The CALIFA survey across the Hubble sequence:

The CALIFA survey across the Hubble sequence:

Spatially resolved stellar population properties in galaxies
R. M. González Delgado Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    R. García-Benito Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    E. Pérez Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    R. Cid Fernandes Departamento de Física, Universidade Federal de Santa Catarina, P.O. Box 476, 88040-900, Florianópolis, SC, Brazil    A. L. de Amorim Departamento de Física, Universidade Federal de Santa Catarina, P.O. Box 476, 88040-900, Florianópolis, SC, Brazil    C. Cortijo-Ferrero Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    E. A. D. Lacerda Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( ) Departamento de Física, Universidade Federal de Santa Catarina, P.O. Box 476, 88040-900, Florianópolis, SC, Brazil    R. López Fernández Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    N. Vale-Asari Departamento de Física, Universidade Federal de Santa Catarina, P.O. Box 476, 88040-900, Florianópolis, SC, Brazil    S. F. Sánchez Instituto de Astronomía,Universidad Nacional Autonóma de Mexico, A.P. 70-264, 04510, México,D.F.    M. Mollá Departamento de Investigación Básica, CIEMAT, Avda. Complutense 40, E-28040 Madrid, Spain    T. Ruiz-Lara Departamento de Física Teórica y del Cosmos, University of Granada, Facultad de Ciencias (Edificio Mecenas), E-18071 Granada, Spain    P. Sánchez-Blázquez Depto. de Física Teórica, Universidad Autónoma de Madrid, 28049 Madrid, Spain    C. J. Walcher Leibniz-Institut für Astrophysik Potsdam (AIP), An der Sternwarte 16, D-14482 Potsdam, Germany    J. Alves University of Vienna, Türkenschanzstrasse 17, 1180, Vienna, Austria    J. A. L. Aguerri Instituto de Astrofísica de Canarias (IAC), E-38205 La Laguna, Tenerife, Spain    S. Bekeraité Leibniz-Institut für Astrophysik Potsdam (AIP), An der Sternwarte 16, D-14482 Potsdam, Germany    J. Bland-Hawthorn Sydney Institute for Astronomy, The University of Sydney, NSW 2006, Australia    L. Galbany Millennium Institute of Astrophysics and Departamento de Astronomía, Universidad de Chile, Casilla 36-D, Santiago, Chile Departamento de Astronomía, Universidad de Chile, Casilla 36-D, Santiago, Chile    A. Gallazzi INAF Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, 50125 Firenze, Italy    B. Husemann European Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching b. München, Germany    J. Iglesias-Páramo Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( ) Estación Experimental de Zonas Aridas (CSIC), Ctra. de Sacramento s/n, La Cañada, Almería, Spain    V. Kalinova Department of Physics 4-181 CCIS, University of Alberta, Edmonton AB T6G 2E1, Canada    A. R. López-Sánchez Australian Astronomical Observatory, PO BOX 296, Epping, 1710 NSW, Australia    R. A. Marino CEI Campus Moncloa, UCM-UPM, Departamento de Astrofísica y CC. de la Atmósfera, Facultad de CC. Físicas, Universidad Complutense de Madrid, Avda. Complutense s/n, 28040 Madrid, Spain    I. Márquez Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    J. Masegosa Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    D. Mast Instituto de Cosmologia, Relatividade e Astrofísica - ICRA, Centro Brasileiro de Pesquisas Físicas, Rua Dr.Xavier Sigaud 150, CEP 22290-180, Rio de Janeiro, RJ, Brazil    J. Méndez-Abreu School of Physics and Astronomy, University of St. Andrews, North Haugh, St. Andrews, KY169SS, UK    A. Mendoza Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    A. del Olmo Instituto de Astrofísica de Andalucía (CSIC), P.O. Box 3004, 18080 Granada, Spain. ( )    I. Pérez Departamento de Física Teórica y del Cosmos, University of Granada, Facultad de Ciencias (Edificio Mecenas), E-18071 Granada, Spain    A. Quirrenbach Landessternwarte, Zentrum fur Astronomie der Universitat Heidelberg, Königstuhl 12, D-69117 Heidelberg, Germany    S. Zibetti Departamento de Astronomía, Universidad de Chile, Casilla 36-D, Santiago, Chile    CALIFA collaboration
Feb. 2015
Key Words.:
Techniques: Integral Field Spectroscopy; galaxies: evolution; galaxies: stellar content; galaxies: structure; galaxies: fundamental parameters; galaxies: bulges; galaxies: spiral

