galaxy separation using PSF minus Kron i-band magnitudes.

Unfortunately, the PS1 star-galaxy classifications often mislabel saturated stars as galaxies at r-band magnitudes brighter than 12–14 mag⁹ , leaving ∼300 candidates that are clearly stars. We can remove ∼60% of these contaminants via a bright Hα magnitude cut of 12 mag, given the brightest quoted saturation magnitude in PS1. We note that the brightest confirmed galaxy in the preliminary fields is ∼13.4 mag. The remaining contaminating stars are removed via visual classification.

The removal of cosmic rays and chip defects is achieved by requiring a source to be spatially coincident with a PS1 source. Since the PS1 detection limits (e.g., r ∼ 23.2 mag) are deeper than the CLU-Hα single-image exposures, any cosmic ray or chip defect with no spatial overlap with a PS1 source will be easy to flag and remove. Visual inspection of sources removed from our analysis via this method show that no real sources have been removed. Furthermore, visual inspection of our galaxy candidates show no contamination from cosmic rays, but do show a small percentage of contamination from chip-column defects that randomly coincide with a PS1 source; these contaminants are easily removed via visual inspection. We note that the removal of cosmic rays and chip defects will be greatly reduced or unnecessary for future analyses using stacked images.

The final sources of contamination are high-redshift galaxies whose emission or absorption lines have been shifted into the wavelength range of our filters. It is not possible to distinguish between strong emission lines redshifted into the wavelength range of our filters and Hα emission at lower redshift. We note that there is a dearth of strong emission lines blue-ward of our filters, where the closest lines are the [O iii]$\lambda 4959,\lambda 5007$ doublet near λ = 5000 Å (z ∼ 0.3). However, galaxies with extreme [O iii]$\lambda 4959,\lambda 5007$ emission at z ∼ 0.3 are likely to be interesting objects (i.e., green peas) and can be used for studies of galaxy evolution and star formation. In fact, we find a newly discovered z ∼ 0.3 green pea contaminant in the preliminary fields (see Section 5.1), and future CLU efforts will be aimed at finding and studying more of these objects.

Absorption lines in elliptical or red-sequence galaxies at intermediate redshifts are another source of contamination. Strong NaD absorption at redshifts of 0.11–0.17 can cause a lower flux in an Hα filter relative to the flux in its pair filter resulting in a significant "On-Off" color. Since red-sequence galaxies tend to be redder than star-forming galaxies (Salim et al. 2007), we can remove these contaminants via simple optical color cuts. Figure 9 shows the g − r versus NUV − r color–color diagram for galaxies in our preliminary fields with low and high Hα EWs, as measured via spectra from SDSS and CLU-Hα follow-up. We find that a g − r color cut of g − r greater than to 0.9 mag removes 60% of red-sequence galaxies at z > 0.047, while removing only four galaxies with moderate EW and a redshift less than 0.047.

Zoom In Zoom Out Reset image size

We have obtained spectra for 334 candidates in the preliminary fields with no redshift information (except obvious stars, cosmic rays, and chip defects) with a Σ value above 2.5. The spectra were taken with the 200 inch Hale telescope atop Palomar Mountain on multiple nights over 2016 and 2017 using the Double-Beam Spectrograph instrument (DBSP; Oke & Gunn 1982) and on the 2.3 meter Wyoming Infra-red Observatory (WIRO) with the Longslit spectrograph. The DBSP data were reduced using the PyRAF-based pipeline¹⁰ of Bellm & Sesar 2016, and the WIRO data were reduced with standard IRAF procedures.

We took spectra of 334 galaxies, where we confirm that 124 galaxies were indeed galaxies with redshifts less than 0.047, while the remainder were higher-redshift contaminants. In addition to finding new galaxies in the target volume, we also found two new galaxies (a Seyfert 1 galaxy and a green pea) at intermediate redshift via strong [O iii]$\lambda 4959,\lambda 5007$ emission lines redshifted into our filters (see Section 5.1).

In the rest of this section, we compare the photometric fluxes measured in our survey to with those derived from the spectroscopic measurements. In Figure 10, we plot the photometric Hα line flux versus the spectroscopic line flux in the left panel. The photometric Hα line flux is measured via

$$\begin{eqnarray}&&{F}_{\mathrm{line}}(\mathrm{erg}\,{{\rm{s}}}^{-1}\,{\mathrm{cm}}^{-2})={\rm{\Delta }}{\lambda }_{\mathrm{On}}({f}_{\mathrm{On}}-{f}_{\mathrm{Off}}),\end{eqnarray} \tag{ 7 }$$

where ${\rm{\Delta }}{\lambda }_{\mathrm{On}}$ is the FWHM of the "On" filter, f_On is the flux density of the "On" filter, and f_Off is the flux density of the "Off" filter. The flux densities of the "On" and "Off" filters are calculated via

$$\begin{eqnarray}&&{f}_{\mathrm{on},\mathrm{off}}(\mathrm{erg}\,{{\rm{s}}}^{-1}\,{\mathrm{cm}}^{-2}{\mathring{\rm A} }^{-1})=\displaystyle \frac{{\rm{c}}}{{\lambda }_{\mathrm{on},\mathrm{off}}^{2}}{10}^{-0.4({m}_{\mathrm{on},\mathrm{off}}+{\mathrm{Zpt}}_{\mathrm{AB}})},\end{eqnarray} \tag{ 8 }$$

where c is the speed of light, Zpt_AB = 48.59, and λ_on,off are the central wavelengths of the "On" and "Off" filters.

