SDSS-IV MaNGA: pyPipe3D Analysis Release for 10,000 Galaxies

The analysis is performed using a new implementation of the Pipe3D pipeline (Sánchez et al. 2016a), fully transcribed to python (pyPipe3D; Lacerda et al. 2022). This version of the code uses similar algorithms and the same analysis sequence as the previous version, adapted to make use of the unique computational capabilities of the new adopted coding language, improving its performance, and correcting bugs when needed. Pipe3D has been extensively used to analyze IFS data of very different natures, in particular data from the CALIFA (e.g., Cano-Díaz et al. 2016; Espinosa-Ponce et al. 2020), MaNGA (e.g., Ibarra-Medel et al. 2016; Barrera-Ballesteros et al. 2018; Sánchez-Menguiano et al. 2019; Sánchez et al. 2019a; Bluck et al. 2020), and SAMI surveys (e.g., Sánchez et al. 2019b), as well as individual (Sánchez et al. 2015) and large MUSE data sets (Sánchez-Menguiano et al. 2018; López-Cobá et al. 2020). It has been described in detail in previous articles (e.g., Sánchez et al. 2016a, 2016b, 2021), and thoroughly tested using both ad hoc simulations and mock data sets based on hydrodynamical simulations (Guidi et al. 2018; Ibarra-Medel et al. 2019). To avoid unnecessary repetition, we here provide a very brief summary, emphasizing the few novelties of the new code.

In summary, the code separates the stellar component and the ionized gas line emission in each spectrum, by fitting the former with a combination of simple stellar population (SSP) spectra. Prior to this decomposition, the SSP spectra are shifted to account for the stellar velocity (vel_⋆) and convolved with a Gaussian function to account for the velocity dispersion (σ_⋆). Finally, the SSPs are dust attenuated, adopting the Cardelli et al. (1989) extinction law. The parameters defining the kinematics (vel_⋆ and σ_⋆) and the intrinsic dust extinction (A_V,⋆) are estimated using a limited set of SSP templates, which restrict the spaces of the parameters to avoid or limit the intrinsic degeneracies between these parameters and the intrinsic properties of stellar populations (e.g., metallicity/dispersion; Sánchez-Blázquez et al. 2011). This two-step procedure and its benefits are described in detail in Sánchez et al. (2016a) and Lacerda et al. (2022). The stellar decomposition provides the light fraction or weight (w_⋆,L ) contributed by each SSP to the problem spectrum at a fixed spectral range (5450–5550 Å, similar to the V-band central wavelength), in addition to the vel_⋆, σ_⋆, and A_V,⋆ parameters obtained in the first step. Each observed spectrum (S_obs) is fitted to a model spectrum of the stellar component (S_mod), following the expression

$$\begin{eqnarray}&&\begin{array}{l}{S}_{\mathrm{obs}}(\lambda )\approx {S}_{\mathrm{mod}}(\lambda )\\ \ \ =\ \left[{{\rm{\Sigma }}}_{\mathrm{ssp}}{w}_{\mathrm{ssp},\star ,L}{S}_{\mathrm{ssp}}(\lambda )\right]{10}^{-0.4\ {A}_{{\rm{V}},\star }\ E(\lambda )}* G({\mathrm{vel}}_{\star },{\sigma }_{\star }),\end{array}\end{eqnarray} \tag{ 1 }$$

where S_ssp is the spectrum of each SSP in the template library, E(λ) is the adopted extinction law, and G(vel_⋆, σ_⋆) is the Gaussian function describing the line-of-sight stellar velocity distribution.

After subtracting the best stellar population model from the observed spectrum, the fitting algorithm models the ionized gas emission with a set of individual Gaussian functions that are fitted to a set of predefined emission lines at known wavelengths, shifted by the observed velocity of the gas component. In this way, we derive the flux intensity (F_el), velocity (vel_el), and velocity dispersion (σ_el) for each emission line (el). The procedure is repeated iteratively, with the emission-line analysis being performed after each step in which the stellar population spectrum is modeled and subtracted from the original spectrum. The figure of merit used to define the best-fitting model (a reduced χ²) takes into account both the stellar and emission-line models provided by the full analysis just described. In summary, our procedure involves a combination of brute-force exploration of the nonlinear parameters within a limited range (vel_⋆, σ_⋆, and A_V,⋆ for the stellar population, and vel and σ for each emission line), together with a pure linear inversion, to derive the weights (w_ssp) of the decomposition of the stellar population, plus an iterative automatic selection of the SSP templates to be included in the decomposition, based on the results of the previous fit.

The procedure just outlined is then applied to each data cube. However, it is not applied spaxel by spaxel. The results from simulations indicate that a minimum signal-to-noise ratio (S/N) is required for the fit to the stellar population to provide reliable results (Sánchez et al. 2016b). We thus perform a spatial binning, known as continuum segmentation (CS; described in Sánchez et al. 2016a), which groups adjacent spaxels, whose S/Ns are below a defined threshold, to produce a higher-S/N average spectrum. In this procedure, contrary to other ones found in the literature (e.g., Cappellari & Copin 2003), we do not group the spectra of adjacent spaxels if their intensities differ by more than a predefined percentage. In this way, the spatial shape of the original galaxy is preserved (to some extent), while the S/N is increased (although it does not always reach our goal value). From this analysis, we recover the parameters of the stellar population for each spatial bin (tessella or voxel), i.e., the weights of the decomposition of the SSPs (see Section 5.1.2 below), the kinematic parameters, and the dust extinction, together with the best spectral model. The model spectrum for each tesella is then scaled to the flux intensity of each spaxel, by using the so-called dezonification parameter (DZ; Cid Fernandes et al. 2013), which is the ratio between the flux intensity of each spaxel with respect to the average value in the tesella in which this spaxel was grouped. Restricting this procedure to model the stellar spectrum, we obtain a model cube of the stellar population. Subtracting this model from the original cube, we obtain the so-called pure-GAS cube, a cube comprising only the information of the ionized gas emission lines (plus noise and residuals from the imperfect subtraction of the stellar component).

The pure-GAS cube is then analyzed using two different procedures. In the first one, a limited set of strong emission lines are fitted with individual Gaussian functions for each individual spaxel, providing a set of maps with the flux velocity and the velocity dispersion of each fitted emission line, and their corresponding errors (Section 5.1.4). In a second step, a much larger set of emission lines is analyzed using weighted moment analysis, deriving maps of the three parameters indicated before (F_el, vel_el, and σ_el), together with the spatial distribution of the equivalent width (EW_el) of the considered emission line, el (Section 5.1.5). Two main differences have been introduced into the analysis of the emission lines with respect to previous analyses of MaNGA data using Pipe3D (Sánchez et al. 2018). (i) The Gaussian fitting is now performed in two steps. The first step replicates the brute-force exploration of the range of parameters described in Sánchez et al. (2016b). This exploration avoids falling into a local minimum. However, it is not very accurate. A second step is introduced by executing a Levenberg–Marquardt minimization algorithm, which uses the results of the first exploration as a guess. This additional fitting significantly increases the accuracy and precision of the results (Lacerda et al. 2022). (ii) A new and larger set of emission lines, with improved definitions of wavelength, has been adopted for the weighted momentum analysis, as will be explained below. The former version of the emission-line treatment has been performed as well, for a simpler comparison with previous results.

Finally, the pipeline obtains the spatial distribution (map) of a set of stellar spectral indices (Section 5.1.3). To do this, we subtract the best model for the emission lines from the spectrum of each tessella, in order to generate a stellar spectrum without contamination by the ionized gas contribution. Then, for each stellar spectral index, we derive its equivalent width by defining a wavelength range at which we measure the median flux intensity, and two adjacent wavelength ranges at which the continuum is estimated. We adopt the definitions of the different stellar indices and the procedures described in Cardiel et al. (2003). We note that we depart from the classical definition of D4000, derived using the flux intensity in units of frequency (i.e., F_ν ; Bruzual 1983), and adopt the more convenient functional form proposed by Gorgas et al. (1999; their Equation (2)). As in the previous cases, the pipeline provides maps of the spectral index and its estimated error.

The errors provided by Pipe3D are based on a set of Monte Carlo (MC) iterations, in which the original spectrum is perturbed within the errors (a needed input in the preprocessed data cubes; see Section 3). Every time a stellar or an emission-line model is subtracted from the original spectrum, the uncertainties in the model are also propagated into the errors. The final errors thus include both the noise level and the uncertainties in the modeling of the galaxy component. More details of this procedure are given in Lacerda et al. (2022).

One of the major changes with respect to the previous versions of the delivered data products, besides the use of an improved and transcribed version of the code, is the use of a new SSP spectral library in our analysis. For SDSS DR14 and DR15, we delivered versions v2_1_2 and v2_4_3 of the MaNGA data products, produced by adopting the GSD156 SSP library (Cid Fernandes et al. 2013). This library resulted from the combination of two sets of SSP model spectra. For stellar populations older than 65 Myr, the GSD156 library uses synthetic spectra from Vazdekis et al. (2010) and Falcón-Barroso et al. (2011), based on the MILES empirical stellar library (Sánchez-Blázquez et al. 2006). For stellar populations younger than 65 Myr (not included in the cited models), GSD156 uses synthetic spectra from the GRANADA library (González Delgado et al. 2005; Martins et al. 2005), based on theoretical stellar spectra. We adopted the Salpeter (1955) Initial Mass Function (IMF) for stellar masses between 0.1 and 100 M_⊙. GSD156 comprises 156 SSP spectra, sampling 39 ages from 1 Myr to 14 Gyr (on a near logarithmic scale), and four metallicities (Z/Z_⊙ = 0.2, 0.4, 1, and 1.5).

For the current implementation of the code, we adopt a completely different library, which we call MaStar_sLOG. This SSP library uses the recently delivered MaNGA stellar library (MaStar; Yan et al. 2019), which includes 8646 spectra for 3321 unique stars. This is a considerable increase in the number of stars and range of sampled atmospheric properties over the previously adopted stellar library (MILES comprises a total of ∼1000 stars). In Appendix A, we present a summary of the updated set of the Bruzual & Charlot (2003; hereafter,BC03) stellar population synthesis models that use several stellar libraries, including MaStar. The full MaStar SSP library comprises a total of 3520 individual spectra, covering 220 ages and 16 metallicities (assuming a solar [α/Fe] ratio and a solar metallicity of [Fe/H] = 0.017). This library is far too impractical to be applied to a stellar decomposition technique like the one performed by pyFIT3D, outlined in Section 4.1. Thus, we experiment with different combinations of SSPs extracted from this full library, adopting linear, logarithmic, and mixed samplings of both age and metallicity, comparing the results with previous ones and with theoretical expectations. In all cases, we repeat the full analysis described in this article for a subset of the full MaNGA data set, comprising 9500 data cubes (the so-called MPL-10 internal release). A detailed description of the experiments and their results will be presented elsewhere (S. F. Sánchez 2022, in preparation). In summary, we conclude that a sampling in age and metallicity, in which the steps between consecutive values increases are multiplicative (i.e., a pseudologarithmic sampling), results in a good compromise between: (i) an adequate sampling of the spectral properties; (ii) an efficient exploration (in terms of computing time); and (iii) the estimation of accurate and precise results. We must recall that, based on previous experiments (Sánchez et al. 2016b), an arbitrary increase of the sampling of the stellar parameters by an SSP library is not feasible for data with limited S/Ns. The finally adopted MaStar_sLOG library was built following this scheme and comprises 273 SSP spectra, sampling 39 ages from 1 Myr to 13.5 Gyr, and seven metallicities (Z = 0.0001, 0.0005, 0.002, 0.008, 0.017, 0.03, and 0.04) or (Z/Z_⊙ = 0.006, 0.029, 0.118, 0.471, 1, 1.764, and 2.353), as indicated in Tables 4 and 8.

Morphology is known to be a fundamental property that affects (and is affected by) galaxy evolution, being directly connected with the dynamical stage, the stellar content, the star formation stage, the gas content, and even the presence of active galactic nuclei. Therefore, it is essential to have a morphological classification of the galaxies as a basis for comparing with the properties derived by our analysis. The catalogs available publicly, which are based on visual classifications of galaxy morphology, only include the galaxies in the releases up to DR15, comprising ∼4700 objects, less than half the galaxies available in the final sample. For the purposes of this article, we do not need a highly accurate determination of the morphology of each individual galaxy. It is enough to have a consistent classification in terms of statistical properties. As such, we use the accurate visual morphology determinations from the SDSS VAC by Vázquez-Mata et al. (2022), which includes ∼6000 galaxies and is an extension of the publicly available SDSS VAC, ¹¹ to obtain training and testing samples for a machine-learning algorithm to classify the rest of the galaxies in the full sample. We select features that are expected to be informative of morphological class, namely: the Sérsic index, the NSA stellar mass, the line-of-sight velocity-to-velocity dispersion ratio at the effective radius, the ellipticity, the concentration index, and the $u^{\prime} g^{\prime} r^{\prime} i^{\prime} z^{\prime}$ colors (extracted from the NSA catalog). For such a task, we implemented a Gradient Tree Boosting algorithm, which is a type of ensemble method that successively trains a predefined number of decision trees, each improving upon its predecessor errors. We found the resulting classification to be satisfactory for our purposes: the deviation between the estimated morphology and that of the VAC for the testing sample is consistent with the deviation observed between this VAC and other morphological classifications (discussed below). We will detail this method in a forthcoming publication (A. Mejía-Narváez et al. 2022, in preparation).

We estimate a set of photometric and structural properties directly from the MaNGA data cubes in addition to the different parameters derived by pyPipe3D. Among them, we compute the broadband photometry in the Gunn u, g, r, and i and the Johnson B, V, and R filters, adopting the Vega photometric system and using the filter parameters provided by Fukugita et al. (1995), redshifted to the rest frame of each object. Thus, no additional corrections, such as K or E corrections, have to be considered. From this photometry, we derive each galaxy's observed and absolute magnitudes, again using the redshift to estimate the cosmological distance. ¹² In addition, we estimate the radii R50 and R90, which include, respectively, 50% and 90% of the integrated flux in the V band inside the MaNGA FOV, and the concentration index R90/R50. In the literature, there are several estimates of similar properties derived for the MaNGA galaxies using direct imaging, like the ones provided by SDSS (Blanton et al. 2017; Fischer et al. 2019) or the DESI survey (Arora et al. 2021). However, they present some disadvantages and obvious intrinsic differences: (i) none of them can derive a direct rest-frame estimate of the photometry, requiring a K correction based on the modeling of a spectral energy distribution to obtain them; (ii) the same limitation affects the structural properties, such as R50 or the concentration index; and (iii) these photometric values are usually derived for the full (optical) extension of the galaxies, and not limited to the FOV of the IFS data. This is very useful for deriving the global integrated galaxy properties (e.g., stellar mass or absolute magnitude). However, they are not appropriate for comparison with aperture-limited quantities, as derived by the pyPipe3D analysis.