Various different physical processes contribute to the star formation and stellar mass assembly histories of galaxies. One important approach to understand the significance of these different processes on galaxy evolution is the study of the stellar population content of today’s galaxies in a spatially resolved manner. The aim of this paper is to characterize in detail the radial structure of stellar population properties of galaxies in the nearby universe, based on a uniquely large galaxy sample considering the quality and coverage of the data. The sample under study was drawn from the CALIFA survey and contains 300 galaxies observed with integral field spectroscopy. These cover a wide range of Hubble types, from spheroids to spiral galaxies, while stellar masses range from to . We apply the fossil record method based on spectral synthesis techniques to recover the following physical properties for each spatial resolution element in our target galaxies: the stellar mass surface density (), stellar extinction (), light-weighted and mass-weighted ages (, ), and mass-weighted metallicity (). To study mean trends with overall galaxy properties, the individual radial profiles are stacked in seven bins of galaxy morphology (E, S0, Sa, Sb, Sbc, Sc and Sd). We confirm that more massive galaxies are more compact, older, more metal rich, and less reddened by dust. Additionally, we find that these trends are preserved spatially with the radial distance to the nucleus. Deviations from these relations appear correlated with Hubble type: earlier types are more compact, older, and more metal rich for a given M, which evidences that quenching is related to morphology, but not driven by mass. Negative gradients of  are consistent with an inside-out growth of galaxies, with the largest  gradients in Sb–Sbc galaxies. Further, the mean stellar ages of disks and bulges are correlated, with disks covering a wider range of ages, and late type spirals hosting younger disks. However, age gradients are only mildly negative or flat beyond HLR, indicating that star formation is more uniformly distributed or that stellar migration is important at these distances. The gradients in stellar mass surface density depend mostly on stellar mass, in the sense that more massive galaxies are more centrally concentrated. Whatever sets the concentration indices of galaxies obviously depends less on quenching / morphology than on the depth of the potential well. There is a secondary correlation in the sense that at the same early type galaxies have steeper gradients. The gradients outside 1 HLR show no dependence on Hubble type. We find mildly negative  gradients, shallower than predicted from models of galaxy evolution in isolation. In general, metallicity gradients depend on stellar mass, and less on morphology, hinting that metallicity is affected by the depth of both - potential well and morphology/quenching. Thus, the largest  gradients occur in Milky Way-like Sb–Sbc galaxies, and are similar to those measured above the Galactic disk. Sc spirals show flatter  gradients, possibly indicating a larger contribution from secular evolution in disks. The galaxies from the sample have decreasing-outwards stellar extinction; all spirals show similar radial profiles, independent from the stellar mass, but redder than E’s and S0’s. Overall we conclude that quenching processes act in manners that are independent of mass, while metallicity and galaxy structure are influenced by mass-dependent processes.

1 Introduction

Galaxies are a complex mix of stars, gas, dust, and dark matter, distributed in different components (bulge, disk, and halo) whose present day structure and dynamics are intimately linked to their assembly and evolution over the history of the Universe. Different observational and theoretical approaches can be followed to learn how galaxies form and evolve.

Theoretically, the formation of large-scale structures arise through the evolution of cold dark matter. In this picture, small-scale density perturbations in the dark matter collapse and form the first generation of dark matter halos, that subsequently merge to form larger structures such as clusters and superclusters (Springel et al. 2005; De Lucia et al. 2006). This basic hierarchical picture is able to explain the global evolution of the star formation rate density of the universe, with galaxy peak formation epoch at redshift 2–3 (e.g. Madau & Dickinson 2014, and references therein). The stellar components formed at earlier epochs likely evolved into elliptical galaxies and bulges through mergers of the primordial star-forming disks (Elmegreen et al. 2007; Bournaud et al. 2007). However, this framework fails to explain how the galaxy population emerges at , and how the present day Hubble sequence of galaxies was assembled.