Zoom In Zoom Out Reset image size

There is reasonable agreement for the majority of galaxies above a few ×10⁻¹⁵ erg s⁻¹ cm⁻² for line fluxes derived from our Hα photometry and those derived from spectroscopy. However, there is a small offset (a few tenths of a dex) at lower line fluxes where the photometrically derived fluxes are systematically higher than the spectroscopic fluxes. This offset is likely due to a mismatch in the apertures used to derive the line fluxes, and from contaminating flux from [N ii] in the Hα imaging. We use a 5'' aperture in the Hα imaging, while the spectroscopic line fluxes are derived from either a 3'' fiber from SDSS or a 2'' slit from Palomar observations. In addition, the [N ii]/Hα ratio for star-forming galaxies in the local volume can range from 0.05 to 0.5 with an average of 0.2 (Kennicutt et al. 2008). These ratios translate into an average and max offset in Figure 10 of 0.1 and 0.3 dex, respectively, and are likely the major source of the offsets seen in Figure 10. We also note that the strength of the [N ii] contamination depends on the redshift of the galaxy where the [N ii] lines can be shifted out of the "On" filter if the Hα line is near the filter edge.

In this section, we evaluate our selection criteria by cross-matching our galaxy candidates to currently cataloged galaxies with spectroscopic redshifts available from NED, SDSS, or our CLU-Hα follow-up in the preliminary fields. Since our selection methods rely on the presence of a moderately strong Hα emission line, we expect that we should recover a large fraction of the Hα flux for galaxies in these fields, but recover a relatively low fraction by number of these galaxies due to the numerous elliptical galaxies in these fields (i.e., in the Coma cluster).

First, we examine the composition of galaxy catalogs produced when using different Σ cuts: 2.5, 3, 4, 5, 6, and 7. Table 3 shows the total number of galaxy candidates, contaminant galaxies (i.e., those with no emission line found in our narrowband filters), galaxies found in the volume of our survey (z < 0.047), galaxies with an emission line found in the correct filter (i.e., an emission line has been confirmed in the filter identified by our narrowband colors), and galaxies with new distance measurements (i.e., with no previous distance measurements) now confirmed in the local universe by our survey in the correct filter.

We find that the percentage of contaminant galaxies falls significantly above a Σ cut of 3, and that the percentage of galaxies with an emission line found in the correct filter is ≈90% above a Σ cut of 5. Thus we conclude that a Σ cut above 5 produces a high fidelity galaxy catalog. However, a Σ cut above 2.5 does produce a catalog with a higher number of galaxies with an emission line found in the correct filter, but with a high contamination percentage. In this and future publications, we will release galaxy candidates with Σ cuts above 2.5, and as a consequence we focus the text in the rest of this section and the paper on the galaxy catalogs from two extreme cuts (Σ > 2.5 and Σ > 5) but provide the statistics for all Σ cuts in the tables.

Next, we quantify the success rates of our selection methods both by number and by Hα flux in Table 4 via a comparison to galaxies with known redshifts in our survey volume and spatially located in our preliminary fields. The simplest comparison we can make is by number where we successfully recover 25.1% and 12.9% for Σ cuts of 2.5 and 5, respectively (column 2). However, there are 1226 known galaxies in our preliminary fields with known redshifts, where 722 have an Hα EW lower than 7.5 Å (≈60%). The majority of these galaxies are located in a single pointing that is coincident with the Coma cluster (i.e., "p3967"). Thus a low success rate by number is not surprising, given the large fraction of elliptical galaxies in our preliminary fields. If we limit the sample to galaxies with an Hα EW larger than the limits for the Σ = 2.5, we find that we recover 54.5% and 29.8% by number for Σ cuts of 2.5 and 5, respectively (column 4).

Since our selection methods are based on the strength of the Hα emission line, it is more appropriate to quantify our success rates via Hα flux. Column 3 of Table 4 shows the fraction of Hα flux captured by each of the Σ cuts, where we recover 82.5% and 67.8% for Σ cuts of 2.5 and 5, respectively. If we limit the sample of known galaxies to those with a moderate Hα EW (>7.5 Å), we recover 86.7% and 72.0% by Hα flux for Σ cuts of 2.5 and 5, respectively. The success rates of other Σ samples can be found in column 5 of Table 4.

Next, we examine why galaxies with an EW greater than 7.5 Å are not selected as candidates (i.e., our false-negative rate). The top panel of Figure 11 shows the Hα "On" magnitude histogram for these galaxies with Σ >= 2.5 and a Σ < 2.5, and we find that the galaxies missed in our survey that should be selected tend to be fainter. In these fainter galaxies, the significance of their narrowband colors is reduced by the increased photometric noise and thus these require a stronger EW line to be selected by our methods (see the bottom panel of Figure 11). We note that the photometric errors will be reduced in the final stacked CLU-Hα images, resulting in higher success rates in future analyses.

Zoom In Zoom Out Reset image size

In this section we explore the limits of our selection criteria and quantify them by Hα flux. We find that our survey is 90% complete at an Hα flux of 1 × 10⁻¹⁴ erg s⁻¹ cm⁻² and 4 × 10⁻¹⁴ erg s⁻¹ cm⁻² for the Σ = 2.5 and Σ = 5, respectively.