From our photometry, we estimate each galaxy's stellar mass from its M/L ratio, assuming the relation

$$\begin{eqnarray}&&\mathrm{log}({M}_{\star ,\mathrm{phot}}/{M}_{\odot })=-0.95+1.58({\text{}}B\,-\,{\text{}}V)+0.43\ast (4.82-{V}_{\mathrm{abs}})\end{eqnarray} \tag{ 2 }$$

of Bell & de Jong (2000), valid for the Chabrier (2003) IMF, using (ϒ_B−V ) as the B − V color and our V-band absolute magnitude. These stellar-mass estimates are considered to be less accurate, but most probably more precise, than estimates based on the spectroscopic analysis performed by pyPipe3D. We acknowledge that there are more recent ϒ_color derivations (e.g., Zibetti et al. 2009; Taylor et al. 2011; Zhang et al. 2017; García-Benito et al. 2019), some of them based on IFS data, and more precise estimates of the stellar mass that take into account multiband photometry (e.g., Arora et al. 2021). However, we adopt Equation (2) for easy comparison with previous calculations (Sánchez et al. 2016a). Nevertheless, since we provide colors and absolute magnitudes, it should be easy to define other ϒ_color stellar-mass estimators.

Finally, we derive each galaxy's V-band surface brightness at its effective radii R50 and Re, encircling half of the light within the FOV and half of the total integrated light of the galaxy, respectively, by averaging the flux values in an elliptical ring (using the known position angle and ellipticity of the object) of width 0.15 (in units of Re or R50). Again, the surface brightness is measured in the galaxy rest frame, which is not directly accessible when using broadband imaging. As mentioned above, parameter errors are estimated from MC iterations, perturbing the observed spectra using the error spectra included in the MaNGA data cubes.

The analysis by pyPipe3D runs in an automatic way through the entire MaNGA data set, without human supervision. Therefore, we must perform a quality control on the results, to identify and flag any possible issues relating to either the data or the analysis itself. This procedure comprises different steps. First, an automatic exploration is done to check that the analysis has produced all the required files, and that they are of the required format with the expected values (i.e., no map or table is filled with NaN, Inf, or zeros). Then, a second automatic exploration is done, by comparing two basic parameters derived from our analysis (the redshift and the integrated stellar mass) with those provided by the NSA catalog. If significant differences (≳30%) are found in any of these quantities, the corresponding data cube is flagged with a "WARNING."

In addition, a more detailed by-eye exploration is performed to determine the quality of the data. In doing so, we created an interactive web page on which we can show for each galaxy/data cube: (i) its central spectrum, corresponding to an aperture of 25, together with the best-fitting model provided by pyPipe3D; (ii) the mass assembly and chemical enrichment curves (e.g., Pérez et al. 2013; Ibarra-Medel et al. 2016; Camps-Fariña et al. 2021); (iii) three true color images generated using (a) the original SDSS u-, g-, and r-band images; (b) the same image generated using synthetic broadband images extracted from the original MaNGA data cubes for the same filters; and (c) the emission-line maps for [O iii]5007 (blue), Hα (green), and [N ii]6584 (red); (iv) maps of the luminosity-weighted (LW) age (${{ \mathcal A }}_{\star ,L}$), metallicity (${{ \mathcal Z }}_{\star ,L}$), and dust extinction, derived as part of the pyPipe3D process (described in detail in Section 5.1.1); and (v) a set of spatially resolved emission-line diagnostic diagrams, as described in López-Cobá et al. (2020), including (a) the WHAN diagram (Cid Fernandes et al. 2010); (b) the classical Baldwin–Phillips–Terlevich (BPT; Baldwin et al. 1981) diagrams involving the [O iii]/Hβ line ratio as a function of the [N ii]/Hα, [S ii]/Hα, and [O i]/Hα ones; (c) the spatial distribution of the emission-line velocity dispersion, the [N ii]/Hα ratio, and the distribution of one as a function of the other; and (d) the gas and stellar velocity maps and their difference.

Figures 2, 3, and 28 (Appendix C) illustrate an example of the quality control analysis for a showcase data cube (manga-7495-12703). We search for: (i) evident problems in the quality of the explored spectra or issues in the fitting process; (ii) issues with the observations themselves (large masked areas and/or strong contamination by foreground stars); and (iii) the presence of strong AGNs or clear merging systems that may affect the stellar population and/or the kinematic analysis (just to warn the user). The visual quality control was performed by seven different people, who explored a subsample of the full data set, with an average overlapping of three different inspections per target/data cube. When there was a different appreciation of the quality by different inspectors, we adopted the worst-reported quality control flag.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The reported quality control flags (QCFLAG) are listed in Table 1, indicating the level of importance ("OK," "BAD," or "WARNING"). Of the 10,245 cubes corresponding to galaxies analyzed by pyPipe3D, the code fails to analyze 25 cases, due to different issues with the data (very low S/Ns in most cases and evident empty fields). Of the remaining 10,220 cubes, in six cases we flagged our results as BAD (for the reasons described before), meaning that they should be discarded. In addition, 386 cubes had different issues, and we strongly recommend not using them without paying attention to the visual inspection of the original spectra, the results from the fitting, and, in particular, the nature of the warning. In summary, we provide good-quality analyses for 9828 data cubes (corresponding to 9784 individual galaxies).

The results from the analysis described above for each individual spectrum are a set of parameters derived for each particular spaxel. Each of these parameters is stored as a 2D array or map of the given quantity, which preserves the World Coordinate System (WCS) of the original data. These maps are the the final products of the analysis. For an optimal (compact and easily accessible) distribution of the products, we pack them in data cubes, i.e., 3D arrays in which each channel (slide on the z-axis) corresponds to one of the maps described before and, therefore, to one of the derived parameters. Instead of storing all the products in a single data cube, they are packed into a set of data cubes based on a physically motivated association of parameters. In the current implementation of Pipe3D, there are five data cubes: (i) SSP, which contains maps of the spatial distribution of the main properties of the stellar populations; (ii) SFH (star formation history), which comprises the distribution of the light fraction derived by the stellar decomposition for each SSP in our template library; (iii) INDICES, which comprises the maps of the stellar indices; (iv) ELINES, which includes the properties of a set of strong emission lines derived by the Gaussian fitting procedure; and (v) FLUX_ELINES, which includes the properties of a much larger set of emission lines, based on the weighted moment analysis. In the current release, we distribute two versions of the ELINE analysis, based on two different sets of emission lines, with the new version labeled FLUX_ELINES_LONG. Finally, all the data cubes are stored as individual extensions of a single FITs file (the Pipe3D file). In this release, we include, in addition to these six extensions: (i) a primary extension with the header of the original MaNGA cube (ORG_HDR); (ii) a mask of the brightest field stars recovered from the Gaia survey (GAIA_MASK); and (iii) a mask of the spaxels within the hexagon with a high enough S/N to provide reliable estimates of the stellar population properties (SELECT_REG).

Table 2 summarizes the structure of the Pipe3D FITs file. Each file can be easily associated with its corresponding MaNGA cube from the file naming convention, i.e., manga-plate-ifudsg.Pipe3D.cube.fits, where plate and ifudsg correspond, respectively, to the number of the physical plate used during the pointing and the IFU bundle design and number. These two numbers uniquely define a fixed pointing to a certain galaxy, as indicated before. In this way, manga-7495-12704.Pipe3D.cube.fits corresponds to the MaNGA observation of a particular galaxy, observed using plate number 7495, and the fourth IFU (04), comprising 127 fibers. It is important to note that one galaxy may be observed using different plates and IFUs, corresponding to repeated observations that have been treated or analyzed individually, as if they were different objects. In the final MaNGA distribution, there are 90 of these repeated galaxies (i.e., 2 × 45 cubes). Since they are different observations, taken under different atmospheric conditions, and calibrated using different stars, they offer a unique opportunity to evaluate the quality of the data. We discuss them in detail in Appendix E.

5.1.1. SSP Extension

As indicated before, this extension comprises a data cube with the average properties of the stellar populations derived from the multi-SSP fitting procedure outlined in Section 4.1. The content of each channel with index N is described in the header entry named DESC_N, starting with N = 0 for the first channel, corresponding to the properties listed in Table 3. An example of the content of this extension, for data cube manga-7495-12704, is shown in Figure 4.

Zoom In Zoom Out Reset image size

Table 3. Description of the SSP Extension

Channel	Units	Stellar Index Map
0	10−16 erg s−1 cm−2	Unbinned flux intensity at ∼5500 Å, fV
1	None	Continuum segmentation index, CS
2	None	Dezonification parameter, DZ
3	10−16 erg s−1 cm−2	Binned flux intensity at ∼5500 Å, fV,CS
4	10−16 erg s−1 cm−2	StdDev of the flux at ∼5500 Å, efV,CS
5	log10(yr)	LW age, ${{ \mathcal A }}_{\star ,L}$, (log scale)
6	log10(yr)	MW age, ${{ \mathcal A }}_{\star ,M}$, (log scale)
7	log10(yr)	Error of both ${{ \mathcal A }}_{\star }$, (log scale)
8	dex	LW metallicity, Z⋆,L
		In logarithmic scale, normalized to
		the solar value (Z⊙ = 0.017)
9	dex	LW metallicity, Z⋆,M
		In logarithmic scale, normalized to
		the solar value (Z⊙ = 0.017)
10	dex	Error of both Z⋆
11	mag	Dust extinction of the st. pop., AV,⋆
12	mag	Error of AV,⋆, ${{\rm{e}}}_{{{\rm{A}}}_{V}}$
13	km s−1	Velocity of the st. pop., vel⋆
14	km s−1	Error of the velocity, evel
15	km s−1	Velocity dispersion of the st. pop., σ⋆
16	km s−1	Error of σ⋆, eσ
17	log10(M⊙ /L⊙)	Mass-to-light ratio of the st. pop., ϒ⋆
18	log10(M⊙ /sp2)	Stellar-mass density per spaxel., Σ⋆
19	log10(M⊙ /sp2)	Dust-corrected Σ⋆, Σ⋆,dust
20	log10(M⊙ /sp2)	Error of Σ⋆

Note. "Channel" indicates the z-axis of the data cube, starting from 0.

Download table as:  ASCIITypeset image

The first channel comprises the median flux intensity in the natural units of the Pipe3D preprocessed MaNGA data cube (Section 3), at the wavelength range 5450–5550 Å. The second channel includes the index of the tessella in which each spaxel is grouped, after the binning has been performed to increase the S/N. The third channel comprises the dezonification map, as described in Section 4.1. The average flux intensity and the standard deviation within the same wavelength range as for the unbinned flux in the first channel, after applying the spatial binning, are included in the following two channels. Up to now, the stored properties correspond to parameters derived directly from the original data cubes, before performing the spectral fit. However, the remaining channels correspond to properties extracted directly from our fits, including the LW and mass-weighted (MW) age (${{ \mathcal A }}_{\star ,L}$ and ${{ \mathcal A }}_{\star ,M}$) and metallicity (${{ \mathcal Z }}_{\star ,L}$ and ${{ \mathcal Z }}_{\star ,M}$), respectively, as well as the dust extinction, the stellar velocity, the stellar velocity dispersion, the stellar mass-to-light ratio, and the stellar-mass density (with and without correction for dust extinction).

A detailed description of how these parameters are derived is presented in Sánchez et al. (2016b, Section 2.3) and Sánchez et al. (2020, Section 3.1, Equations (2)–(4)). The kinematic parameters (vel_⋆ and σ_⋆), and the dust extinction (A_V,⋆), are derived directly from the first step of the spectral decomposition analysis described in Section 4.1, and expressed in Equation (1). The LW parameters (P_L ) are derived from

$$\begin{eqnarray}&&\mathrm{log}{P}_{L}={{\rm{\Sigma }}}_{\mathrm{ssp}}{w}_{\mathrm{ssp},\star ,L}\mathrm{log}{P}_{\mathrm{ssp}},\end{eqnarray} \tag{ 3 }$$

and the MW parameters (P_M ) from

$$\begin{eqnarray}&&\mathrm{log}{P}_{M}=\displaystyle \frac{{{\rm{\Sigma }}}_{\mathrm{ssp}}{w}_{\mathrm{ssp},\star ,L}{{\rm{\Upsilon }}}_{\mathrm{ssp},\star }\mathrm{log}{P}_{\mathrm{ssp}}}{{{\rm{\Sigma }}}_{\mathrm{ssp}}{w}_{\mathrm{ssp},\star ,L}{{\rm{\Upsilon }}}_{\mathrm{ssp},\star }},\end{eqnarray} \tag{ 4 }$$

where w_ssp,⋆,L is the light fraction/weight of each SSP in the library at the normalization wavelength derived from the spectral fit, according to Equation (1), P_ssp is the value of the given parameter for the corresponding SSP (e.g., age, metallicity), and ϒ_ssp,⋆ is the stellar mass-to-light ratio. Note that both the LW and MW average values are indeed geometrical averages (or averages of logarithmic parameters). Thus, these maps should be interpreted as a realization of the statistical distribution of the considered parameter (e.g., age or metallicity distributions; Mejía-Narváez et al. 2020).