The growth of galaxies is not related in a simple way to the build up of dark matter; the interplay of energy and matter exchange (between the process of gas accretion and cooling and star formation) is essential to grow the gaseous and stellar components in galaxies. Feedback mechanisms resulting from stellar winds, supernova explosions, and AGN are relevant to stop the gas collapse and cooling, quenching the star formation and hence galaxy growth (Silk & Rees 1998; Hopkins et al. 2011). Although these processes are difficult to implement in theoretical models, they are essential to explain the masses, structures, morphologies, stellar populations, and chemical compositions of galaxies, and the evolution of these properties with cosmic time.

Recently, a new set of cosmological hydrodynamic simulations have started to predict how the spatially resolved information of the properties of stellar populations in galaxies can constrain the complex interplay between gas infall, outflows, stellar migration, radial gas flows, and star formation efficiency, in driving the inside-out growth of galactic disks (Brook et al. 2012; Gibson et al. 2013; Few et al. 2012; Pilkington et al. 2012a; Minchev et al. 2014). Radiative cooling, star formation, feedback from supernovae, and chemical enrichment are also included in simulations to predict radial metallicity gradients as a function of merging history. Shallow metallicity gradients are expected if elliptical galaxies result from major mergers (e.g. Kobayashi 2004), but a minor merger picture for the formation of ellipticals can successfully explain the strong size evolution of massive galaxies (Oser et al. 2012). This late-time accretion of low mass and metal poor galaxies (dry mergers) into the already formed massive galaxy can produce a variation of the age and metallicity radial structure of the galaxy as it increases in size. Galactic stellar winds and metal cooling have also an important effect on these ex-situ star formation models, predicting different behaviour of the mass and metallicity assembly in massive early type galaxies, and in the radial gradient of present stellar populations properties of galaxies (Hirschmann et al. 2013, 2015).

In summary, these theoretical works show that observational data with spatial information of the mass and metallicity assembly and their cosmic evolution, and the present radial structure of stellar population properties (stellar mass surface density, age, metallicity) contain relevant information to constrain the formation history of galaxies, and the physics of feedback mechanisms involved.

Observationally, a first step is to study what kinds of galaxies are there in the Universe, and which are their physical properties. Attending to their form and structure, galaxies can be grouped into a few categories. Results show that most of the massive nearby galaxies are ellipticals, S0’s, or spirals (Blanton & Moustakas 2009) following the Hubble tuning fork diagram. In this scheme, S0’s are a transition between spirals and ellipticals (Cappellari et al. 2013), and the bulge/disk ratio increases from late to early type spirals. At the same time, galaxy properties such as color, mass, surface brightness, luminosity, and gas fraction are correlated with Hubble type (Roberts & Haynes 1994), suggesting that the Hubble sequence somehow reflects possible paths for galaxy formation and evolution. However, the processes structuring galaxies along the Hubble sequence are still poorly understood.

Integral Field Spectroscopy (IFS) enables a leap forward, providing 3D information (2D spatial + 1D spectral) on galaxies. Such datacubes allow one to recover two-dimensional maps of stellar mass surface density, stellar ages, metallicities, extinction and kinematics, as well as a suit of nebular properties such as gas kinematics, metallicity, excitation, and etc. Until a few years ago IFS was used to target small samples of galaxies. Detailed programs such as SAURON (Bacon et al. 2001), VENGA (Blanc et al. 2013), (U)LIRs at z 0.26 (Arribas et al. 2010), PINGS (Rosales-Ortega et al. 2010), or DiskMass Survey (Bershady et al. 2010), have been limited to less than a hundred galaxies, but it is more than fair to recognize that to get these amounts of IFU data was a challenge at the time. ATLAS3D (Cappellari et al. 2011) represented a step forward, with the observation of a volume-limited sample of 260 galaxies, but with three important limitations: the sample only includes early-type galaxies, the field of view is limited to 1 effective radius, and the spectral range is restricted from H to [NI]5200.