To quantify the limits of this survey, we present the Hα flux histograms for our confirmed galaxy candidates and all galaxies in the preliminary fields with spectroscopic information from either SDSS or our own follow-up. In the top panel of Figure 12, the unfilled, red-filled, and blue-filled histograms represent all galaxies with spectroscopy and our CLU-Hα galaxies with Σ > 2.5 and with Σ > 5. The completeness (defined as the fraction of all galaxies recovered in our survey in each line flux bin) is illustrated in the bottom panel of Figure 12. We find that we recover 90% and 50% of known galaxies at F_line = 1 ×10⁻¹⁴ and 3 × 10⁻¹⁵ erg s⁻¹ cm⁻² for Σ > 2.5 and F_line = 4 × 10⁻¹⁴ and 1 × 10⁻¹⁴ erg s⁻¹ cm⁻² for Σ > 5.

Zoom In Zoom Out Reset image size

Next, we explore the redshift-magnitude distribution of the two CLU-Hα galaxy catalogs in Figure 13 to illustrate the resulting CLU-Hα galaxy catalogs and their magnitudes. The top panel presents the full redshift range of our galaxy candidates, while the bottom panel is a zoom-in of the local universe (z < 0.05). In addition, the dashed- and solid-curved lines represent the relationship between an apparent magnitude limit of 17.8 (i.e., the limit of the SDSS spectroscopic galaxy survey) and 19 mag, respectively, with redshift.

Zoom In Zoom Out Reset image size

The density of all galaxies with spectroscopic redshifts drops off significantly below the r-band magnitude limit of the SDSS spectroscopic galaxy survey (r = 17.8 mag). In addition, the density of CLU-Hα galaxies also significantly decreases near that of SDSS, suggesting that our selection methods result in an effective magnitude limit similar to the SDSS limit. Our methods do not select bright galaxies with weak emission lines, but do find 69 CLU-Hα galaxies with r-band magnitudes fainter than 18 mag with large Hα EWs (a median value of 45 Å), illustrating that our selection methods can detect fainter galaxies with strong emission lines. We note that the "X" symbols in Figure 13 represent galaxies with no previous distance information in NED or SDSS.

Here we present examples of CLU-Hα galaxies with no previous distance information that are now well constrained to be in the local universe. Figure 14 presents a representative sample mosaic of these galaxies where the Hα emission line has been spectroscopically confirmed in the filter identified by our narrowband colors. The left image cutout in Figure 14 is the SDSS gri color composite, the four panels to the right are the cutouts for all four Hα filters, and the green boxes highlight which filter the emission line has been confirmed. The galaxies in this mosaic span a range of Hα EW, where EW increase toward the bottom of the figure from EW = 10 Å at the top to EW = 100 Å at the bottom. The galaxies also show a range of morphologies from compact to irregular to those showing spiral structure.

Zoom In Zoom Out Reset image size

Here we explore the observable properties of our confirmed galaxy candidates in comparison to cataloged galaxies in our preliminary fields with spectroscopic information from SDSS or our CLU-Hα follow-up. Figure 15 presents the SDSS g − r color, apparent r-band magnitude, the absolute r-band magnitude, and the Hα luminosity of all galaxies with spectroscopic information, a subset of these galaxies with Hα EW greater than 7.5 Å, and our CLU-Hα Σ = 2.5 galaxies. We do not show the observable properties for the Σ = 5 catalog, as their distributions are similar to the Σ = 2.5 catalog. The subsample of galaxies with an Hα EW greater than 7.5 Å represent those above our minimum Hα EW selection threshold for the Σ = 2.5 galaxy list. Thus these comparisons will reveal the properties of the galaxies with EW values above our limit that were not selected.

Zoom In Zoom Out Reset image size

Panel (a) of Figure 15 shows that the majority of galaxies in our preliminary fields have g − r colors that peak around 0.75 mag, largely due to the presence of early-type galaxies in the Coma field. However, both the subsample with Hα EW > 7.5Å and the CLU-Hα galaxies show a bluer peak around 0.45 mag. The different color distributions show that our selection methods tend to select blue, star-forming galaxies, as expected.

Panel (b) of Figure 15 shows the apparent r-band magnitude where we find median values of 16.5, 16.9, and 16.6 mag for galaxies in our preliminary fields with spectroscopy, the subsample with an Hα EW > 7.5 Å, and CLU-Hα galaxies, respectively. In addition, the distribution of the Hα EW > 7.5 Å subsample is peaked at fainter r-band magnitudes between 17 and 18 mag when compared to the CLU-Hα galaxies. This suggests that a larger fraction of galaxies with moderate Hα EWs that are missed by our selection methods have fainter apparent magnitudes with a median value at 17.2 mag. These fainter galaxies are not selected because the signal-to-noise criteria in our method is an increasing function of "On-Off" color toward fainter magnitudes. In other words, the random "On-Off" color scatter of sources due to the noise in the images is greater at fainter magnitudes and thus requires a larger Hα EW to be selected (see the signal-to-noise curves in Figures 5–8).

Panel (c) of Figure 15 shows the absolute M_r magnitude where we find median values of –18.7, –18.5, and –18.9 mag for galaxies in our preliminary fields with spectroscopy, the subsample with an Hα EW > 7.5 Å, and CLU-Hα galaxies, respectively. We find that the subset of galaxies with an Hα EW > 7.5 Å and not selected in our survey span a range of magnitudes, including both intrinsically faint and bright galaxies. These galaxies tend to have fainter apparent magnitudes due to their distances, where 75% are at distances greater than 100 Mpc with a median r-band magnitude of 17.2 mag. Thus these missed galaxies with moderate Hα EWs would require stronger Hα emission to be selected in our survey.