Finally, the mass-to-light ratio is derived from a simple weighted average:

$$\begin{eqnarray}&&{{\rm{\Upsilon }}}_{\star }={{\rm{\Sigma }}}_{\mathrm{ssp}}{w}_{\mathrm{ssp},\star ,L}{{\rm{\Upsilon }}}_{\mathrm{ssp},\star }.\end{eqnarray} \tag{ 5 }$$

The stellar surface density is then derived by multiplying ϒ_⋆ by the observed surface density at the normalization wavelength (approximately the rest-frame V band), corrected for extinction by dust and luminosity distance (D_L ), and divided by the area of each spaxel (a_sp):

$$\begin{eqnarray}\begin{array}{rcl}{{\rm{\Sigma }}}_{* } & = & {\mu }_{V}{{\rm{\Upsilon }}}_{\star },\\ {\mu }_{V} & = & \displaystyle \frac{4\pi {D}_{L}^{2}{f}_{V}}{{a}_{\mathrm{sp}}}{10}^{0.4{A}_{{\rm{V}},* }}.\end{array}\end{eqnarray} \tag{ 6 }$$

5.1.2. SFH Extension

This extension comprises a data cube where each slice/channel includes the spatial distribution of the light fraction (weights w_⋆,L in Equation (1)) of each SSP within the adopted library at the normalization wavelength (5500 Å), following the WCS of the original MaNGA cube. Since the decomposition of the stellar population is done for the spatial binning described in Section 4.1, the distributed w_⋆,L have the same value for all the spaxels within the same tessella. As indicated in Section 4.2, the adopted MaStar_sLOG library comprises 273 SSP spectra, covering 39 ages and seven metallicities (Table 4). The spatial distributions of the w_⋆,L for each SSP are then stored in the first 273 channels of the SFH extension. Then, 39 additional channels (from N = 273 to N = 311, with the first channel corresponding to index N = 0) comprise the weights for each individual age (i.e., the values derived by coadding the seven w_⋆,L for the same age, but different metallicities). An example of the spatial distribution of the latter weights for the same data cube as shown in Figure 4, manga-7495-12704, is shown in Figure 5. Since the weights w_⋆,L are not weighted by the flux intensity, i.e., they are just relative quantities, the original shape of the galaxy is not evident in almost all of the panels. Only for the youngest ages (t < 30 Myr) is it possible to trace the location of the disk of this galaxy (somehow). Curiously, most of the light corresponds to stars with ages between ∼1 and ∼6 Gyr, which are more homogeneously distributed than younger stars. Finally, w_⋆,L increases slightly for older stars toward the center, up to ∼12 Gyr. These weights/light fractions correspond to the spatially resolved age distribution function, which is usually explored in the analysis of resolved stellar populations (e.g., Hasselquist et al. 2020).

Zoom In Zoom Out Reset image size

Table 4. Description of the SFH Extension

Channel	Description of the Map
0	w⋆,L for (${{ \mathcal A }}_{\star }$,Z⋆) = (0.001 Gyr, 0.0001)
272	w⋆,L for (${{ \mathcal A }}_{\star }$,Z⋆) = (13.5 Gyr, 0.04)
273	w⋆,L for ${{ \mathcal A }}_{\star }$ = 0.001 Gyr
311	w⋆,L for ${{ \mathcal A }}_{\star }$ = 13.5 Gyr
312	w⋆,L for Z⋆ = 0.0001
318	w⋆,L for Z⋆ = 0.04

Note. "Channel" indicates the z-axis of the data cube, starting from 0, and w_⋆,L indicates the fraction (weight) of light at 5500 Å, corresponding to an SSP of: (i) a certain age (${{ \mathcal A }}_{\star }$) and metallicity (Z) (channels 0–272); (ii) a certain age, i.e., coadding all w_⋆,L corresponding to SSPs with the same age, but different metallicities (channels 273–311); and (iii) a certain metallicity, i.e., coadding all w_⋆,L corresponding to SSPs with the same metallicity, but different ages (channels 312–318). The adopted MaStar_sLOG SSP library comprises 39 ages, ${{ \mathcal A }}_{\star }$/Gyr = (0.001, 0.0023, 0.0038, 0.0057, 0.008, 0.0115, 0.015, 0.02, 0.026, 0.033, 0.0425, 0.0535, 0.07, 0.09, 0.11, 0.14, 0.18, 0.225, 0.275, 0.35, 0.45, 0.55, 0.65, 0.85, 1.1, 1.3, 1.6, 2, 2.5, 3, 3.75, 4.5, 5.25, 6.25, 7.5, 8.5, 10.25, 12, and 13.5), and seven metallicities, Z_⋆ = (0.0001, 0.0005, 0.002, 0.008, 0.017, 0.03, and 0.04).

Download table as:  ASCIITypeset image

In addition, seven more channels are included in the SFH extension, comprising the weights for each individual metallicity (i.e., the values derived by coadding the 39 w_⋆,L for the same metallicity, but different ages). Figure 6 shows the spatial distribution of these light fractions for each of the metallicities included in the SSP library. Despite the clumpy distribution, which is a consequence of the local uncertainties in the derivation of w_⋆,L , there are clear patterns that emerge. For instance, the higher light fractions are more distributed in the outer regions of the galaxies for the lowest metallicities, while the distributions are more homogeneous for the highest metallicities. This is a clear consequence of the spatial variation of the metallicity distribution function (MDF; as reported by Mejía-Narváez et al. 2020). Indeed, these weights are the basic ingredients for deriving the MDFs using our analysis.

Zoom In Zoom Out Reset image size

Table 4 summarizes the information included in each channel of the SFH cube. This information is included in the header with a set of keywords named DESC_N, indicating the content of channel N, with N running from 0 to 318. For completeness, the original file generated by pyPipe3D, from which the information in channel N was obtained, has been listed in the header as FILE_N.

5.1.3. INDICES Extension

This extension comprises a data cube in which each channel/slice corresponds to the spatial distribution of the stellar indices derived for the emission-line-subtracted spectra, following the procedure outlined in Section 4.1 and described in detail in Lacerda et al. (2022) and Sánchez et al. (2016b). As in the previous cases, this analysis has been done for the averaged spectra within each spatial bin, and therefore the reported values are the same for all the spaxels within the same tessella. The content of each channel/slice is listed in Table 5, indicating the explored stellar index and its error. The adopted spectral range for measuring each index, together with the ranges adopted to estimate the adjacent continuum (at bluer and redder wavelength ranges, with respect to the bandwidth that defines the index) are included in the table. For completeness, we include the average flux within the full wavelength range covered by the data and its standard deviation, as proxies for the signal and noise within each tessella. The corresponding information is stored in the header keywords named INDEXN, where N corresponds to each channel. Figure 7 shows an example of the content of this extension, corresponding to the same data cube shown in previous figures (i.e., manga-7495-12704; e.g., Figure 4) and including both the measurements and the estimated errors. As expected for a spiral galaxy, there are clear radial patterns in the different stellar indices that follow the light distribution: (i) D4000 presents a radial gradient, with the highest values being found in the center of the galaxy; (ii) all the explored Balmer indices (Hβ, Hγ, and Hδ), show a decline in the inner regions, most probably associated with the location of the bulge of this galaxy, and a rise, spatially coincident with the disk and the spiral arms; and (iii) the indices more sensible to stellar metallicity, in particular Mgb and Fe5335, present a mild decline from the center to the outer regions. The error maps show a near-constant distribution for all the indices, suggesting that the adopted spatial binning has indeed increased the S/Ns of the outer regions by adding the required spectra. Only in the outermost regions, at both the east and west edges of the FOV, do the errors rise up, suggesting that the surface brightness of the galaxy has dropped beyond the ability of the binning procedure to produce reliable (high enough S/N) spectra. Finally, the map of the standard deviation of the flux intensity presents higher values at the locations of the spiral arms, as expected, since in these regions the emission-line-subtracted spectra (used for the analysis of the stellar indices) present stronger variations and residuals. Similar patterns are appreciated for all the explored galaxies, with their own peculiarities and signatures.

Zoom In Zoom Out Reset image size

Table 5. Description of the INDICES Extension

ID	Channel	Units	Channel Content	Index λ Range (Å)	Blue λ Range (Å)	Red λ Range (Å)
Hd	0/9	Å	Hδ index/error	4083.500–4122.250	4041.600–4079.750	4128.500–4161.000
Hβ	1/10	Å	Hβ index/error	4847.875–4876.625	4827.875–4847.875	4876.625–4891.625
Mgb	2/11	Å	Mgb index/error	5160.125–5192.625	5142.625–5161.375	5191.375–5206.375
Fe5270	3/12	Å	Fe5270 index/error	5245.650–5285.650	5233.150–5248.150	5285.650–5318.150
Fe5335	4/13	Å	Fe5335 index/error	5312.125–5352.125	5304.625–5315.875	5353.375–5363.375
D4000	5/14	Å	D4000 index/error	4050.000–4250.000	3750.000–3950.000
Hdmod	6/15	Å	H${\delta }_{\mathrm{mod}}$/error	4083.500–4122.250	4079.000–4083.000	4128.500–4161.000
Hg	7/16	Å	Hϒ/error	4319.750–4363.500	4283.500–4319.750	4367.250–4419.750
Flux	8/17	10−16 erg s−1 cm−2	Median flux/standard deviation a

Notes. "Channel" indicates the z-axis of the data cube starting from 0.

^a This is measured along the entire wavelength range covered by the spectroscopic data.

Download table as:  ASCIITypeset image

5.1.4. ELINES Extension

As outlined in Section 4.1, and described in detail in Lacerda et al. (2022) and Sánchez et al. (2016a), the pipeline explores the properties of the emission lines by performing two different analyses. In the first analysis, a set comprising the strongest emission lines in the optical range ([OII]3727, [O iii]4959,5007, Hβ, [N ii]6548,83, Hα, and [S ii]6717,31) is fitted with a single-Gaussian function for each line at each spaxel within the pure-GAS cube (once the best model spectra for the stellar populations are subtracted). The fitting procedure recovers the flux intensity, the velocity, and the velocity dispersion for each emission line. The pipeline stores the results of this analysis in the ELINES extension of the Pipe3D FITs file. As in the previous cases, this extension comprises a data cube, in which each channel/slice includes the spatial distribution of a different property derived from this fitting procedure, following the scheme described in Table 6. The corresponding information is stored in the header keywords DESC_N, where N corresponds to the channel index (starting with 0). An example of the content of this extension for data cube (galaxy) manga-7495-12704 is included in Figure 8. For this particular galaxy, the Hα velocity map included in the first channel shows a clear rotational pattern, with an S-shaped distortion in the central regions, suggesting the possible presence of a bar. The ionized gas presents a low velocity dispersion over all the disk near to the instrumental dispersion, rising only in the central regions, at the location of the bulge. The intensity maps of all the emission lines show a clumpy structure, as expected from ionized gas dominated by H ii regions and associations (at the spatial resolution of these data), with most of these regions being distributed along the spiral arms of this galaxy. This is the expected ionized gas distribution for a low-inclination spiral galaxy (e.g., Sánchez 2020; Sánchez et al. 2021).

Zoom In Zoom Out Reset image size

Table 6. Description of the ELINES Extension

Channel	Units	Description of the Map
0	km s−1	Hα velocity
1	Å	Hα velocity dispersion a
2	10−16 erg s−1 cm−2	[O ii]3727 flux intensity
3	10−16 erg s−1 cm−2	[O iii]5007 flux intensity
4	10−16 erg s−1 cm−2	[O iii]4959 flux intensity
5	10−16 erg s−1 cm−2	Hβ flux intensity
6	10−16 erg s−1 cm−2	Hα flux intensity
7	10−16 erg s−1 cm−2	[N ii]6583 flux intensity
8	10−16 erg s−1 cm−2	[N ii]6548 flux intensity
9	10−16 erg s−1 cm−2	[S ii]6731 flux intensity
10	10−16 erg s−1 cm−2	[S ii]6717 flux intensity

Notes. "Channel" indicates the z-axis of the data cube, starting from 0.

^a FWHM, i.e., 2.354σ. The instrumental velocity dispersion has not been removed.

Download table as:  ASCIITypeset image

5.1.5. FLUX_ELINES and FLUX_ELINES_LONG Extensions

The second analysis performed by pyPipe3D to extract the main parameters of the emission lines is based on a weighted moment analysis, summarized in Section 4.1 and described in detail in Lacerda et al. (2022) and Sánchez et al. (2016a). This analysis is performed over a larger list of emission lines than the Gaussian fitting procedure discussed in the previous section. For former distributions of the MaNGA Pipe3D analysis (DR14 and DR15; Sánchez et al. 2018), we adopted the list of emission lines included in Sánchez et al. (2016a), which was mostly focused on the emission lines detectable in the wavelength range bluer than λ = 7000 Å, since it was originally compiled to explore the ionized gas in the CALIFA data set (Sánchez et al. 2012). This list comprises a total of 56 emission lines. In this new distribution, we decided to update the list, enlarging the number of emission lines for better coverage of the redder wavelength range of the MaNGA data set, and using a single/homogeneous set of wavelengths (and not a compilation). We adopted the list of emission lines in Fesen & Hurford (1996), selecting those lines within the formal wavelength range covered by the MaNGA data set. The final list comprises 192 emission lines. For simpler compatibility and comparison with previous versions of the data set, we include in the Pipe3D file two extensions: one for analysis of the former list of emission lines (FLUX_ELINES), and another one for the new updated list (FLUX_ELINES_LONG). Both extensions follow the same basic scheme, described in Table 7, comprising eight different parameters for each analyzed emission line: flux intensity, velocity, velocity dispersion, and equivalent width (and their corresponding errors). The details of the two adopted sets of emission lines are presented in Appendix B, Tables 13 and 14. We note that in the ELINES and the two FLUX_ELINES extensions, the velocity dispersion is given as an FWHM in angstroms, without subtracting the instrumental dispersion. The following transformation should be applied to derive the velocity dispersion in kilometers per second:

$$\begin{eqnarray}&&{\sigma }_{\mathring{\rm A} }=\mathrm{FWHM}/2.354,\ \ {\sigma }_{\mathrm{km}\ {{\rm{s}}}^{-1}}=\displaystyle \frac{c}{\lambda }\sqrt{{\sigma }_{\mathring{\rm A} }^{2}-{\sigma }_{\mathrm{inst}}^{2}},\end{eqnarray} \tag{ 7 }$$

where c is the speed of light in kilometers per second, λ is the wavelength of the emission line, and σ_inst is the instrumental resolution (∼1.6 Å).

Table 7. Description of the FLUX_ELINES Extensions

Channel	Units	Description of the Map
I	10−16 erg s−1 cm−2	Flux intensity
I+N	km s−1	Velocity
I+2N	Å	Velocity dispersion a
I+3N	Å	Equivalent width b
I+4N	10−16 erg s−1 cm−2	Flux error
I+5N	km s−1	Velocity error
I+6N	Å	Velocity dispersion error
I+7N	Å	Equivalent width error

Notes. I is a running index over the set of emission lines listed in Appendix B, and N is the number of analyzed lines. N = 57 for FLUX_ELINES and N = 192 for FLUX_ELINES_LONG.

^a FWHM, i.e., 2.354σ. The instrumental velocity dispersion has not been removed. ^b We follow the convention by which the EW for an emission line is negative and positive for an absorption line.