CALIFA (Calar Alto Legacy Integral Field Area) is our ongoing survey of 600 nearby galaxies at the 3.5m at Calar Alto (Sánchez et al. 2012)111 The data set provided by the survey (see Husemann et al. 2013 for DR1; García-Benito et al. 2015 for DR2) is unique to advance in these issues not only because of its ability to provide spectral and spatial information, but also because: a) It includes a large homogeneous sample of galaxies across the color-magnitude diagram, covering a large range of masses ( to , González Delgado et al. 2014c), and morphologies from Ellipticals (E0-E7), Lenticulars (S0-S0a), to Spirals (Sa to Sm) (see Walcher et al. (2014) for a general description of the sample). b) It has a large field of view () with a final spatial sampling of 1 arcsec, and a resolution of arcsec, allowing to spatially resolve well the stellar population properties, and to obtain the total integrated properties, such as galaxy stellar mass, and stellar metallicity. c) It covers the whole rest-frame optical wavelength at intermediate spectral resolution, including the most relevant absorption diagnostics for deriving the stellar population properties.

Previous papers in this series have used the first 100 datacubes of the survey to derive spatially resolved stellar population properties by means of full spectral fitting techniques. We have obtained that: 1) Massive galaxies grow their stellar mass inside-out. The signal of downsizing is shown to be spatially preserved, with both inner and outer regions growing faster for more massive galaxies. The relative growth rate of the spheroidal component (nucleus and inner galaxy), which peaked 5–7 Gyr ago, shows a maximum at a critical stellar mass (Pérez et al. 2013). 2) The inside-out scenario is also supported by the negative radial gradients of the stellar population ages (González Delgado et al. 2014c). 3) Global and local relations between stellar mass, stellar mass surface density and stellar metallicity relation were investigated, along with their evolution (as derived from our fossil record analysis). In disks, the stellar mass surface density regulates the ages and the metallicity. In spheroids, the galaxy stellar mass dominates the physics of star formation and chemical enrichment (González Delgado et al. 2014c, a). 4) In terms of integrated versus spatially resolved properties, the stellar population properties are well represented by their values at 1 HLR (González Delgado et al. 2014c, a). The CALIFA collaboration has also compared the age and metallicity gradients in a subsample of 62 face-on spirals and it was found that there is no difference between the stellar population properties in barred and unbarred galaxies (Sánchez-Blázquez et al. 2014).

In this paper we extend our study of the spatially resolved star formation history of CALIFA galaxies to derive the radial structure of the stellar population properties as a function of Hubble type, and galaxy stellar mass, . The goals are: 1) To characterize in detail the radial structure of stellar population properties of galaxies in the local universe. 2) To find out how these properties are correlated with Hubble type, and if the Hubble sequence is a scheme to organize galaxies by mass and age, and/or mass and metallicity. 3) To establish observational constraints to galaxy formation models via the radial distributions and gradients of stellar populations for disk and bulge dominated galaxies.

This paper is organized as follows: Section 2 describes the observations and summarizes the properties of the CALIFA galaxies analyzed here. In Sec. 3 we summarize our method for extracting the SFH, based on the fossil record method, and we explain the main differences between the analysis presented here and that in previous works. Sec. 4 presents results on the galaxy stellar mass, half light and half mass radii (HLR, HMR, respectively), and galaxy averaged stellar metallicity. Sec. 5 deals with the spatially resolved properties of the stellar population: stellar mass surface density, ; luminosity weighted mean age, ; mass weighted mean metallicity, ; and stellar extinction, . We discuss the results in Sec. 6; and Sec. 7 presents the conclusions.

2 Sample, and Observations, data reduction

2.1 Sample and morphological classification

The CALIFA mother sample consists of 939 galaxies selected from SDSS survey in the redshift range –0.03, and with -band angular isophotal diameter of 45–80. These criteria guarantee that the objects fill well the FoV. The sample includes a significant number of galaxies in different bins in the color-magnitude diagram (CMD), ensuring that CALIFA spans a wide and representative range of galaxy types.