Panel (d) of Figure 15 shows the Hα luminosity, where we find median Log L_Hα (erg s⁻¹ cm²) values of 39.1, 39.7, and 39.9 for galaxies in our preliminary fields with spectroscopy, the subsample with an Hα EW > 7.5 Å, and CLU-Hα galaxies, respectively. This panel clearly illustrates that a large fraction (>50%) of galaxies with spectroscopy in our preliminary fields have weak Hα emission (i.e., small Hα EWs). We note that the early-type galaxies with Hα line non-detections are not plotted in this panel. A comparison between the subset of galaxies with EW > 7.5 and our CLU-Hα galaxies shows that the CLU-Hα galaxies have a larger median value by 0.2 dex. Thus our methods tend to select galaxies with the intrinsically higher Hα luminosities. We note that the moderate EW galaxies missed by our selection methods tend to be fainter with a median r-band magnitude of 17.1 mag, and would require larger Hα emission to be selected.

Here we present the physical properties of the galaxy sample by cross-matching against the ALLWISE 3.4 μm and 22 μm fluxes. These data will provide stellar masses (M_⋆) and extinction corrections due to dust. In addition, we use the spectroscopic Hα fluxes of our survey to derive SFRs. The Hα fluxes have been corrected for Milky Way extinction via the prescription of Schlafly & Finkbeiner (2011).

The stellar masses are derived from mass-to-light ratios (ϒ_⋆) using the WISE 3.4 μm fluxes. We utilize the fluxes derived from ALLWISE catalog profile fitting photometry; thus these fluxes should encompass each galaxies' full radial extent. The WISE 3.4 μm bandpasses provides a robust tracer of a galaxy's stellar mass, as this light is dominated by an older stellar population (which make up the majority of a galaxy's stellar mass) and is less affected by attenuation from dust than shorter wavelengths.

Many studies over the past few years have made comparisons between a variety of observationally derived stellar masses (e.g., baryonic Tully–Fisher relationship and resolved star color–magnitude diagrams) and luminosities in Spitzer 3.6 μm and WISE 3.4 μm bandpasses (Oh et al. 2008; Eskew et al. 2012; Barnes et al. 2014; Meidt et al. 2014; McGaugh & Schombert 2014, 2015; Norris et al. 2014; Querejeta et al. 2015). The results of these comparisons show that a constant Spitzer ${{\rm{\Upsilon }}}_{\star }^{3.6\mu m}$ of 0.4–0.55 M_⊙/L${}_{\odot }$ provides a robust estimation of a galaxies stellar mass with a relatively low error of ∼0.1 dex (Meidt et al. 2014; McGaugh & Schombert 2015). A constant mass-to-light ratio of the same value for WISE 3.4 μm has been shown to yield similar stellar masses as Spitzer 3.6 μm (Jarrett et al. 2013; Norris et al. 2014). We adopt the constant mass-to-light ratio of ${{\rm{\Upsilon }}}_{\star }^{3.4\,\mu m}\,=0.5\,$M_⊙/L_⊙,3.4μm, where M_⊙/L_⊙,3.4μm is the mass-to-light ratio in units of solar masses per the solar luminosity in the WISE 3.4 μm filter bandpass (m_{⊙,3.4 μm} = 3.24 mag; ${L}_{\odot ,3.4\mu m}\,=1.58\times {10}^{32}$ erg s⁻¹; Jarrett et al. 2013).

SFRs for our galaxies can be estimated from many different luminosity tracers (e.g., Hα, FUV, etc.), where measured luminosities are transformed into SFRs via scaling prescriptions (e.g., Kennicutt 1998; Murphy et al. 2011). We utilize the updated scaling relationships of Murphy et al. (2011), which assumes a Kroupa IMF (Kroupa 2001). The Hα SFRs are derived via a combination of Hα and 22 μm luminosities, which account for internal dust extinction (Calzetti et al. 2010; Murphy et al. 2011),

$$\begin{eqnarray}&&{\mathrm{SFR}}_{{\rm{H}}\alpha ,\mathrm{corr}}({M}_{\odot }{\mathrm{yr}}^{-1})={\rm{C}}\times \nu {L}_{{\rm{H}}\alpha }+0.031\times \nu {L}_{22\mu {\rm{m}}},\end{eqnarray} \tag{ 9 }$$

where νL are the observed monochromatic luminosities of both Hα and the WISE4 band at 22 μm in ergs per second and C = 5.37 × 10⁻⁴² (Murphy et al. 2011).

The two panels of Figure 16 show the Hα-derived SFR and stellar mass distributions for the same samples as in Figure 15. Panel (a) shows the SFR histograms, where we find median Log SFR (M_⊙ yr⁻¹) values of –1.0, –0.7, and –0.5 for the galaxies in our preliminary fields with spectroscopy, the subsample with an Hα EW > 7.5 Å, and CLU-Hα galaxies, respectively. Our selection methods tend to select galaxies with higher SFRs. However, we are able to recover some galaxies with moderately low SFRs near Log SFR (M_⊙ yr⁻¹) ∼ −2.

Zoom In Zoom Out Reset image size

Panel (b) of Figure 16 shows the stellar mass histograms where we find median Log M_⋆ (M_⊙) values of 9.5, 9.4, and 9.6 for the galaxies in our preliminary fields with spectroscopy, the subsample with an Hα EW > 7.5 Å, and CLU-Hα galaxies, respectively. The median values suggest that our methods select galaxies of roughly the same masses as other galaxies in the preliminary fields. We also find that the CLU-Hα galaxies span a range of stellar mass (8 ≲ Log M_⋆(M_⊙) to ≲11), suggesting that CLU-Hα will recover some lower-mass dwarfs as well as larger spirals. However, CLU-Hα will miss a significant fraction of dwarfs at greater distances.