Download table as:  ASCIITypeset image

As for the previous extensions, we present in Figure 9 an example of the content of the FLUX_ELINES extension, corresponding to the properties derived for the [O ii]3727 emission line for the data cube (galaxy) manga-7495-12704. As expected, we observe a similar distribution in the flux intensity reported for this emission line as based on the moment analysis, rather than the one extracted using the Gaussian fitting (shown in Figure 8). The patterns in the velocity and velocity dispersion are also similar to the ones found for Hα using the Gaussian fitting, although they look noisier, as expected, since the [O ii]3727 emission line is weaker through the FOV in this galaxy, and it is a blended doublet. Finally, EW(Hα) presents clear negative values, with an absolute value larger than 3 Å through most of the FOV of this data set, in agreement with a scenario in which the ionization is dominated by young massive OB stars associated with recent star formation activity (e.g., Sánchez et al. 2021).

Zoom In Zoom Out Reset image size

5.1.6. GAIA_MASK Extension

5.1.7. SELECT_REG Extension

The described set of data products included in the current distribution comprises a vast data set that can be used to explore a large number of spatially resolved properties of galaxies. As a simple showcase, we use this data set to explore how the ionization conditions change across the optical extension of galaxies, as well as their dependence on galaxy properties. To do this, we emulate Figure 5 of Sánchez (2020), a review aimed at characterizing the main spatially resolved properties of galaxies in the nearby universe. This figure shows the distribution across the classical [O iii]/Hβ versus [N ii]/Hα diagnostic diagram of the full set of spatially resolved regions extracted from a compilation of data from four different IFS-GSs (CALIFA, SAMI, MaNGA, and AMUSSING++). This compilation comprises ∼8000 galaxies (i.e., it is of the same order as the final MaNGA data set), from which we select a subsample of ∼2000 well-resolved and well-sampled objects. Despite the size of this sample, and the efforts to homogenize the data products (for more details, see Sánchez 2020), no compilation has the advantages of a well-defined sample, like the one offered by MaNGA, considering that all the data were observed, reduced, and analyzed using the same procedures and tools. Following a similar philosophy, we select a subset of MaNGA data cubes with the best characteristics in order to explore the spatial distribution of the ionized gas. From the full sample, we retain only the galaxies/cubes observed with the largest IFUs (i.e., those comprising 127, 91, and 61 fibers), whose FOVs cover at least 2 Re of the galaxy. This guarantees good sampling and coverage of the spatially resolved properties, based on the simulations presented by Ibarra-Medel et al. (2019). Furthermore, we keep only those galaxies that are clearly resolved, i.e., whose Re is larger than the PSF FWHM of the MaNGA observations, following Belfiore et al. (2017b). Finally, we exclude highly inclined galaxies, i.e., the ones whose ellipticity is larger than 0.75. This latter condition is adopted to guarantee that the observed patterns result from changes with galactocentric distance, and are not an effect of a change of the ionization with vertical distance, due to the presence of shocks associated with galactic outflows (e.g., Heckman et al. 1990; Bland-Hawthorn 1995; López-Cobá et al. 2020) or extraplanar diffuse ionized gas (e.g., Flores-Fajardo et al. 2011; Levy et al. 2018). The final subsample fulfilling all these criteria comprises ∼5500 well-resolved/well-sampled galaxies/cubes.

Figure 10 shows the distribution along the BPT diagram of the full set of spatially resolved regions (spaxels) included in the final sample of well-resolved/well-sampled galaxies, segregated by mass and morphology, and color coded by the average EW(Hα). The average location within this diagram for different galactocentric distances is indicated by a solid circle (whose size grows with radius). To generate this figure, the contents of the FLUX_ELINES extensions in the pyPipe3D FITs file were used, to extract the flux intensity maps of [O iii]λ5007, Hβ, [N ii]λ6584, and Hα, and the map of EW(Hα). Then, for each galaxy, two maps were derived, each one comprising the two line ratios involved in the diagram. We use these maps to derive: (i) the azimuthally averaged radial distribution of both line ratios, with the galactocentric distance normalized to the effective radius; (ii) the density distribution of the spaxels across the BPT diagram; and (iii) the distribution of EW(Hα) across the same diagram. Finally, for each morphological and stellar-mass subsample (and for the complete sample), we derive the averages of all these properties. In this way, each galaxy contributes equally to the final distribution, irrespective of its number of ionized regions (represented by the individual density distribution).

Zoom In Zoom Out Reset image size

The emerging patterns as seen in Figure 10 are very similar to those already discussed in Sánchez (2020), although there are some mild but evident differences that are clearly related to the differences in the explored samples. For the complete data set (the upper left panel), the peak of the density distribution traces the classical location of H ii regions (Osterbrock 1989), associated with recent SF processes. The highest densities are found at the lower right end of the distribution. There is a clear segregation of the locations in the BPT diagram with EW(Hα), with most of the high(low)-EW regions being located on the left(right) side of the diagram. This bimodality is a direct consequence of the LINER-like regions (less frequent) being preferentially located in the central regions of galaxies, while the SF-like regions (more frequent) are located much farther away. The above is reflected in the distributions of the average values at different galactocentric distances, being clearly above the Kewley et al. (2001) demarcation line for R < 0.6 Re (the region dominated by the bulges), but below the Kauffmann et al. (2003a) demarcation line for R > Re (the region dominated by the disk, in late-type galaxies). The most plausible explanation for this pattern is that LINER-like ionization is dominated by the contribution of hot-evolved low-mass stars (HOLMES; Flores-Fajardo et al. 2011), also known as post–asymptotic giant branch (pAGB) ionization (Binette et al. 1994). The above would be the reason why this ionization is ubiquitous in galaxies (Singh et al. 2013), and more frequently found in the central regions, dominated by old stellar populations (Belfiore et al. 2017a; Lacerda et al. 2018).

This pattern is modulated by both the stellar mass and the morphology. As galaxies are more massive and earlier, the presence of SF-like regions is less frequent. The peak of the density distribution is shifted toward the LINER-like region, which is more relevant for these kinds of galaxies. If there is still any SF activity, it is mostly located in the outer regions (at least for M_⋆ > 10^9.5 M_⊙). On the contrary, as galaxies are less massive and later, the LINER-like component becomes less relevant and the dominance of the ionization by SF processes becomes more evident (to the extent that, in the latest morphological bins (Sc/Sd), and in particular for M_⋆ < 10¹¹ M_⊙, there is almost no LINER-like component in the average distribution). An additional pattern is observed in the radial trend traced by the solid gray circles in the SF-dominated regions. For galaxies more massive than M_⋆ > 10^9.5 M_⊙, there is a shift from the bottom right end of the classical location of the H ii region toward its upper left end, from the inner (0.6–0.8 Re) to the outer (1.8–2.1 Re) regions of the disks. This radial trend is directly associated with the oxygen abundance gradient observed in galaxies (e.g., Searle et al. 1973; Vila-Costas & Edmunds 1992; Sánchez et al. 2014). On the contrary, this trend is absent or even reverted for low-mass galaxies (M_⋆ < 10^9.5 M_⊙). This suggests a possible flattening of the oxygen abundance gradient in this mass bin (e.g., Belfiore et al. 2017b). These results are in agreement with those presented in Sánchez (2020), and discussed in detail in Sánchez et al. (2021), highlighting the clear connection between the ionization processes and the properties of the underlying stellar populations, which are the dominant ionizing sources in galaxies (on average). This connection is the main driver for the observed patterns and trends.

In this example, we have made use of a tiny fraction of the information distributed in the current data set. A similar analysis, using different stellar properties instead of the EW(Hα), would reveal the outlined connection. Explorations using different line ratios or other physical properties, like the stellar or gas velocity dispersion (e.g., D'Agostino et al. 2019; Law et al. 2021), are easily obtained by introducing simple modifications to the outlined procedure.

From the results of the analysis performed by pyPipe3D, we derive a set of spatially resolved properties of the stellar populations and emission lines in the sample galaxies, following Sánchez et al. (2018), Sánchez (2020), and Sánchez et al. (2021). We do not distribute the entire set of resolved properties in the current release. However, we provide for each galaxy property either its integrated value (for extensive parameters, such as the stellar mass) or a characteristic value (for intensive parameters, such as the oxygen abundance), and/or a value at a certain aperture. For the integrated extensive quantities, we just coadd the corresponding values spaxel by spaxel, excluding those spaxels for which the quantity cannot be derived (we will include some examples below). For intensive properties, which present clear gradients along the galactocentric distance, we derive their azimuthally averaged radial distribution. To do so, we use the position angle, ellipticity, and effective radius (Re) provided by the NSA catalog ¹⁵ for each galaxy, to create elliptical apertures of 0.15 Re width, covering the galactocentric distance from 0 to 3.5 Re. Then, we estimate the average value for each parameter (and its standard deviation). From these radial distributions, we derive the value at the effective radius and the slope of the average gradient, based on a linear regression of the considered parameter along the radius (normalized to Re). This fitting is restricted to the galactocentric distance range between 0.5 and 2.0 Re. When the FOV does not reach this galactocentric distance, the regression is restricted to the largest distance covered by the FOV. We acknowledge that fitting the radial gradients of the explored physical properties to a linear relation is an oversimplification, in many cases, as clearly demonstrated by the results based on previous IFS-GSs (e.g., CALIFA; González Delgado et al. 2014, 2015). For a more detailed description of the radial gradients and the validity of this approximation, we refer the reader to J. K. Barrera-Ballesteros et al. (2022, in preparation). It is worth noticing that for most of the explored quantities in the galaxies, the value at the effective radius is usually considered a good proxy of the average one (e.g., Moustakas et al. 2010; Sánchez et al. 2016a), being a better representation than the mean value for data with variable FOVs (with respect to the optical extension of the galaxy).

For some parameters, we also include the central value, derived from a fixed aperture of 25 diameter (i.e., 1 FWHM of the spatial PSF). These values reveal the effects of physical processes and structures that are present in the central regions of galaxies, such as AGNs, outflows, or strong bulges, without necessarily being representative of their average properties (for a deeper discussion of this topic, see Sánchez 2020 and Sánchez et al. 2021). For some particular parameters, we also provide the values within a 1 Re aperture (i.e., not the value at the effective radius, but the value within one effective radius).

Table 18 lists all the integrated, characteristic, and aperture-limited properties derived for each cube/galaxy. For completeness, we have included additional parameters extracted from other catalogs, together with the morphological classification, photometric/structural parameters, and quality control flags described in Sections 4.3, 4.4, and 4.5. In general, the error estimated for each quantity is labeled with the same name, but with the prefix e_ (e.g., e_log_Mass corresponds to the error of the parameter log_Mass). The final catalog is included as an extension of the SDSS17Pipe3D_v3_1_1.fits FITs file distributed online. ¹⁶ This FITs file comprises three extensions. The first extension, HDU[0], is empty, just to fulfill the standard format adopted by the SDSS collaboration. The second extension, HDU[1], includes the described catalog, comprising 535 entries for each galaxy. The third extension, HDU[2], consists of row-stacked spectra, comprising the individual spectra included in the MaStars_sLOG SSP library (Section 4.2), in the format required to run pyFIT3D and pyPipe3D (Lacerda et al. 2022).

The derivation of the delivered parameters listed in Table 18 has been described in detail in many previous articles. Below, we provide a summary of the adopted procedures and some details for those parameters that require extra information.

5.3.1. Parameters Inherited from the DRP

5.3.2. Photometric and Structural Properties

5.3.3. Morphological Properties

5.3.4. Stellar Population–related Quantities

Stellar masses. Most of the stellar parameters from which we derive the integrated and characteristic properties of each galaxy result from the decomposition of the stellar spectra in the set of SSPs described in Sections 4.1 and 5.1.1. We estimate the integrated stellar mass (M_⋆; labeled as log_Mass) by coadding the stellar surface density (Σ_⋆; Equation (6)) through the unmasked region (Section 5.1.7) within the FOV of the data. The regions contaminated by foreground stars included in the GAIA_MASK extension of the Pipe3D file have been masked in the derivation of any integrated quantity, including M_⋆. From Σ_⋆, we derive not only M_⋆, but also: (i) the mass within R50 (log_Mass_corr_in_R50_V), using the parameters described in Section 4.4, and the mass within 1 Re (log_Mass_in_Re); (ii) the radius enclosing 50% of the stellar mass in kpc (R50_kpc_Mass; not to be confused with R50_kpc_V, the radius enclosing 50% of the V light in kpc, i.e., R50 transformed to kpc); (iii) the stellar-mass surface density in the central aperture (Sigma_Mass_cen), at 1 Re (Sigma_Mass_Re), and averaged over all the FOV (Sigma_Mass_ALL). The derivation of Σ_⋆ is based on estimates of the mass-to-light ratio (ϒ_⋆; Equation (5)), spaxel by spaxel, from which we also derive its average value across the FOV (ML_avg). A different method for estimating the average ϒ_⋆ across the entire galaxy is to divide the integrated stellar mass by the integrated luminosity (ML_int). Both quantities are listed in the catalog.

Figure 11 shows the comparison between the stellar mass derived by us from our photometry (M_⋆,phot; Section 5.3.2), the stellar mass provided by the NSA catalog, based on multiband photometry (M_⋆,NSA; Section 5.3.1), and the stellar mass derived by spectral fitting using pyPipe3D (M_⋆; Section5.3.4). We applied a systematic offset to the photometric masses. In the case of M_⋆,phot, the offset is a pure consequence of the different IMF adopted (0.28 dex, according to González Delgado et al. 2016). However, in the case of M_⋆,NSA, applying a similar offset produces an overcorrection, with the NSA masses being systematically larger than the masses derived by pyPipe3D. Aperture issues may be behind this difference. It is known that MaNGA apertures miss ∼22% of the total flux (e.g., Pace et al. 2019). On average, we find that an offset of 0.15 dex is a better correction for matching the NSA and pyPipe3D stellar masses. Once corrected, the photometric stellar masses agree with those derived using the stellar synthesis code within ∼0.17 dex, which is similar to the differences found by other groups (e.g., Pace et al. 2019). Furthermore, this difference is of the order of the expected error in the stellar mass due to uncertainties in the ϒ_⋆–color relation, estimated to be ∼0.09–0.20 dex (e.g., Fraser-McKelvie et al. 2019; García-Benito et al. 2019).