The galaxies were morphologically classified as Ellipticals (E0–7), Spirals (S0, S0a, Sab, Sb, Sbc, Sc, Scd, Sd, Sm), and Irregulars (I). The classification was carried out through visual inspection of the -band images averaging the results (after clipping outliers) from five members of the collaboration. Galaxies are also classified as for barred, otherwise , or if it is unsure, and as if it shows ”merger” or ”interaction features” (Walcher et al. 2014).

The sample for this paper comprises the 312 CALIFA galaxies observed in both V1200 and V500 setups as of January 2014. The 12 galaxies showing ”merger or interacting features” are not discussed here, leaving a main sample of 300 objects with a well defined morphology. For this work we have grouped galaxies into 7 morphology bins: E, S0 (including S0 and S0a), Sa (Sa and Sab), Sb, Sbc, Sc (Sc and Scd), and Sd (13 Sd, 1 Sm and 1 Irr).

Figure 1: Left: Comparison of the distribution of Hubble types in the CALIFA mother sample (empty bars) and the galaxies analyzed here (filled bars). The number of galaxies in our sample are labeled in colors. The histograms are normalized to form a probability density, i.e., each bar scales with the ratio of the number of galaxies in each bin and the total number of galaxies, such that the two distributions are directly comparable. Right: Color-magnitude diagram. Mother sample galaxies are plotted in grey, while the 300 galaxies analyzed in this work are marked as colored points.

Fig. 1 shows that these 300 galaxies provide a fair representation of the CALIFA survey as a whole. The left panel shows scaled histograms of the Hubble type in the mother sample (empty bars) and in our sample (filled bars). The number of objects in each morphology bin for our sample is indicated at the top, with a brown to blue color palette that represents Hubble types from ellipticals to late spirals. This same color scheme is used throughout this paper. The similarity of the distributions reflects the random sampling strategy of CALIFA, with targets being picked from the mother sample on the basis of visibility criteria alone. The right panel in Fig. 1 shows the versus CMD, with grey points representing the mother sample and colored points the 300 galaxies. As for the Hubble type distribution, a simple visual inspection shows that our subsample is representative of the full CALIFA sample in terms of CMD coverage.

2.2 Observations and data reduction

The observations were carried out with the Potsdam Multi-Aperture Spectrometer (Roth et al. 2005, PMAS,) in the PPaK mode (Verheijen et al. 2004) at the 3.5m telescope of Calar Alto observatory. PPaK contains 382 fibers of diameter each, and a Field of View (FoV Kelz et al. 2006). Each galaxy is observed with two spectral settings, V500 and V1200, with spectral resolutions Å (FWHM) and 2.3 Å, respectively. The V500 grating covers from 3745 to 7300 Å, while the V1200 covers 3650–4840 Å. Detailed descriptions of the observational strategy and of the data can be found in Sánchez et al. (2012), and Husemann et al. (2013).

The datacubes analyzed here have been calibrated with version 1.5 of the reduction pipeline. The main issues addressed by this new version are: (i) correction of the sensitivity curve for the V500 grating; (ii) new registering method to determine, for each galaxy, the relative positioning of the 3 pointings of the dithering pattern, and absolute WCS registration; (iii) a new cube interpolation method. CALIFA pipeline v1.5 improves the flux calibration to an accuracy of 2–3% and is the current official data release. A detailed account of this new pipeline is presented in the Data Realease 2 article (García-Benito et al. 2015).

In order to reduce the effects of vignetting on the data, we combine the observations in the V1200 and V500 setups. The combined datacubes were processed as described in Cid Fernandes et al. (2013). Our analysis requires that spectra have signal to noise ratio S/N in a 90 Å window centered at 5635 Å (rest-frame). When individual spaxels do not meet this S/N threshold, they are coadded into Voronoi zones (Cappellari & Copin 2003). Further pre-processing steps include spatial masking of foreground/background sources, rest-framing and spectral resampling. The resulting 253418 spectra were then processed through starlight and pycasso (the Python CALIFA starlight Synthesis Organizer), producing the stellar population properties discussed here as described in detail in the next section.