There have been several blind emission-line and UV-excess surveys undertaken prior to CLU-Hα, including the Markarian survey (Markarian et al. 1981 and references therein), the Case Northern Sky survey (Pesch & Sanduleak 1983), the KPNO International Spectroscopic survey (KISS; Salzer et al. 2000), the University of Michigan survey (UM; MacAlpine et al. 1977, 1977a, 1977b; MacAlpine & Lewis 1978; MacAlpine & Williams 1981), and the Hamburg/SAO survey for emission-line galaxies (HSS; Ugryumov et al. 1999). Here we compare our galaxy catalogs (Σ > 2.5 and Σ > 5) to the objects in the first three surveys, as there is no overlap between our preliminary fields and both UM and HSS.

The Markarian (i.e., the first Byurakan; Markarian & Stepanian 1983; Markarian et al. 1989; Petrosian et al. 2007) survey used an objective prism and photographic plates to find a total of 1544 UV-excess galaxies via a visual search for steep UV continuum slopes in their low-dispersion spectra. This was the first systematic search for active galaxies and covered 17,000 square degrees down to a continuum brightness of 17.5 mag in the V band. The Case survey used a combination of steep UV continuum slopes and emission line selection ([O iii]) to find more galaxies per area, resulting in 2339 galaxies in ∼1200 square degrees down to 18 mag. Finally, the KISS survey used an objective prism and line selection (Hα) but implemented modern CCD detectors to survey fainter sources (20–21 mag) and found 2425 galaxies in ∼200 square degrees. The use of CCDs and line selection methods facilitated deeper surveys and the detection of galaxies with a wider range of properties because the Hα line (compared to UV slope and [O iii] line selection) will be stronger in more massive galaxies.

The CLU-Hα survey utilizes CCD detectors and narrowband imaging for low resolution line selection (Hα) to increase the survey area (3π for the full survey) and to ensure the detection of galaxies with a wide range of properties. Using single-image exposures in preliminary fields, we found 258 confirmed galaxies with Σ > 2.5 in 100 deg² and can detect the continuum of sources down to 18.5 mag. Thus the CLU survey depth and galaxy density are in between previous surveys that used photographic plates (Markarian and Case) and modern CCDs (KISS).

Figure 17 shows the sky regions covered by the CLU-Hα preliminary fields presented in this paper and each of the three comparison surveys with spatial overlap: Markarian (Petrosian et al. 2007), Case (Pesch & Sanduleak 1983, 1986, 1988, 1989; Sanduleak & Pesch 1984, 1987, 1989, 1990; Pesch et al. 1991), and KISS (Salzer et al. 2000, 2001; Gronwall et al. 2004; Jangren et al. 2005). In addition, the bottom panel of Figure 17 is a zoomed-in version covering our preliminary fields in more detail between an R.A. of 180° and 270°. We first spatially cross-matched CLU-Hα to the galaxies in each of the catalogs that are spatially coincident with our preliminary fields (within 4'') and have limited these catalogs to galaxies with a redshift below 0.047 (i.e., redshift probed by our survey). We also note that one of our CCD chips (#3) has been nonfunctional for all of our pointings since the beginning of the survey. In addition, the two lower-left chips for the field "p3967" (center near R.A. = 195, decl. = 28) in the Hα4 filter had poor data quality and could not be reduced. We remove galaxies that overlap with chip 3 in all pointings and the poor data chips for "p3967" in this comparison. The number of galaxies from the three comparison catalogs with a redshift less than 0.047 and inside our preliminary fields is 8, 18, and 80 for the Markarian, Case, and KISS surveys, respectively.

Zoom In Zoom Out Reset image size

A comparison with the Markarian galaxies shows that we successfully recover 7 of the 8 galaxies (88%). However, CLU-Hα found an additional 247 objects in the overlapping volume. The one Markarian galaxy not selected in CLU-Hα has a star within 1'' of the galaxy center and was mislabeled as a star in our catalog. We note that this galaxy still exhibits a Σ value of 15.

A comparison with the Case galaxies shows that we successfully recover 14 of the 18 galaxies (78%). However, CLU-Hα found an additional 30 objects in the overlapping survey regions. Two of the Case galaxies missed by CLU-Hα have small EW values below 2 Å. The third galaxy missed by CLU-Hα is an extended galaxy whose central region shows a small Σ value; however, we did find several of this galaxy's H ii  regions with high Σ values. The last galaxy missed by CLU-Hα shows two nearby H ii  regions where our photometric aperture is located between two clumps and thus is missing some of the Hα flux.

A comparison with the KISS galaxies shows that we successfully recover 42 of the 80 galaxies (53%). However, CLU-Hα found an additional 24 objects in the overlapping survey regions. Nearly a third of the KISS galaxies missed by CLU-Hα (N = 12) were below our detection limits of 18.5 mag. The remaining KISS galaxies were missed by CLU for the following reasons: 8 had low Hα EWs less than 7.5 Å, 6 were labeled as stars, and 12 had moderate Hα EWs combined with fainter magnitudes between 17.5 and 18.5 mag. The 12 missed galaxies with moderate EWs were cut in our survey, as their fainter magnitudes required higher EWs to exhibit a significant narrowband color.