Zoom In Zoom Out Reset image size

Age and metallicity. We derive from our analysis the LW and MW stellar age and metallicity, spaxel-wise, as described in Section 5.1.1, following Equations (3) and (4). We report their characteristic values at the effective radius, PAR_LW_Re_fit and PAR_MW_Re_fit, the slope of their radial gradients, alpha_PAR_LW_Re_fit and alpha_PAR_MW_Re_fit, and the corresponding errors, labeled with an e_ prefix. PAR corresponds to either the Age or the metallicity ZH. These parameters are listed in the catalog in rows 24 to 39. Additionally, we include the dust extinction at the effective radius as derived from the analysis of the stellar population, Av_ssp_Re (row 173) and its error e_Av_ssp_Re (row 174).

Star formation and chemical enrichment histories. Σ_⋆, M_⋆, ${{ \mathcal A }}_{\star }$, and ${{ \mathcal Z }}_{\star }$ can be estimated at any lookback time (τ_lbt) by integrating the corresponding equations (e.g., Equation (3)) from the oldest age in the SSP decomposition down to the age matching τ_lbt. This is the standard procedure, as broadly applied in the literature, for deriving, for instance, the mass assembly history (MAH; Pérez et al. 2013; Ibarra-Medel et al. 2016, 2019), the SFH (Panter et al. 2007; González Delgado et al. 2017; López Fernández et al. 2018; Sánchez et al. 2019a), or the chemical enrichment history (ChEH; Vale Asari et al. 2009; Camps-Fariña et al. 2021, 2022). This derivation can be applied to integrated, characteristic, and/or spatially resolved properties. In the left panels of Figure 3, we show the normalized MAH and the ChEH at different galactocentric distances for the example galaxy/cube manga-7495-12704. Figure 12 illustrates the derivation of a spatially resolved parameter for different τ_lbt, by showing the cumulative Σ_⋆,t for the same prototype/example galaxy, derived by making use of the content of the SFH extension, and applying Equations (5) and (6) for a range of ages, labeled in each panel of the figure, together with the corresponding slices of the SFH cube. Integrating Σ_⋆,t through the FOV of the IFU versus τ_lbt, we derive the stellar mass versus time, the MAH. A correction to account for the mass of stars that are dead at the observing time, but that were still shining at τ_lbt, should be applied (e.g., Courteau et al. 2014). The SFH (integrated or resolved) can be obtained from the MAH (or Σ_⋆,t ), by dividing the differential mass accumulated at two consecutive τ_⋆,t by the corresponding time interval, namely:

$$\begin{eqnarray}&&{\mathrm{SFR}}_{\mathrm{ssp},{\rm{t}}}=\displaystyle \frac{{\rm{\Delta }}{M}_{\star ,{\rm{t}}}}{{\rm{\Delta }}{\rm{t}}},\ \ {{\rm{\Sigma }}}_{\mathrm{SFR},\mathrm{ssp},{\rm{t}}}=\displaystyle \frac{{\rm{\Delta }}{{\rm{\Sigma }}}_{\star ,{\rm{t}}}}{{\rm{\Delta }}{\rm{t}}},\end{eqnarray} \tag{ 8 }$$

where SFR_ssp,t (Σ_SFR,ssp,t) is the star formation (density) at a particular lookback time t, Δt is the time interval between two consecutive τ_lbt (ages in the SSP library), and ΔM_⋆,t (ΔΣ_⋆,t) is the differential mass (density) assembled during Δt. When τ_lbt is short enough, the estimated SFR corresponds to the current one (e.g., González Delgado et al. 2016). In our catalog, we include three estimates of SFR based on the analysis of the stellar population: one corresponding to the last 32 Myr (log_SFR_ssp), adopting the time interval proposed by González Delgado et al. (2016) and already tested in Sánchez et al. (2019a), and two additional values corresponding to 10 and 100 Myr, log_SFR_ssp_10Myr and log_SFR_ssp_100Myr, respectively. The latter intervals correspond to the maximum time at which an OB star can still produce some ionizing photons before it fades out (10 Myr) and the typical time interval associated with the SFR measured using far-infrared emission (100 Myr; e.g., Kennicutt 1998; Catalán-Torrecilla et al. 2015).

Zoom In Zoom Out Reset image size

Star formation history timescales. Following Pérez et al. (2013) and García-Benito et al. (2017), we estimate the age at which a fraction X (from 30% to 99%) of the current stellar mass was formed, based on the integrated MAH derived for each galaxy. Then, T30 (T80) indicates the time at which 30% (80%) of the stellar mass was formed. In addition, making use of the integrated and spatially resolved ChEH, we derive: (i) the average stellar metallicity, normalized to the solar value, in logarithmic scale at the same lookback times—e.g., ZH_T30 corresponds to the metallicity at the time at which 30% of the current mass was formed; (ii) the metallicity at the effective radius (ZH_Re_T%); and (iii) the radial gradient of the metallicity at the same time (a_ZH_T%). These quantities are listed in the catalog in rows 107 to 142.

Stellar spectral indices. The pyPipe3D analysis provides us with the spatial distribution of a set of stellar spectral indices for each galaxy/cube, as described in Section 5.1.3. For each index, we estimate its value at the effective radius (ID_Re_fit), the slope of its radial gradient (ID_alpha_fit), and the corresponding errors (labeled with the e_ prefix), where ID is the index identification in Table 5. The indices are listed in rows 436–467 of the final catalog (Table 18).

Spectral indices are frequently used to derive the average properties of stellar populations, such as age and metallicity (e.g., Gallazzi et al. 2005). Some indices, such as D4000, are sensitive to the age of the stellar population, tracing the fraction of young to intermediate-to-old stars, or Hδ, which is more sensitive to the presence of young stars. Other indices, like Mgb, are sensitive to metallicity. Since indices permit an alternative method of analysis to the stellar decomposition at the core of pyPipe3D, it is worth comparing the results from both methods. Figure 13 shows: (i) the distribution of Hδ as a function of D4000; (ii) the distribution of ${{ \mathcal A }}_{\star ,L}$ as a function of these two indices; and (iii) the distribution of ${{ \mathcal Z }}_{\star ,L}$ as a function of Mgb, where all the index values are estimated at the effective radius. As additional information, we indicate the locations of the different morphological types in these diagrams. The Hδ–D4000 diagram shows the typical distribution observed for the bulk population of nearby galaxies (e.g., Kauffmann et al. 2003b), a clear anticorrelation, in the sense that galaxies with higher Hδ (young stellar populations) have a lower D4000 (a lower fraction of old stars with respect to young ones). Late-type (early-type) galaxies are found in the upper left (lower right) area of the distribution. ${{ \mathcal A }}_{\star ,L}$ shows a well-defined positive (negative) relation with D4000 (Hδ), with the different morphological types following the expected behavior. Finally, ${{ \mathcal Z }}_{\star ,L}$ presents a well-defined positive relation with Mgb, which seems to be less tight than that tracing ${{ \mathcal A }}_{\star ,L}$ with D4000 or Hδ. As expected, early-type (late-type) galaxies are found in the regime of high (low) ${{ \mathcal Z }}_{\star ,L}$ and Mgb values. This comparison suggests that our estimated stellar indices are indeed good tracers of the stellar content.

Zoom In Zoom Out Reset image size

5.3.5. Emission-line-related Quantities

We derive a wide variety of parameters based on the emission-line analysis performed by pyPipe3D. First, we determine which are the most frequently detected emission lines. To do so, we explore how often the different emission lines included in the FLUX_ELINES_LONG extension (Section 5.1.5, Table 14) are detected in all the analyzed data cubes. We select emission lines that are detected with an S/N larger than 3 in at least 5% of all galaxies/cubes (i.e., in at least ∼500 galaxies). For these emission lines, we estimate the flux intensity at the effective radius and the slope of the radial gradient (flux_ELINE_Re_fit and flux_ELINE_alpha_fit), as well as their corresponding errors (labeled with the e_ prefix), following the procedure described above (Section 5.3). For each parameter, ELINE corresponds to the Id+wavelength listed in Table 14, where the most frequently detected emission lines are labeled with an "*" symbol. All these parameters and errors comprise the rows 180 to 307 of the final catalog, as listed in Table 18.

Line ratios. For those emission lines that are more frequently used in the classical diagnostic diagrams (i.e., [O ii]λ3727, [O iii]λ5007, [O i]λ6300, [N ii]λ6584, and [S ii]λ6717,31), we derive their ratios with respect to Hβ (in the case of [O ii] and [O iii]) and Hα (for the remaining lines), using the values included in the FLUX_ELINES extension at: (i) the central aperture (_cen suffix); (ii) the effective radius (_Re suffix); and (iii) the average across the full FOV of the data cube (_ALL suffix). In addition, we estimate the equivalent width of Hα (EW_Ha_), and the Hα-to-Hβ ratio (Ha_Hb) at the three locations. The corresponding errors have been estimated, too. Rows 13 to 21 and 70 to 100 comprise all these parameters. A simple nomenclature, including the two line ratios involved, is adopted to label these quantities (e.g., log_OI_Ha_cen corresponds to the line ratio between [O i] and Hα in the central aperture in logarithmic scale, while including an e_ prefix means that it corresponds to its error). The Hα flux in the central aperture is also included for the purpose of comparison with the SDSS single-aperture fiber data (rows 162–163). A practical example of the use of these line ratios is included in Section 7.

Dust extinction. We use the Hα-to-Hβ ratio to derive the dust extinction of the ionized gas (A_V,gas), by assuming a nominal ratio of 2.86. This value corresponds to a nebula fulfilling the case-B recombination scenario, with an electron density of n_e = 100 cm⁻³ and an electron temperature of T_e = 10⁴ K (Osterbrock 1989). We adopted an MW-like dust extinction (Cardelli et al. 1989), with a total-to-selective extinction R_V = 3.1. By conducting this derivation spaxel by spaxel, we are able to derive a dust extinction map for each galaxy/cube (A_V ). From those maps, we estimate the dust extinction at the effective radius (Av_gas_Re) and its error, included in the final catalog in rows 171–172. These A_V maps are used to correct all the observed emission lines by their dust extinction, providing dust-corrected emission lines that can be used in the derivation of additional parameters.

Oxygen and nitrogen abundances. The emission-line maps can be used to examine the possible ionizing sources spaxel by spaxel. In particular, following the criteria described in detail in Sánchez (2020) and Sánchez et al. (2021), we classify as SF areas those regions below the Kewley et al. (2001) demarcation line in the classical BPT diagram involving [O iii]/Hβ and [N ii]/Hα with EW(Hα) > 3Å. For those regions, and only for them, it is possible to estimate the oxygen and nitrogen abundances, and the ionization parameter, based on strong line indicators (since those are only calibrated for ionization due to young OB stars, associated with recent SF activity). The dust-corrected emission-line maps are used for the derivation of the oxygen abundances using the different calibrators adopted along this exploration. For the central aperture, we estimate the oxygen abundance using the calibrators adopted in Sánchez et al. (2019b), which comprise a mixed set, including calibrators anchored to measurements based on the direct method and calibrators derived using photoionization models. Among them, we include here the O3N2- and N2-based calibrators proposed by Marino et al. (2013; OH_O3N2_cen and OH_N2_cen), the ONS calibrator by Pilyugin et al. (2010; OH_ONS_cen), the R23-based calibrator by Kobulnicky & Kewley (2004; OH_R23_cen), the pyqz calibrator of Vogt et al. (2015; OH_pyqz_cen), the calibrators of Maiolino et al. (2008; OH_M08_cen), Tremonti et al. (2004; OH_T04_cen), Dopita & Sutherland (1996; OH_dop_cen), and Pérez-Montero & Contini (2009; OH_O3N2_EPM09_cen), and the t2-corrected calibrator of Peña-Guerrero et al. (2012), as derived by Sánchez et al. (2019b; OH_t2_cen). All these oxygen abundances and their errors are included in rows 50 to 69 of the final catalog.

A larger and more complete set of calibrators was recently used by Espinosa-Ponce et al. (2022) in their exploration of the properties of H ii regions. We adopt most of those calibrators to derive the oxygen and nitrogen abundances (or the nitrogen-to-oxygen abundance ratio), together with the ionization parameter, for those spaxels whose ionization is compatible with young massive OB stars (i.e., associated with recent SF). The complete list of adopted calibrators is included in Table 15 of Section D. Once derived, those parameters spatially resolved (i.e., spaxel by spaxel), and we estimate their corresponding values at the effect radius (OH_CAL_Re_fit, where CAL corresponds to the ID of each calibrator), the slope of their radial gradient (OH_CAL_alpha_fit), and their errors (labeled with the prefix e_). All these values are included in the final catalog, covering rows 308 to 431. They comprise a total of 24 oxygen abundance estimates, four different nitrogen (and nitrogen-to-oxygen) estimates, and four estimates of the ionization parameter.

Figure 14 shows a comparison of a subset of the oxygen abundance calibrators included in the final catalog as a function of the values derived using the Ho (2019) calibrator, selected as an arbitrary fiducial oxygen abundance calibrator following Espinosa-Ponce et al. (2022). This calibrator implements a state-of-the-art machine-learning routine to provide a calibrator that anchors the oxygen abundance to direct measurements using the direct method. It uses a large number of emission-line ratios, as listed in Table 15, providing a quite accurate estimation of the oxygen abundance. A similar comparison for the full set of calibrators is included in Appendix D. It is beyond the scope of the current study to discuss in detail the well-known discrepancies among different O/H calibrators and the possible transformations between them (e.g., Kewley & Ellison 2008; López-Sánchez et al. 2012; Curti et al. 2020; Espinosa-Ponce et al. 2022). The main aim of this comparison is to show that there is indeed a correspondence between the oxygen abundances derived using different calibrators included in our catalog. This correspondence is far from being a one-to-one relation in most of the cases. In a few cases, the oxygen abundances present just a constant offset for all the considered dynamical range (e.g., the Kew02 N2O2 and Pil16 R and S calibrators). For a large number of calibrators, the relation is well characterized by an almost linear relation, with a slope different than 1 (e.g., Cur20 N2, Mar N2, Pet04 N2). In some of them, the dynamical range is considerably different from calibrator to calibrator (e.g., Cur20 R2 or RS32), which would translate into a large variety of slopes for the linear relation that would match them. Finally, there are cases in which there are clear deviations from the linear relation, with a change in the slope or even a plateau found at high abundances (e.g., Pil10 ON and NS), or even large discrepancies in the same regime (e.g., T04). In summary, any analysis using the oxygen abundances included in the delivered catalog should acknowledge the described differences among the different calibrators, and the impacts on the results. For instance, explorations of the shape of the mass–metallicity relation (MZR) or the oxygen abundance gradient are deeply affected by the adopted calibrator, both qualitatively (e.g., the presence of plateaus in the distribution of the MZR; Sánchez et al. 2019b; Alvarez-Hurtado et al. 2022) and quantitatively (e.g., the actual values of the radial gradients; Belfiore et al. 2017b; Sánchez-Menguiano et al. 2018).