3 Stellar population analysis: Differences with respect to previous work

Our method to extract stellar population properties from datacubes has been explained and applied to CALIFA in Pérez et al. (2013), Cid Fernandes et al. (2013, 2014), and González Delgado et al. (2014c, a). In short, we analyse the data with the starlight code (Cid Fernandes et al. 2005), which fits an observed spectrum () in terms of a model () built by a non-parametric linear combination of Simple Stellar Populations (SSPs) from a base spanning different ages () and metallicities (). Dust effects are modeled as a foreground screen with a Cardelli et al. (1989) reddening law with . Windows around the main optical emission lines and the NaI D absorption doublet (because of its interstellar component) are masked in all spectral fits222To test the effect of this process in the estimation of ages, we have compared the results for 60 galaxies in common with Sánchez-Blázquez et al. (2014). This work uses Steckmap (Ocvirk et al. 2006) and the H line (previously corrected for emission). Statistically, we find that there is no difference in ages (mean = -0.04, std = 0.15 dex) if the same SSP models are used in the two methods.. Bad pixels (identified by the reduction pipeline) are also masked. Results for each spectrum are then packed and organized with the pycasso pipeline.

This working scheme is preserved here, but with three new developments:

  1. The datacubes used in this paper come from the version 1.5 of the reduction pipeline (García-Benito et al. 2015).

  2. Larger and more complete SSP bases are employed.

  3. A somewhat different definition of mean stellar metallicity is adopted (see González Delgado et al. 2014a).

This section describes the novelties related with the stellar population synthesis. Improvements resulting from the new reduction pipeline are described in Appendix A.

3.1 SSP spectral bases

SSP models are a central ingredient in our analysis, linking the results of the spectral decomposition to physical properties of the stellar populations. Our previous applications of starlight to CALIFA explored spectral bases built from three sets of SSP models, labeled as GM, CB and BC in Cid Fernandes et al. (2014). The first two are again used in this study, but extended to a wider range of metallicities, producing what we will denote as bases GMe and CBe.

Base GMe is a combination of the SSP spectra provided by Vazdekis et al. (2010) for populations older than Myr and the González Delgado et al. (2005) models for younger ages. The evolutionary tracks are those of Girardi et al. (2000), except for the youngest ages (1 and 3 Myr), which are based on the Geneva tracks (Schaller et al. 1992; Schaerer et al. 1993; Charbonnel et al. 1993). The IMF is Salpeter. In our previous studies of the first 100 CALIFA galaxies we defined base GM as a regular grid of these models, with 39 ages spanning –14 Gyr and four metallicities from 0.2 to 1.5 . We now extend the range to use of all seven metallicites provided by Vazdekis et al. (2010) models: , , , , , 0, and . Because these models lack ages below 63 Myr, these young ages are only covered by the four largest metallicities, such that our extended GM base is no longer regular in and . Base GMe contains elements.

Base CBe is built from an update of the Bruzual & Charlot (2003) models (Charlot & Bruzual 2007, private communication), replacing STELIB (Le Borgne et al. 2003) by a combination of the MILES (Sánchez-Blázquez et al. 2006; Falcón-Barroso et al. 2011) and granada (Martins et al. 2005) spectral libraries (the same ones used in base GMe). The evolutionary tracks are those collectively known as Padova 1994 (Alongi et al. 1993; Bressan et al. 1993; Fagotto et al. 1994a, b; Girardi et al. 1996). The IMF is that of Chabrier (2003). Whereas in previous works we limited the range to solar, we now extend this base to six metalicities: , , , , 0, and . Base CBe contains elements (41 ages from 0.001 to 14 Gyr and the 6 metallicities above).

The main similarities and differences between bases GMe and CBe are the same as between the original GM and CB bases, thoroughly discussed in Cid Fernandes et al. (2014). Throughout the main body of this paper we focus on results obtained with base GMe, but we anticipate that our overall qualitative findings remain valid for base CBe. The role of base CBe in this paper is to allow a rough assessment of the uncertainties associated to model choice.