The CLU-Hα survey is able to recover the majority of the galaxies in both the Markarian and Case surveys, and was able to find additional galaxies. Our comparison with the KISS survey shows that CLU-Hα recovers roughly half as many emission-line galaxies as KISS due to our brighter limits. However, CLU-Hα will cover a much larger area of the sky (3π sr) compared to all previous emission-line galaxy surveys to a moderate depth. Thus the CLU-Hα survey will discover many emission-line galaxies across the sky.

Among our emission-line candidates we find interesting extreme objects, some of which have no previous distance information. In Figure 18 we present a mosaic of four interesting candidates, where panel (a) shows the known planetary nebula, panel (b) shows one of the BCDs, panel (c) shows the green pea, and panel (d) shows Seyfert 1 galaxy. Due to the large "On-Off" magnitudes of the planetary nebula, the central pixels in the top panel of Figure 18 in the first Hα filter were masked to provide a more consistent background and visual comparison across the four Hα images.

Zoom In Zoom Out Reset image size

The planetary nebula is PN H 4-1 and is located at a high Galactic latitude: b = 8814757, l = 493065 (Haro 1951). Panel (a) of Figure 18 shows the SDSS gri color and our four Hα filters (from left-to-right, respectively). PN H 4-1 shows a large flux excess in Hα1 compared to Hα2 (i.e., "On-Off") equal to three magnitudes (Δ mag = −3.05). The measured spectroscopic Hα line flux and EW are 1.21 × 10⁻¹² erg s⁻¹ cm⁻² and 1200 Å, respectively. The spectrum of this PN is shown in top panel of Figure 19 and exhibits emission lines similar to other known planetary nebulae.

Zoom In Zoom Out Reset image size

The detection of this planetary nebula is interesting, as our choice of preliminary fields avoided the Galactic plane and PN H 4-1 is located at a high Galactic latitude (b ∼ 88°). We anticipate that the CLU-Hα survey can be used as a discovery engine for new planetary nebulae at intermediate galactic latitudes. This expectation is based on the galactic latitude limits of previous emission-line galaxy and Galactic Plane surveys in the northern sky. Previous galaxy surveys (Markarian, Case, UM, KISS, HSS, etc.) avoided the galactic plane above $| b| \gt 20^\circ$, while the largest-area galactic plane survey in the northern hemisphere was limited to $| b| \lt 5^\circ$ (the INT Photometric Hα survey, IPHAS; Drew et al. 2005). Thus there is a gap at intermediate galactic latitudes that has not been uniformly searched for emission-line sources. In addition, the large-area galactic plane search for planetary nebulae in the southern hemisphere (the Macquarie/AAO/Strasbourg Hα, MASH; Parker et al. 2006) found a few hundred newly discovered planetary nebulae at galactic latitudes of $5\,\lt | b| \lt 15$ degrees (≈4000 deg²) with a detection limit of 5 × 10⁻¹⁶erg s⁻¹ cm⁻². Despite the higher selection limit of our survey (1 × 10⁻¹⁴erg s⁻¹ cm⁻²), we anticipate finding 10s-to-100 new PNe in the area not covered by IPHAS and MASH (≈3500 deg²) at intermediate galactic latitudes, assuming the same distribution of PNe as in the southern hemisphere.

In the CLU-Hα sample, there are nine BCDs, where seven have distances measured for the first time in this survey. Panel (b) of Figure 18 shows the image cutouts of an example BCD with new distance measurements, where the source is brightest in the fourth Hα filter. The second panel of Figure 19 shows the spectrum of the BCD where the redshifted wavelength of the strong Hα line confirms the identification in the fourth filter. We note that the Hα emission line has a wavelength that is only a few angstroms greater than the separation between the third and fourth filter, and is clearly brighter in the fourth filter image. The spectrum of the example BCD is representative of the other BCDs, where both the [O iii]$\lambda 4959,\lambda 5007$ and Hα lines exhibit strong emission lines: EWs of 418 Å and 446 Å for [O iii]$\lambda 4959,\lambda 5007$ and Hα, respectively. We measure the 12+Log(O/H) metallicity of 8.02 from O3N2 methods (Pettini & Pagel 2004) and a dust-corrected SFR of 0.41 M_⊙ yr⁻¹.

The SDSS and CLU-Hα images of the green pea are shown in panel (c) of Figure 18. The spectrum of this object is shown in the third panel of Figure 19, where the [O iii]$\lambda 4959,\lambda 5007$ lines confirm the correct identification in the Hα3 filter. Both the [O iii]$\lambda 4959,\lambda 5007$ and Hα lines exhibit strong emission lines: EWs of 510 Å and 560 Å for [O iii]$\lambda 4959,\lambda 5007$ and Hα, respectively. We have measured metallicity of 12+log(O/H) = 8.09 via the O3N2 method. The Hα flux of 9. × 10⁻¹⁵ erg s⁻¹ cm⁻² corresponds to a dust-corrected SFR of 24 M_⊙ yr⁻¹, given the measured redshift of 0.337 and H₀ of 72 km s⁻¹ Mpc⁻¹.

The image cutouts of the new Seyfert 1 galaxy found in our survey of preliminary fields is shown in panel (d) of Figure 18, where the object is brightest in the third Hα filter. In addition, the spectrum of this object is shown in the bottom panel of Figure 19, which shows strong [O iii]$\lambda 4959,\lambda 5007$ lines confirming the correct identification in our third filter. Furthermore, the broadened Hα and Hβ emission lines clearly indicate the presence of a strong AGN, which allows us to classify this object as a Seyfert 1 galaxy.