Zoom In Zoom Out Reset image size

A less relevant difference is found among the three estimates of the nitrogen-to-oxygen relative abundance included in our catalog, as seen in Figure 15. They present clear linear relations among them, with a slope near to 1, and small offsets (Δlog(N/O) ∼ 0.03–0.08 dex) and scatters (${\sigma }_{{\rm{\Delta }}\mathrm{log}({\rm{N}}/{\rm{O}})}\sim$ 0.03–0.07 dex). In this case, the use of a different calibrator would produce little quantitative difference and essentially no qualitative differences.

Zoom In Zoom Out Reset image size

Star formation rate. The dust-corrected Hα luminosity is used to estimate the SFR spaxel-wise, by applying the relation proposed by Kennicutt (1998) for the Salpeter (1955) IMF:

$$\begin{eqnarray}&&{\rm{S}}{\rm{F}}{\rm{R}}\,({M}_{\odot }\,{{\rm{y}}{\rm{r}}}^{-1})=0.79\times {10}^{-41}\,{L}_{{\rm{H}}\alpha }({\rm{e}}{\rm{r}}{\rm{g}}\,{{\rm{s}}}^{-1}).\end{eqnarray} \tag{ 9 }$$

From this distribution, it is possible to derive the SFR surface density (Σ_SFR). As in the case of the stellar mass, we can derive the integrated SFR just by coadding the spaxel-by-spaxel values across the FOV of the IFU data. As already discussed in previous articles (Cano-Díaz et al. 2016; Sánchez et al. 2018), this SFR is an upper limit to the real one, since in this derivation all Hα flux is integrated, irrespective of the nature of the detected ionization. In other words, this calculation would derive an SFR even for those galaxies in which there is no ionization that could be directly associated with a recent SF event (i.e., in the case of RGs, such as elliptical ones). However, it is still useful as an upper limit, in the event that a fraction of this ionization is still due to SF, but it is so weak that it is overshadowed by ionization due to other sources (i.e., shocks, ionization by old stars, like pAGBs and HOLMESs, or AGNs; see Sánchez et al. 2021, and references therein). We include this integrated SFR in the catalog (log_SFR_Ha), together with its error (e_log_SFR_Ha), in rows 7 and 11. However, for completeness, we also derive the SFR just by coadding those regions in each galaxy where ionization is compatible with recent SF, using the combination of the BPT diagram plus the EW(Hα), as described above (log_SFR_SF). Furthermore, we estimate the SFR after considering the possible contamination from a diffuse ionization gas, due to old stars (e.g Binette et al. 1994; Flores-Fajardo et al. 2011), assuming a constant EW(Hα) = 1 Å for this component (log_SFR_D_C).

Figure 16 shows the comparison between the SFR derived using the dust-corrected Hα luminosity (SFR_Hα ) and the values derived based on the stellar synthesis analysis (SFR_ssp) for three different timescales (10, 32, and 100 Myr), and for the average of the three of them. We find a clear correspondence between the four SFR_ssp with SFR_Hα , with the average of the three timescales providing the lower offset with respect to the one-to-one relation (Δlog(SFR) = 0.06 ± 0.30 dex), followed by the value for 32 Myr. On average, the SFR derived for 10 Myr (100 Myr) is higher (lower) than the one estimated using the Hα luminosity. These results agree with previous explorations using different IFS data sets (e.g., González Delgado et al. 2016; Barrera-Ballesteros et al. 2021a). In summary, the use of SFR_ssp instead of SFR_Hα would provide similar statistical results—for instance, when exploring global relations such as the SF main sequence (e.g., Sánchez et al. 2018, 2019a). In principle, this quantity should be more accurate, since it does not involve any assumed SFH or ChEH, as it is necessary to derive the scaling between the Hα luminosity and the SFR (Kennicutt 1998). However, it is most probably less precise, due to the additional uncertainties introduced by the stellar synthesis analysis in the derivation of this quantity. As a consequence, the derived relations may present a larger scatter.

Zoom In Zoom Out Reset image size

Molecular gas estimation. The dust extinction is a tracer of the molecular gas content, via the dust-to-gas relation (e.g., Brinchmann et al. 2004, and references therein). We adopt the recent calibrator from Barrera-Ballesteros et al. (2021b) to estimate the molecular gas surface density (Σ_mol) through the spaxel-by-spaxel A_V,gas parameter, using the formula

$$\begin{eqnarray}&&{{\rm{\Sigma }}}_{{\rm{m}}{\rm{o}}{\rm{l}}}\,({M}_{\odot }\,{{\rm{p}}{\rm{c}}}^{-2})=1.06\,{A}_{{\rm{V}},{\rm{g}}{\rm{a}}{\rm{s}}}^{2.58}\,({\rm{m}}{\rm{a}}{\rm{g}});\end{eqnarray} \tag{ 10 }$$

then, by coadding throughout the FOV of each IFU, we estimate the integrated molecular gas (log_Mass_gas_Av_gas_log_log, row 170). For completeness, we provide the same parameter estimated using the linear calibrator proposed by Barrera-Ballesteros et al. (2020):

$$\begin{eqnarray}&&{{\rm{\Sigma }}}_{{\rm{m}}{\rm{o}}{\rm{l}}}\,({M}_{\odot }\,{{\rm{p}}{\rm{c}}}^{-2})=23\,{A}_{{\rm{V}},{\rm{g}}{\rm{a}}{\rm{s}}}\,({\rm{m}}{\rm{a}}{\rm{g}}),\end{eqnarray} \tag{ 11 }$$

which is included in row 40 (log_Mass_gas). We do not provide the formal errors of these parameters, since the error budget is dominated by the calibrators themselves (∼0.2 dex), rather than by the uncertainties in the derivation of the dust extinction.

Two additional estimates of the integrated molecular gas mass have been included in the catalog. One of them is derived by applying a possible correction that takes into account the dependence of the dust-to-gas ratio on the oxygen abundance (log_Mass_gas_Av_gas_OH), and the other is derived by adopting this correction, but by using the stellar dust extinction instead of the ionized gas one in the estimation of the molecular gas surface density (log_Mass_gas_Av_ssp_OH). To derive these estimates, we followed the main procedures described in Barrera-Ballesteros et al. (2021b). However, the dynamical range of the metallicity explored in that study was too narrow to provide a clear conclusion regarding the improvements as a result of introducing these two new calibrators. We include them here in order to allow future explorations to compare them with other estimates of the molecular gas.

Electron density. Finally, we use the [S ii]λ6717,31 line ratio to derive the electron density spaxel by spaxel, solving the equation

$$\begin{eqnarray}&&\displaystyle \frac{[\mathrm{SII}]\lambda 6717}{[\mathrm{SII}]\lambda 6731}=1.49\displaystyle \frac{1+3.77x}{1+12.8x},\end{eqnarray} \tag{ 12 }$$

where x = 10⁻⁴ n_e t^−1/2, t is the electron temperature in units of 10⁴ K (McCall et al. 1985), and n_e is the electron density in units of cm⁻³. As in the case of the dust extinction, we adopted a typical electron temperature of T_e = 10⁴ K in this derivation. We note that the dependence of the electron on this parameter is weak in the adopted formula. As in the cases of other parameters, we deliver the value at the effective radius (Ne_Oster_S_Re_fit), the slope of its radial gradient (Ne_Oster_S_alpha_fit), and their corresponding errors for each galaxy/cube (rows 432–435).

5.3.6. Kinematics-related Quantities

The analysis performed on the data provides the spatial distribution of the stellar and ionized gas velocity and velocity dispersion. From these quantities, we derive different characteristic parameters for each galaxy/cube, which are included in the final catalog: (i) the average velocity-to-velocity dispersion ratio within an aperture of 1 Re ($\tfrac{v}{\sigma }$; labeled as vel_sigma_Re, row 45), and its corresponding error (e_vel_sigma_Re, row 46), estimated as

$$\begin{eqnarray}&&\displaystyle \frac{v}{\sigma }=\sqrt{\tfrac{{\sum }_{r\lt 1\mathrm{Re}}{f}_{\star }{v}_{\star }^{2}}{{\sum }_{r\lt 1\mathrm{Re}}{f}_{\star }{\sigma }_{\star }^{2}}},\end{eqnarray} \tag{ 13 }$$

where f_⋆, v_⋆, and σ_⋆ correspond to the stellar flux intensity in the V band at any position (x,y) within the FOV, the stellar velocity, and the stellar velocity dispersion in each spaxel within the considered apertures; (ii) the stellar and Hα ionized gas velocities (vel_ssp_R and vel_Ha_R), and their corresponding errors (labeled with an e_ prefix), derived in an elliptical ring (following the PA and ellipticity of the galaxy) at 1 and 2 Re (R = 1 or 2), listed in rows 145 to 152; (iii) the stellar and Hα ionized gas velocity dispersion (vel_disp_ssp_R and vel_disp_Ha_R) in the central aperture (R = cen) and at 1 Re (R = 1Re), listed in rows 155 to 158; and (iv) the apparent stellar angular momentum parameter at the effective radius (Emsellem et al. 2007), ${\lambda }_{\mathrm{Re}}$ (labeled as Lambda_Re, row 175), and its corresponding error (e_Lambda_Re, row 176), defined as

$$\begin{eqnarray}&&{\lambda }_{{\rm{R}}}=\displaystyle \frac{{\sum }_{r\lt 1.15\mathrm{Re}}{f}_{\star }r\ | {v}_{\star }| }{{\sum }_{r\lt 1.15\mathrm{Re}}{f}_{\star }r\ \sqrt{{v}_{\star }^{2}+{\sigma }_{\star }^{2}}},\end{eqnarray} \tag{ 14 }$$

where f_⋆, v_⋆, and σ_⋆ correspond to the same parameters as in Equation (13), and r is the deprojected galactocentric distance.

For the current data set, we introduce an inclination correction on λ that was not included in the calculations for DR14 and DR15. This correction is described in the Appendix of Emsellem et al. (2011):

$$\begin{eqnarray}\begin{array}{rcl}{\lambda }^{{\prime} } & = & \displaystyle \frac{\lambda }{{C}_{1}}\\ {C}_{1} & = & {C}_{0}^{2}-{\lambda }^{2}({C}_{0}^{2}-1)\\ {C}_{0} & = & \displaystyle \frac{\sin (i)}{\sqrt{1-0.3* {\cos }^{2}(i)}},\end{array}\end{eqnarray} \tag{ 15 }$$

where λ corresponds to the apparent angular momentum described in Equation (14) and i is the inclination angle of the galaxy.

5.3.7. Volume Correction

The MaNGA sample has a complicated selection function. As already outlined in Section 2, the final sample was built from a set of different subsamples, each of them adopting different selection criteria. Despite this complicated construction, in principle, it is possible to perform a volume correction and derive representative quantities from this sample. Wake et al. (2017) proposed a volume correction in which the volume accessible for each subsample is explored separately, and then the full volume accessible for each individual object is evaluated. In Sánchez et al. (2019a), Appendix E, we proposed a different approach, in which the individual volume accesible for each target in the final sample is estimated a posteriori. The procedure is described in detail in Rodríguez-Puebla et al. (2020). In summary, we adopted the stellar-mass function (Blanton et al. 2017) of the SDSS galaxies to estimate the volume accessible in a set of bins of stellar mass, redshift, and color for this parent sample (as all MaNGA galaxies are extracted from the SDSS spectroscopic sample). Then, we estimate the fraction of MaNGA galaxies in each of those bins with respect to the total number of SDSS galaxies. Considering this fraction and the previously estimated volume, it is possible to estimate, for each galaxy, its accessible volume. We repeat this calculation for the final MaNGA sample, deriving, for each galaxy, the weight to correct for its accessible volume (Vmax_w) and number (Num_w), which are included in the final catalog as rows 533 and 534, respectively. Figure 17 shows the comparison between the original stellar-mass function adopted to estimate the accessible volume for each galaxy (i.e., the SDSS one) and the one derived from our data using these estimated volume corrections. There is considerable agreement between both distributions in the regime at which the sample may be considered representative and statistically significant, i.e., for M_⋆ > 10⁹ M_⊙. Below this stellar mass, the current sample is clearly incomplete.

Zoom In Zoom Out Reset image size

Different groups have addressed the morphological classification of the MaNGA galaxies. Among them, we highlight the visual classification presented by Vazquez-Mata et al. (2022) as a MaNGA VAC, ¹⁹ which has been the basic training data set for our own classification. Domínguez Sánchez et al. (2022; hereafter, DS22) present a detailed state-of-the-art morphological classification based on supervised deep-learning models applied directly to SDSS images. ²⁰ They use two different visual classification catalogs to train their method (Nair & Abraham 2010; Willett et al. 2013). This analysis provides a T-type for each galaxy, together with the automatic identification of edge-on and barred galaxies. Our morphological classification is expected to provide a statistical segregation of the different morphological types, matching the visual classifications mentioned above. In Figure 18, we compare the results from both methods as a violin plot of the DS22 T-types versus our morphological classification (the best_type_n parameter). Despite the significant differences between both methods, the agreement between both morphological classifications for most types is remarkable. The largest differences are found at the extremes (best_type_n < 1 and best_type_n > 8). On the one hand, mild but clear differences are found in the merging spiral (Sm) and irregular (Irr) galaxy types. This is expected, since both schemes rely on different visual classifications of the galaxies. The more regular the galaxy, the better the expected agreement between both methods, as is apparent in Figure 18. On the other hand, the E and S0 galaxies in our catalog present very similar T-type distributions as in the DS22 catalog. We stress that DS22 consider that their T-type number is not enough to distinguish between these two morphological types, proposing a more complex decision tree based on additional parameters. ²¹ Our comparison indeed supports their results in this regard.

Zoom In Zoom Out Reset image size

The largest differences are found in the cD-class, which is included in our classification scheme as best_type_n = − 2, but absent from DS22. Based on this result, we consider that our classification for these galaxies is dubious. In some cases, such as manga-11968-6103, it may indeed be a cD galaxy from visual inspection. However, there are other galaxies, such as manga-8144-3703, which we consider should be classified as Irr (it was classified as Scd/Sd by DS22). In summary, the cD morphological type should be excluded from any further analysis. In any case, its removal has a negligible impact from a statistical point of view, since it only comprises eight galaxies. For the remaining galaxies, our morphological classification is at least as good as the published ones for statistical analysis.