A minor technical difference with respect to our previous analysis is that we now smooth the spectral bases to 6 Å FWHM effective resolution prior to the fits. This is because the kinematical filter implemented in starlight operates in velocity-space, whereas both CALIFA and the SSP model spectra have a constant spectral resolution in -space, so that effects of the instrumental broadening can only be mimicked approximately by starlight. We have verified that this modification does not affect the stellar population properties used in this paper.

Appendix B presents some comparisons of the results obtained with these two bases. Experiments were also performed with bases which extend the age range to 18 Gyr, and configuring starlight to allow negative values of . These tests are also discussed in Appendix B, which adds to the collection of “sanity checks” on the results of our analysis.

4 Galaxy mass, metric, and stellar metallicity

This section addresses three relatively unrelated aspects, which are all important to better understand the results presented in the next section, where we examine how the spatial distribution of stellar population properties relates to a galaxy’s stellar mass and morphology. First, §4.1 reviews the relation between stellar mass and morphological type for our sample. This strong relation is imprinted on virtually all results discussed in §5. Secondly, §4.2 compares our measurements of the Half Light (HLR) and Half Mass Radii (HMR). As discussed by González Delgado et al. (2014c), these two natural metrics for distances are not identical due to the inside-out growth of galaxies. Here we inspect how the HMR/HLR ratio varies as a function of Hubble type and stellar mass in our sample. Finally, §4.3 presents our definition of mean stellar metallicity. González Delgado et al. (2014c) showed that stellar mass surface densities, mean ages, and extinction values defined from the integrated spectrum, from galaxy-wide spatial averages, and measured at HLR all agree very well with each other. Here we extend this test to stellar metallicities. Throughout this section, results for the two SSP models discussed in §3.1 are presented.

4.1 Stellar masses

(M) bin E S0 Sa Sb Sbc Sc Sd
9.1 - - - - - 2 2
9.1-9.6 - - - - - 9 8
9.6-10.1 - - - - 2 10 5
10.1-10.6 - - 7 11 16 21 -
10.6-10.9 3 8 9 14 21 4 -
10.9-11.2 8 14 22 17 16 3 -
11.2-11.5 17 8 13 10 3 1 -
11.5-11.8 12 2 - 1 - -
11.8 1 - - - - -
total 40 32 51 53 58 50 15
Table 1: Number of galaxies for each Hubble type and M interval ()

To obtain the total stellar mass of a galaxy we add the mass in each zone, thus taking into account spatial variations of the stellar extinction and ratio. Masked spaxels (e.g., foreground stars) are accounted for using the radial profile as explained in González Delgado et al. (2014c).

Fig. 2 shows the distribution of as a function of Hubble type. Table 1 shows the distribution of galaxies by Hubble type in several bins of . The masses range from to for fits with GMe (Salpeter IMF). CBe-based masses (Chabrier IMF) are on average smaller by a factor 1.84. As for the general galaxy population, mass is well correlated with Hubble type, decreasing from early to late types. High bulge-to-disk ratios (E, S0, Sa) are the most massive ones (), while galaxies with small bulges (Sc–Sd) have . The average is 11.4, 11.1, 11.0, 10.9, 10.7, 10.1, and 9.5 for E, S0, Sa, Sb, Sbc, Sc, and Sd, respectively. The dispersion is typicaly 0.3 dex, except for Sc galaxies, that have a dispersion of dex.

Because CALIFA is not complete for , this distribution in mass is not completely representative of the local Universe. In particular, it is important to remember that dwarf ellipticals are not included, so or any other property discussed here for E’s are restricted to massive ellipticals.

Figure 2: Distribution of the stellar masses obtained from the spatially resolved spectral fits of each galaxy for each Hubble type (grey small points). The colored dots (stars) are the mean galaxy stellar mass in each Hubble type obtained with the GMe (CBe) SSP models. The bars show the dispersion in mass.