In this section, we show where the CLU-Hα galaxies are located on established trends of star-forming galaxies. Previous studies of star-forming galaxies have found a relatively tight correlation between the current SFR and the total stellar mass for both local universe and higher-redshift galaxy samples: the galaxy "Main Sequence." This trend can provide insights into how galaxies evolve over time, since the normalization increases with redshift (e.g., Heinis et al. 2014). We use this relationship to illustrate the distribution of CLU-Hα galaxy properties.

Figure 20 shows the dust-corrected Hα-SFR versus stellar mass (M_⋆) for the CLU-Hα galaxies, where the symbols are the same as those in Figure 13. The blue and green lines represent the relationships found in the LVL sample (D < 11 Mpc; Cook et al. 2014b) and SDSS sample (z ∼ 0.1; Peng et al. 2010). The majority of the CLU-Hα galaxies follow the previous "Main Sequence" relationship and span a wide range in both SFR and stellar mass from low-mass, low-SFR dwarfs to high-mass, high-SFR spirals.

Zoom In Zoom Out Reset image size

The BCDs found in the CLU-Hα sample show high SFRs, given their stellar masses illustrating that these galaxies are undergoing an episode of enhanced star formation. The new green pea, on the other hand, shows good agreement with the SDSS "Main Sequence"relationship, suggesting that this galaxy is not undergoing an enhanced period of star formation; this is contradictory to previous studies, which find that green peas show signs of significant star formation activity (Cardamone et al. 2009; Izotov et al. 2011). However, we do not have enough statistics in this study to draw meaningful conclusions as to the connection between BCDs and green peas. Such an analysis is left for a future CLU-Hα study that will collect a larger sample of both types of objects.

We also explore a relationship between physical and observed properties of galaxies, where local universe studies have found that the specific SFR (sSFR ≡ SFR/M_⋆) forms a relatively tight trend with optical colors (e.g., Cook et al. 2014b; Schawinski et al. 2014). Figure 21 shows the sSFR versus the SDSS g − r color of the CLU-Hα galaxies, where the symbols are the same as those in Figure 20. We find that the CLU-Hα galaxies follow a similar trend to the LVL galaxies (Cook et al. 2014b), but that several CLU-Hα BCDs tend to have sSFRs (Log(sSFR) as high as −8.7) and bluest optical colors due to the large volume sampled. The CLU-Hα green pea galaxy shows a sSFR and color similar to that of the other CLU-Hα galaxies. A future CLU-Hα study will include larger numbers of BCDs and green pea galaxies to study this relationship.

Zoom In Zoom Out Reset image size

The detection of GWs via the LIGO (Abbott et al. 2016) allows for a new way of observing the universe and thus provides new tools with which to understand fundamental physics. Associating GWs and EM counterparts will have dramatic impacts on our understanding of the universe. However, the size of the 90% sky localizations of LIGO GW events have ranged from 30 to 1000 square degrees on the sky, making the search for EM counterparts a challenge.

Significant efforts of large-area surveys (e.g., Kasliwal et al. 2016, 2017; Coulter et al. 2017; Evans et al. 2017; Hallinan et al. 2017; Troja et al. 2017) have scanned the most probable sections of gravitational wave localizations in search of an EM counterpart. However, the efficiency of this effort can be greatly enhanced by utilizing the locations of known galaxies in the LIGO sensitivity volume for neutron star mergers (D ≤ 200 Mpc; Aasi et al. 2015). Targeted follow-up observations of likely host galaxies can narrow down the search area by a factor of 100 (Nissanke et al. 2013; Gehrels et al. 2016).

Previous efforts have been made to provide lists of galaxies specifically designed for EMGW follow-up (Kopparapu et al. 2008; White et al. 2011), where the latest catalog from White et al. 2011 shows a B-band luminosity completeness of 65% at 100 Mpc. However, the galaxies in both previous catalogs were limited to 100 Mpc, which is an eighth of the LIGO sensitivity volume out to 200 Mpc. In addition, several new or updated spectroscopic galaxy surveys have published secure distances via redshifts for new galaxies in the local universe (e.g., SDSS, ALFALFA, 6dFRGS; Jones et al. 2009; Haynes et al. 2011; Alam et al. 2015), thus leaving previous EMGW galaxy catalogs less complete than previously estimated. There are efforts that utilize increased numbers of galaxies with photometric redshifts (e.g., GLADE¹¹ ); however, we focus on galaxy catalogs with secure distance measurements. The lack of an EMGW catalog that extends to the full LIGO sensitivity volume and the existence of new galaxies found in the local universe motivate the construction of a new compiled galaxy catalog to be used in the search for EM counterparts to GW events.

In an effort to provide the most complete list of galaxies with measured distances in the LIGO sensitivity volume, our team has compiled a catalog of all known galaxies out to 200 Mpc, hereafter referred to as CLU-compiled. The galaxies were taken from existing galaxy databases: NASA/IPAC Extragalactic Database (NED),¹² Hyperleda¹³ (Makarov et al. 2014), Extragalactic Distance Database¹⁴ (EDD; Tully et al. 2009), the Sloan Digital Sky Survey DR12 (SDSS; Alam et al. 2015), the 2dF Galaxy Redshift Survey (6dFRGS; Jones et al. 2009), and The Arecibo Legacy Fast ALFA (ALFALFA Haynes et al. 2011). Distances based on Tully–Fischer methods were favored over kinematic (i.e., redshift) distances; however, the majority of the distances are based upon redshift information. The catalog contains ∼234,500 galaxies with existing distances less than 200 Mpc.