As already mentioned, we have published previous versions of the analysis presented here, using an outdated version of the Pipe3D pipeline and a different set of SSP templates for the stellar decomposition analysis. In Sánchez et al. (2016a), we presented this analysis for the IFS data of the CALIFA DR2 data set (García-Benito et al. 2015). This distribution is a very limited catalog, with just 12 stellar and ionized gas characteristic/integrated properties, using a previous version of the Pipe3D data products for each of the 200 analyzed cubes (including only the SSP, SFH, FLUX_ELINES, and INDICES extensions). A similar analysis was presented in Sánchez et al. (2018) for the MaNGA DR14 data set for ∼2800 galaxies. In that case, we distributed a catalog ²² comprising almost 100 characteristic/integrated properties, in addition to the corresponding set of Pipe3D files (one per galaxy). Finally, we distributed exactly the same set of properties and data products for the MaNGA DR15 data set, ²³ including ∼4500 galaxies.

Here, we present a comparison between a set of selected properties derived for DR17 (the current analysis) with their DR15 counterparts, for the objects in common. Figure 19 shows this comparison for frequently used stellar parameters, including: (i) integrated stellar mass (M_⋆); (ii) average mass-to-light ratio (ϒ_⋆); (iii) SFR derived from the analysis of the stellar population (SFR_ssp); (iv) LW age at the effective radius (${{ \mathcal A }}_{\star ,L}$); (v) LW metallicity at the effective radius (${{ \mathcal Z }}_{\star ,L}$); (vi) average stellar dust extinction (A_⋆,V ); (vii) stellar velocity dispersion in the central aperture (σ_⋆,cen); (viii) stellar velocity at the effective radius (${v}_{\star ,1\mathrm{Re}}$); and (ix) apparent angular momentum (${\lambda }_{\star ,1\mathrm{Re}}$).

Zoom In Zoom Out Reset image size

The stellar mass, shown in the upper left panel, presents a small systematic offset of ∼−0.07 dex between the two distributions, with the DR15 masses being slightly larger than the DR17 ones. Considering that the two data sets have been reduced using different versions of the MaNGA DRP (v2_4_3 versus v3_1_1), and that we use a completely different SSP library, some systematic differences are expected. This difference is considerably smaller than the scatter between both values, characterized by a standard deviation σ_M⋆ ∼ 0.21 dex. Most of this scatter cannot be attributed to photometric differences, as the expected spectrophotometric precision and accuracy of the MaNGA data is of the order of ∼5% (Yan et al. 2016a, 2016b). Indeed, a simple comparison of the V-band photometry measured directly from the DR15 and DR17 data cubes yields a difference of ∼0.02 ± 0.05 dex. We consider that the scatter observed in M_⋆ is introduced by differences in the estimated ϒ_⋆, which are a combination of the use of a different stellar population template and, to a lesser extent, differences in the fitting procedure. This is evident from the comparison of the results for ϒ_⋆, shown in the upper middle panel of Figure 19. Although there is a clear correspondence between the two values, near a one-to-one relation for ϒ_⋆ < 7.5 M_⊙/L_⊙, there is a deviation, with the DR15 values being slightly lower than the DR17 values for ϒ_⋆ > 7.5 M_⊙/L_⊙. This introduces a bias of −0.26 M_⊙ L ${}_{\odot }^{-1}$ in Δϒ_⋆. The considerable scatter in the distribution, with σ_ϒ⋆ = 1.41 M_⊙ L ${}_{\odot }^{-1}$ in Δϒ_⋆, corresponding to ∼0.15 dex on average, is clearly the main driver for the observed differences in the stellar masses. This scatter is propagated to any other quantity that requires the use of ϒ_⋆, like the SFR derived from the stellar population spectral decomposition (Section 5.3.4), shown in the upper right panel of Figure 19. In this case, we compare the SFR_ssp estimated at 32 Myr. There is clear agreement between the DR17 and DR15 SFR values, following on average a one-to-one relation. This is reflected in the small average difference (∼−0.03 dex). However, the dispersion around the one-to-one relation, ∼0.37 dex, is larger than the dispersions reported for both ϒ_⋆ and M_⋆. This is expected, since the derivation of SFR depends directly on the derivation of the two previous parameters, with the additional uncertainty of estimating the stellar mass for just a tiny fraction of the total stellar component (younger than 32 Myr, according to Equation (8)).

The center left panel of Figure 19 shows the comparison of the LW stellar age. Again, there is a correspondence between the DR15 and DR17 values. However, the relation is not just on top of the one-to-one line. There is a clear systematic offset of ∼0.1 dex, with the ages derived for DR15 being older than those for DR17. The difference is small compared to the dynamical range of the parameter (∼2 dex), but it is well determined, even considering the scatter around the central relation, ${\sigma }_{{{ \mathcal A }}_{\star }}$ = 0.14 dex. We built the age sampling of the new SSP template following a philosophy similar to the one that was adopted for generating the GSD156 template (a pseudologarithmic sampling of the time steps). However, neither the detailed set of ages nor the range of metallicities (and their sampling) are equal between the two sets of SSPs. In addition, the two templates use different stellar libraries (MILES versus MaStar), different sets of isochrones, and were computed by two different synthesis codes. All together, this can easily explain the observed age differences, which, in any case, are rather small.

Larger differences are found for the LW stellar metallicity (the center middle panel of Figure 19). For this parameter, we find a one-to-one correspondence between the two data sets for the stellar ${{ \mathcal Z }}_{\star ,L}$ _,DR17 > −0.5 dex. Below this value, the distribution bends toward a plateau for the DR15 metallicities at ∼−0.3 dex. Despite this deviation from the one-to-one relation, the average difference between the two values is rather small, with a Δ${{ \mathcal Z }}_{\star ,L}$ = 0.06 ± 0.18 dex. The reported differences are most probably due to the different sampling and coverage of stellar metallicity in the newly adopted SSP template, ranging from ${{ \mathcal Z }}_{\star }$ = −2.30 to 0.30 dex, sampled for seven metallicities, compared to the DR15 template, ranging from −0.73 to 0.20 dex, sampled for four metallicities. However, this alone cannot explain the observed difference, since the plateau is not reached at the minimum sampled metallicity for the GSD156. A combination of the difference in the metallicity coverage, the different stellar libraries, the adopted isochrones, and the different synthesis codes must be behind the observed behavior. As a result of this difference, similar qualitative results will be found when using the DR15 and DR17 ${{ \mathcal Z }}_{\star ,L}$ values, but the quantitative results will change. Global relations, like the stellar MZR, or spatially resolved trends, like radial gradients, will present a wider dynamical range for the DR17 data set. However, we do not expect strong changes, either in the shapes of the relations, or in the signs of the gradients.

The right middle panel of Figure 19 shows the comparison between the dust extinction for the two data sets. As in the case of the two previous parameters, there is a well-defined correspondence between the two values. The best agreement is found for low dust extinction values (A_V < 0.3 mag), with a distribution near the one-to-one relation. For larger values, the A_V reported in DR15 is slightly higher than for DR17, which translates into a positive ΔA_V = 0.07 ± 0.17 mag. Again, despite the lack of a one-to-one correspondence, for the three cases shown in the central panels these differences are rather small, being just two or three times larger than the expected errors for the parameters (Figure 6 of Lacerda et al. 2022).

In the lower panels of Figure 19, we compare the DR15 and DR17 results for the kinematical parameters: σ_⋆,cen, vel${}_{\star ,\mathrm{Re}}$, and ${\lambda }_{\star ,\mathrm{Re}}$. For σ_⋆,cen and vel${}_{\star ,\mathrm{Re}}$, the distributions follow a one-to-one relation, with offsets lower than the reported standard deviations: Δσ_⋆,cen = −12 ± 37 km s⁻¹ and Δvel_⋆,cen = −17 ± 61 km s⁻¹. In the case of σ_⋆,cen, the scatter of the difference is of the order of the expected errors for this parameter, based on simulations, ∼22 km s⁻¹ (Lacerda et al. 2022), and slightly smaller than the instrumental resolution, ∼75 km s⁻¹ (e.g., Law et al. 2021). For the stellar velocity at the effective radius, the scatter of the difference is three times larger than the expected errors, ∼20 km s⁻¹ (Lacerda et al. 2022), being of the order of the instrumental resolution. In both cases, the differences are larger for the lowest values of the parameters, in particular for the velocity. In general, we do not anticipate any qualitative or quantitative differences in the results derived from both data sets relating to these two kinematic parameters. Larger differences between the DR15 and DR17 values are found for ${\lambda }_{\star ,\mathrm{Re}}$, due to the inclination correction described in Section 5.3.6 (Equation (15)). Once corrected, ${\lambda }_{\star ,\mathrm{Re},\mathrm{DR}15}^{{\prime} }$ follows the DR17 results along the one-to-one relation (the lower right panel of Figure 19). The two values agree within ${\rm{\Delta }}{\lambda }_{\star ,\mathrm{Re}}^{{\prime} }$ = 0.4 ± 0.19, with the larger discrepancies being found for the larger values of the angular momentum.

Figure 20 shows the same type of comparison for a selected subset of the parameters derived from the ionized gas emission lines included in the final catalog (Section 5.3.5). We compare: (i) the Hα flux (F_Hα,cen); (ii) the equivalent width of Hα (EW_Hα,cen); (iii) the Hα-to-Hβ line ratio; (iv) the [O iii]-to-Hα line ratio; (v) the [N ii]-to-Hα line ratio; and (vi) the [S ii]-to-Hα line ratio, all measured in the central aperture. Additionally, we include: (vii) the average dust extinction across the entire galaxy derived from the Hα-to-Hβ line ratio; (viii) the integrated SFR derived from the Hα luminosity; and (ix) the molecular gas integrated mass derived from the dust-to-gas ratio. In most cases, we find quite good agreement between the two estimates of the parameter. In the case of F_Hα,cen, there is a tight one-to-one relation, with an offset lower than the standard deviation over most of the dynamical range. Most of the scatter and the deviation from the one-to-one relation is found at very low flux intensities, where the differences are driven by the accuracy of the subtraction of the underlying stellar population, rather than by the properties of the emission lines themselves (e.g., their S/Ns). Indeed, the scatter presents a standard deviation of ∼0.34 dex, three times larger than expected, based on the emission-line properties (∼0.1 dex; Figure 9 of Lacerda et al. 2022). For log(F_Hα,cen,DR17) < −1.5 dex, the distribution bends, with the DR15 values reaching a plateau. The newly adopted SSP template is based on a stellar library observed with the same instrument and reduced with the same tools, and therefore its spectral resolution should better match that of the observed data. Therefore, we consider that the new derivation of the emission-line fluxes is more reliable, in particular in the central regions, where the flux intensity of the emission lines is low, in general, and the stellar component is brighter and older (i.e., the worst-case scenario for a proper estimate of the emission-line properties). For the same reasons, a similar trend is observed for the EW(Hα), although in this case the agreement with respect to the one-to-one relation seems to be slightly better, with a Δlog (EW(Hα)) = 0.07 ± 0.19.

Zoom In Zoom Out Reset image size

We find similar results for the four line ratios in Figure 20. In all cases, the bulk of the distribution is located around the one-to-one relation, with tiny offsets of ∼0.0–0.5 dex and a scatter of ∼0.13 dex. The largest differences are found for the Hα/Hβ ratio, which shows a tail toward larger values for DR17 compared to DR15 below <2.8. This tail is not observed in the other line ratios, indicating that it is most probably related to an effect that affects the Balmer lines simultaneously. We consider that the differences in the SSP templates are behind this effect. In general, based on these results, none of the conclusions regarding the ionization conditions, and most probably none of the properties derived from the line ratios (e.g., the oxygen abundance and the ionization parameter), should be affected by the choice of either the DR15 or the DR17 data products.

The lower row of Figure 20 shows the comparison for a set of physical properties derived from the emission-line parameters. The lower left panel shows the comparison of the dust extinction at the effective radius, where we find a systematic offset of ΔA_gas,V = 0.24 ± 0.27 mag. We do not have a clear explanation for the origin of this difference, since, in principle, all the analysis was repeated following exactly the same steps. We can only guess that the masking associated with the minimum S/N required to compute the radial distribution of the dust extinction has introduced this change. In any case, this offset has not introduced any significant changes in the integrated properties included in the two final panels of this figure. The lower middle panel of Figure 20 shows the comparison between the integrated SFR, derived using the dust-corrected Hα luminosity. As for the rest of the properties, the distribution lies around the one-to-one relation, in particular for SFR_Hα,DR17 > 10⁻² M_⊙ yr⁻¹, i.e., for most SFGs. At low SFRs, the DR15 values are slightly larger, following a similar trend (and most probably for the same reasons) as the distributions for F_Hα,cen and EW(Hα). Finally, the lower right panel of Figure 20 shows the comparison between the molecular gas mass estimated from the dust extinction. In this particular case, we use the log_Mass_gas parameter listed in the catalog (Section 5.3.5). The distribution is centered on the one-to-one relation. However, this parameter shows a larger scatter than the previous one (∼0.48 dex), reflecting a considerable variation in the spaxel-by-spaxel dust extinction between the two DRs. Since this parameter is derived from the Hα/Hβ ratio, from the discussion regarding the top right panel of the figure, most likely this scatter is due to differences in the subtraction of the underlying stellar population introduced by the new SSP library (and its effects on the estimates of the shape and depth of the Hα and Hβ stellar absorption).

As already mentioned, the MaNGA DR17 data have been analyzed using other tools. In particular, all the data set was processed using the MaNGA DAP (Westfall et al. 2019). ²⁴ This tool performs a decomposition of the stellar and ionized gas components of the observed spectra, using different spatial binning schemes, delivering the spatial distribution for a set of properties of the emission lines, a set of stellar indices, and the stellar kinematic properties. From these data products, a set of characteristic values for each galaxy estimated at the central regions (for the emission lines) and at the effective radius (for the stellar indices) are extracted, which are then integrated into the DAPall database. In this database, parameters for different analyses are also included, comprising: (i) a treatment of the stellar and ionized gas on individual spectra, spaxel by spaxel; (ii) an analysis of both the stellar and ionized gas using a spatial binning scheme; and (iii) a hybrid analysis, in which the stellar population is analyzed by adopting a binning scheme and the ionized gas is studied spaxel by spaxel. In addition, a different combination of stellar libraries is used to explore the stellar kinematics and to decouple the stellar and ionized gas components, as well as both a Gaussian fitting and a moment analysis for the exploration of the emission lines. The most similar approach to the one performed by pyPipe3D is the hybrid method, which uses a combination of the MILES stellar library for the kinematics and the MaStar SSP library for decoupling the stellar and ionized gas components (the HYB10-MILESHC-MASTARSSP data set in their nomenclature). We compare our DR17 results with this catalog.

Figure 21 shows the comparison of the stellar indices in common between the two data sets. In general, we find a good match between the DAP and pyPipe3D. The best agreement is found for D4000. For this parameter, the pyPipe3D value has been corrected by the scaling factor proposed by Gorgas et al. (1999), transforming the index derived from a flux density in units of wavelength to a flux density in units of frequency (see Section 4.1). For the DAP value, we adopt the average between the two listed parameters, D4000 and Dn4000, since this combination provides the best comparison with our results. There is a tight one-to-one relation between the two estimates, with ΔD4000 = −0.01 ± 0.07. For the remaining spectral indices, there is a clear correspondence between the DAP and pyPipe3D, following a distribution on top of the one-to-one relation (e.g., Fe5270) or just showing a systematic offset with respect to that relation. In the latter cases, the offsets range between −0.47 Å for Mgb and 0.21 Å for H_β . The scatter around these relations is of the order of ∼0.6 Å, corresponding to ∼20%–30% of the typical index value.

Zoom In Zoom Out Reset image size

The main driver of the observed differences is not the definition of the passbands adopted to derive the indices, since a cross-check of our adopted values (Table 5) with the ones published in Westfall et al. (2019) indicates that there is no difference down to the third decimal. Therefore, the difference should be in the details of the procedure. For instance, the DAP adopts a different binning scheme than the one implemented by our code. Thus, the indices are not calculated for exactly the same spectra. For indices affected by the subtraction of the emission lines, like Hβ, there is an additional source of discrepancy, as the treatment of the emission lines is different in the DAP and pyPipe3D. However, both effects should contribute to the observed scatter, but we have found some clear systematic offsets (e.g., Mgb). We do not have a definitive explanation for these offsets, although we suspect that the preprocessing of the data performed as part of our analysis (Section 3), which involves the homogenization of the spectral resolution, may be behind the observed differences. In any case, despite the reported offsets and discrepancies, the similarities between both sets of indices are such that no significant differences should be introduced by using either the DAP or the pyPipe3D indices in further analysis.

Figure 22 shows the comparison of the same set of emission-line properties derived by the DAP and pyPipe3D for the central regions of the galaxies shown in Figure 20. In general, we find the same differences/similarities between the parameters derived using the DAP and pyPipe3D as when we compare the results of DR15 and DR17 (Section 6.2). In some cases, like F_Hα,cen and [S ii]/Hα, the systematic offset is smaller than the one found in the previous comparison. In other cases, like [O iii]/Hβ and [N ii]/Hα, they are slightly larger. However, in no case are the offsets significant compared with the standard deviation of the difference between the two sets of values. The larger differences are found for: (i) EW(Hα), for values lower than 1 Å (measured by pyPipe3D), a range for which the DAP predicts slightly larger values; (ii) Hα/Hβ ratio for values lower than 2.5 (measured by the DAP), a range for which pyPipe3D derives slightly larger values; (iii) F_Hα,cen, for values lower than 0.1 10⁻¹⁶ erg s⁻¹ cm⁻² (measured by pyPipe3D), a regime for which the DAP predicts slightly larger values; and (iv) the [S ii]/Hα ratio, for values larger than 0 dex (measured by pyPipe3D), where the DAP predicts slightly lower values. For cases (i) and (ii), the differences occur in regimes of very low intensity (and S/N) of the emission lines, where any measurement is unreliable. For case (iii), the difference is irrelevant in most of the calculations, since this line ratio is usually adopted to estimate the dust extinction (e.g., Section 5.3.5), and the considered regime corresponds to either unphysical values (Hα/H β < 2.5) or a regime of very low dust extinction. Finally, for case (iv), the difference is so small than it should not affect any further analysis. In summary, we consider that any analysis using the emission-line fluxes derived by both procedures should provide very consistent results.

Zoom In Zoom Out Reset image size

The final parameters in common between the DAP and pyPipe3D correspond to the kinematical properties of galaxies. Figure 23 shows the comparison of the stellar σ_⋆ (left panel) and Hα (σ_Hα ; central panel) velocity dispersions estimated within 1 Re. In both cases, there is good correspondence between the DAP and DR17 estimates. However, there are noticeable differences. In the case of σ_⋆, the DAP values are slightly larger (∼12 km s⁻¹), showing a systematic offset that is evident, even though it is smaller than the scatter (∼27 km s⁻¹). Furthermore, there is a clear bend in the distribution for σ_⋆,DR17 < 70 km s⁻¹, where the DAP values reach a plateau at ∼50 km s⁻¹, while the pyPipe3D values are still decreasing. These bends occur when there is a discrepancy in the treatment of the instrumental velocity dispersion as a consequence of a mismatch between the spectral resolution of the adopted SSPs or stellar templates and the observations. In our case, we use SSPs based on the MaStar library, and therefore they have a similar spectral resolution as the analyzed data. However, the DAP adopts a subset of the MILES stellar library for the kinematics analysis. Without prejudging which of the two approaches provides a more realistic estimate of the velocity dispersion, we just note that the difference in the adopted procedure may explain the observed differences at low σ_⋆.

Zoom In Zoom Out Reset image size

In order to get an independent judgment of the quality of the pyPipe3D estimates of σ_⋆, we compare them with the values expected from the stellar masses of early-type galaxies, as predicted by the well-known Faber–Jackson (FJ) relation (Faber & Jackson 1976). The right panel of Figure 23 shows the distribution of σ_⋆ measured within the effective radius by pyPipe3D, together with the FJ relation estimated by Aquino-Ortíz et al. (2018) for the same aperture. We find that the DR17 results follow the FJ relation down to ∼60 km s⁻¹, a value that roughly corresponds to the spectral resolution of the data. Thus, we consider that our DR17 values of σ_⋆ are reliable at least above this limit. In order to provide a direct comparison between the two estimates of the velocity dispersion, we derive a prescription to transform from one to the other, by following the equation

$$\begin{eqnarray}&&{\sigma }_{\star ,\mathrm{DAP}}=1.15\sqrt{{\sigma }_{\star ,\mathrm{pyPipe}3{\rm{D}}}^{2}+{75}^{2}}-41.\end{eqnarray} \tag{ 16 }$$

After applying this transformation, the two estimates of the velocity dipersion show a one-to-one relation, with Δσ_⋆ = 0.3 ± 22 km s⁻¹, a scatter similar to the expected error in this quantity (Lacerda et al. 2022).

Finally, for the ionized gas velocity dispersion, we find a systematic offset between the values reported by the DAP and pyPipe3D of Δσ_Hα = −16 ± 41 km s⁻¹. We recall that our estimate is based on an analysis of the emission lines included in the FLUX_ELINES extension of the Pipe3D FITs file. In this extension, we include the velocity dispersion measured in angstroms (Table 7). We realized that for the published table we underestimated the instrumental dispersion by 45 km s⁻¹, leading to the observed discrepancy. Therefore, for any further analysis, we recommend applying the following correction to the Hα velocity dispersion included in the final catalog:

$$\begin{eqnarray}&&{\sigma }_{{\rm{H}}\alpha }^{{\prime} }=\sqrt{{\sigma }_{{\rm{H}}\alpha }^{2}-{45}^{2}}.\end{eqnarray} \tag{ 17 }$$

Once corrected, the two quantities agree within a standard deviation, ∼27 km s⁻¹.

The MaNGA DR17 data set has been analyzed using the FIREFLY (Wilkinson et al. 2017) full spectral fitting code to derive the main spatially resolved and stellar population properties of each individual galaxy. Like pyPipe3D, FIREFLY implements a chi-square minimization fitting code that combines a template of SSPs to model the observed spectra. A Bayesian information criterion for selecting the best-fitting model, without assuming any priors, is adopted. Beside this, the major difference with respect to pyPipe3D is the treatment of dust extinction. Although FIREFLY includes a method whereby extinction by dust is included as part of the stellar decomposition process, the default procedure involves the preprocessing of both the observed spectra and the SSP templates, which removes their shapes (over a bandwidth of ∼100 Å), and from this first best-fitted model the dust extinction is derived. This analysis, applied to MaNGA DR17, provides a full set of spatially resolved properties for the stellar populations, including the corresponding maps for LW and MW ${{ \mathcal A }}_{\star }$ and ${{ \mathcal Z }}_{\star }$, ²⁵ from which the values at the effective radius and the slopes of their radial gradients are extracted (Neumann et al. 2022). ²⁶ Here, we compare the values derived for the LW parameters by FIREFLY and pyPipe3D. In this comparison, we should bear in mind that the FIREFLY data products were derived using: (i) a different SSP template (Maraston et al. 2020) than the one adopted in our analysis, although we selected results based on the same stellar library for this comparison (i.e., MaStar); (ii) the spatial binning and the results of the kinematical analysis derived by the DAP (Section 6.3); (iii) a different wavelength to weight the stellar populations—the average observed wavelength (∼6600 Å), instead of the rest-frame 5500 Å adopted by pyPipe3D; (iv) a different method to derive the LW values, as they adopted an arithmetic weighted average (according to Neumann et al. 2022, Equation (2), although this was not the case in previous versions of the code, e.g., Goddard et al. 2017); and (v) a different procedure to derive the radial gradients. As indicated in Section 5.3, our method excludes the central regions of the galaxies (<0.5 Re), since we consider that they are usually affected by PSF/beam effects. Furthermore, the gradient is derived up to 2.0 Re (or the maximum extension covered by the FOV). On the contrary, the values reported by FIREFLY are derived for the region within 1.5 Re. In both cases, an azimuthal average is performed (contrary to previous derivations of this parameter by FIREFLY, e.g., Figure 8 and Section 3.2 of Goddard et al. 2017).

Figure 24 compares the FIREFLY and pyPipe3D values of ${{ \mathcal A }}_{\star ,L}$ and ${{ \mathcal Z }}_{\star ,L}$, measured at the effective radius, and the slopes of their corresponding gradients. ²⁷ For age and metallicity, we observe a clear correspondence between the values reported from both methods, but not a one-to-one relation. On average, FIREFLY derives older stellar populations than pyPipe3D, a bias that is stronger for the younger stellar populations (with a difference of ∼0.5 dex) than for the older ones (with a difference of ∼0.1 dex). As a result, we find a positive offset of ${\rm{\Delta }}{{ \mathcal A }}_{\star ,L}\sim 0.25$ dex, with a scatter of ∼0.25 dex. Both the offset and the scatter are larger than the values found in the comparison for the same parameter in Section 6.2, and larger than for any of our previous comparisons between this tool and other stellar population analysis techniques (e.g., Figure 17 and Section 4 of Sánchez et al. (2016a)). A similar pattern is observed for the metallicity: FIREFLY provides larger values than pyPipe3D, although in this case the offset seems to be similar (or at least of the same order) for low and high metallicities. On average, we find ${\rm{\Delta }}{{ \mathcal Z }}_{\star ,L}=$ 0.21 ± 0.16 dex.

Zoom In Zoom Out Reset image size

The reported discrepancies are a consequence of the combined effects of the differences in the adopted procedures listed in the previous paragraph. Systematic differences are expected just due to the use of a different SSP library (e.g., the appendices in González Delgado et al. 2014, 2015; Sánchez et al. 2016b), differences that are enhanced by the adoption of a different binning scheme (e.g., Ibarra-Medel et al. 2019), a different method for deriving the mean values, and the normalization at a redder wavelength (which enhances the older and more metal-rich stellar populations). In any case, despite these quantitative differences, qualitatively both procedures classify the old (metal-rich) and young (metal-poor) populations in similar ways.

Stronger differences are found for the radial gradients (the right panels of Figure 24). The gradients derived using the two methods present a less clear correspondence to one another than the parameters discussed previously, with a scatter of the same order of the dynamical range covered by the gradients. The ${{ \mathcal A }}_{\star ,L}$ gradient presents a somehow better correspondence, with FIREFLY deriving slightly sharper gradients than pyPipe3D: ${\rm{\Delta }}{\rm{\nabla }}{{ \mathcal A }}_{\star ,L}$ = −0.14 ± 0.27 dex/Re. On the contrary, the relation between the ${{ \mathcal Z }}_{\star ,L}$ gradients is very loose, with an average difference of ${\rm{\Delta }}{\rm{\nabla }}{{ \mathcal Z }}_{\star ,L}$ = 0.09 ± 0.13 dex/Re, and the FIREFLY gradients tending to be shallower.

We consider that the combination of the differences in the FIREFLY procedure outlined above, and, in particular, the use of a different scheme to derive the gradient, can easily explain the reported discrepancies between the FIREFLY and the pyPipe3D gradients. It is worth noticing that despite these differences, both methods predict a slightly sharper (shallower) gradient for both quantities at higher (lower) stellar masses, in agreement with previous results (e.g., González Delgado et al. 2015; Sánchez et al. 2021, and references therein). Thus, both methods provide similar qualitative results in a statistical sense, despite the large reported differences in the gradients for individual galaxies.

We use the first release of MaStar (Yan et al. 2019) to build a test set of the C&B SSP models. This release of the MaStar library contains 8646 spectra of 3321 unique stars. The spectra cover the wavelength range from 3622 to 10354 Å, at a resolving power of R ≈ 1800. Due to the lack of a uniform calibration of the astrophysical parameters for the stars in this library, we opted for a quick and simple manner when assigning a MaStar spectrum to each star in the synthesis model. For each MILES spectrum used in our spectral synthesis, we searched for the closest matching spectrum (in log flux) in the MaStar library. We then replaced the MILES and IndoUS spectra (see lines 3 and 4 in Table 12) with the selected MaStar spectrum, over the full range covered by the latter. We checked that there were no systematic differences between the colors and the line strength indices computed for both sets of models as a function of time. The MaStar SSP models were computed for all the metallicities listed in Table 8. As its MILES counterpart, each MaStar SSP model contains 220 spectra at time steps from 0 to 14 Gyr. The subset of the MaStar C&B templates adopted throughout this article in the pyPipe3D format are accesible through the web. ³⁰

Recently, Mejia-Narvaez et al. (2022) have determined the stellar atmospheric parameters for the latest release of MaStar, comprising over 22 k unique stars. At the moment, we are working on the implementation of this fully calibrated MaStar library in the C&B models.