In addition to distances, the catalog also contains compiled photometric information. We have cross-matched (within a 4'' separation) the CLU-compiled catalog with GALEX all-sky (Martin et al. 2005), WISE all-sky (Wright et al. 2010), and SDSS DR12 (Alam et al. 2015) surveys to obtain fluxes from the ultraviolet (UV) to the infrared (IR). We find ∼154,200 matches for GALEX FUV, ∼216,000 for WISE 3.4 and 22 μm, and ∼100,400 for SDSS r band. With the UV and IR fluxes, we have also derived physical properties (SFRs, stellar masses, and dust extinction) for the CLU-compiled galaxies via similar methods as the CLU-Hα survey in Section 4.3.

The CLU galaxy catalog is being utilized by the GROWTH (Global Relay of Observatories Watching Transients Happen¹⁵ ) collaboration to search for EM counterparts to GW events across the EM spectrum. On 2017 August 17th, GWs from two merging neutron stars were observed in both the LIGO and Virgo detectors (hereafter GW170817; Abbott et al. 2017), and this marks the first GW event to have an EM counterpart found (Abbott et al. 2017). Shortly after the GW event, several collaborations around the world began a large multi-wavelength observational campaign to search for the EM counterpart (e.g., Coulter et al. 2017; Evans et al. 2017; Hallinan et al. 2017; Kasliwal et al. 2017; Troja et al. 2017). Our team cross-matched the CLU-compiled catalog against the LIGO-Virgo volume (initially 31 square degrees on the sky at a distance of 40 ± 8 Mpc; Abbott et al. 2017) and found 49 galaxies (see Kasliwal et al. 2017). In addition, since the physical properties of these galaxies were determined in advance by our team, we were able to promptly prioritize these galaxies by stellar mass and publish this list on the Gamma-ray Coordinate Network (GCN; Cook et al. 2017). The EM counterpart was later found in NGC 4993 (Coulter et al. 2017), which was the third highest priority galaxy on our published list. We note that the CLU galaxy catalog produced the only published GCN to correctly identify NGC 4993 in the localization volume of GW170817. Future efforts to search for the EM counterparts to GW events using CLU will include new galaxies found in the CLU-Hα survey.

Next, we examine the basic properties of our CLU-compiled catalog and investigate how the new galaxies from the CLU-Hα survey will contribute to the full CLU catalog. Figure 22 shows the observable and physical property histograms of both the CLU-compiled galaxies (open histogram) and CLU-Hα galaxies in the preliminary fields (blue-filled histogram), where all histograms have been normalized to the peak.

Zoom In Zoom Out Reset image size

Panel (a) shows the SDSS g − r color histograms, where the CLU-compiled sample shows a double peak while the CLU-Hα sample shows a single peak that overlaps with the CLU-compiled blue peak. The double peak is due to the dichotomy between blue star-forming galaxies and red early-type galaxies with little to no star formation. The absence of a red peak in the CLU-Hα sample shows that the CLU-Hα survey is not sensitive to early-type galaxies. This can also be seen in panel (c), which shows the FUV-derived SFRs, where the CLU-Hα galaxies are peaked at 0.3 dex higher than that of CLU-compiled. Thus the galaxies added by the CLU-Hα survey will be biased toward bluer galaxies with recent star formation.

Panel (b) shows the absolute magnitude M_r histogram, where both CLU-compiled and CLU-Hα samples show two peaks: one for intrinsically bright galaxies (M_r ∼ −19 mag) and one for fainter galaxies (M_r ∼ −14 mag). The reduced number of low luminosity CLU-Hα galaxies suggests that the CLU-Hα survey will not probe the same number densities of dwarfs found in the volume out to 200 Mpc. There is also a deficit of intrinsically bright CLU-Hα galaxies (M_r ≲ −20 mag) compared to CLU-compiled. This can also be seen in panel (d), which shows the stellar mass histograms, where the CLU-compiled sample peaks at 0.5 dex higher in M_⋆. An examination of the morphologies of the galaxies brighter than –20 mag shows that 60% have RC3 T-types less than 1 (e.g., S0 and earlier-type galaxies). Thus, the CLU-Hα survey will not add massive early-type galaxies to existing catalogs. However, the majority of these luminous galaxies are already in CLU-compiled, given the inclusion of SDSS galaxy spectra.

We will add CLU-Hα galaxies found in our narrowband survey with no previous distance information to our CLU-compiled catalog to produce a more complete CLU. This combined catalog will then be utilized to focus the search for future EM counterparts to gravitational wave events.

Future improvements to maximize the completeness and minimize the contamination of the CLU-Hα catalog include (i) stacking all Hα images to produce a deeper Hα source catalog and a resulting deeper galaxy catalog (these data will not only provide deeper images from which to find more galaxies, but will also facilitate removal of cosmic rays and chip defects); (ii) extended source photometry; (iii) machine learning for star-galaxy separation; and (iv) image subtraction of "On" and "Off" filters.

In addition, efforts to use machine learning for candidate selection are currently underway. The preliminary CLU-Hα survey fields are critical to our machine-learning efforts, as they establish a training set of objects with known redshifts. We have run preliminary tests on how well our machine-learning algorithms perform the tasks of classifying objects as simply within 200 Mpc and classifying objects as within one of our narrowband filters. Preliminary results are promising, showing that we can increase our completeness by using our current metrics along with additional information (e.g., Pan-STARRS, WISE, GALEX magnitudes) as features in our machine-learning algorithms (C. Zhang et al. 2018, in preparation).