4.2 The HMR/HLR

Figure 3: Left: The ratio between half mass and half light radius (a/ a) with the Hubble type (left). Big colored dots represent the averaged a/ a in each Hubble type bin, and the lines the dispersion. Stars and big circles show the results obtained with the and bases, respectively. Right: a/ a as a function of the galaxy stellar mass. The black circles show the averaged correlation independently of the morphological type. Large circles represent the averaged relation in mass intervals of 0.25 dex for each color-coded morphological type.

As explained in Cid Fernandes et al. (2013), we define the HLR as the semi-major axis length of the elliptical aperture that contains half of the total light of the galaxy at the rest-frame wavelength 5635 Å. Similarly, the HMR is derived from the 2D distribution of the stellar mass, as the elliptical aperture at which the mass curve of growth reaches 50% of its asymptote. The ratio between the HMR and the HLR (/) reflects the spatial variation of the star formation history in a galaxy. This ratio is lower than 1 in almost all cases (González Delgado et al. 2014c), a signpost of the inside-out growth found by Pérez et al. (2013).

Fig. 3 shows the relation between / and Hubble type (left panel), and galaxy stellar mass (right panel). These plots confirm our earlier finding that galaxies are more compact in mass than in light. If the gradient in stellar extinction is taken into account, the average / for base GMe (CBe). Fig. 3 shows that the ratio decreases from late to early type spirals; while lenticulars and ellipticals have similar /.

These results are also in agreement with our previous result that / shows a dual dependence with galaxy stellar mass: It decreases with increasing mass for disk galaxies but it is almost constant in spheroidal galaxies, as confirmed in the right panel of Fig. 3. Sb-Sbc galaxies are the ones with the lowest /.

4.3 Stellar metallicity

Metallicity is one of the most difficult stellar population properties to estimate. Reasons for this include: (i) the coarse metallicity grid of the SSP bases; (ii) the limitation of the stellar libraries to the solar neighborhood; and (iii) inherent degeneracies like the dependence of the continuum shape on extinction, age, and metallicity, whose effects are hard to disentangle. Notwithstanding these difficulties, meaningful estimates of can be extracted from observed spectra, particularly by means of full spectral synthesis methods (Sánchez-Blázquez et al. (2011)).

starlight-based estimates of for the same CALIFA sample used in this paper were previously used by González Delgado et al. (2014a) to study global and local relations of with the stellar mass and stellar mass surface density. We have shown there that: (i) our sample follows a well defined stellar mass-metallicity relation (MZR), (ii) this relation is steeper than the one obtained from O/H measurements in HII regions, but that considering only young stellar populations the two MZR’s are similar, and (iii) is strongly related to in galaxy disks and to in spheroids. All these results lend confidence to our estimates.

Here we review our definition of the mean stellar metallicity, and test whether its value at 1 HLR matches the galaxy wide average value as well as the one obtained from the spatially collapsed data cube.

4.3.1 Mean stellar metallicity

The main properties analyzed in this paper are the stellar mass surface density (), stellar extinction (), mean age (), and metallicity of the stellar population, whose spatial distributions are studied as a function of Hubble type and total stellar mass (). These properties were defined in previous articles in this series. For instance, the mean light weighted log stellar age is defined as


(eq. 9 of Cid Fernandes et al. 2013), where is the fraction of flux at the normalization wavelength (5635 Å) attributed to the base element with age and metallicity . The mass weighted version of this index, , is obtained replacing by its corresponding mass fraction .

While Cid Fernandes et al. (2013) average the base metallicities linearly (their eq. 10), in this paper, as in González Delgado et al. (2014a), we employ a logarithmic average:


for the mass weighted mean and


for the luminosity weighted mean . The motivation to use this definition is that the extended SSP bases used in this study span a much wider dynamical range in (nearly three orders of magnitude, compared to barely one in our previous papers), which is better handled with a geometric mean (implicit in the use of the logarithm). This is the same reasoning behind the use of instead of .

To some degree, the definition of mean is largely a matter of taste (albeit one with mathematical consequences because of the inequality of the arithmetic and geometric means, ), so much so that one finds both types of averaging in the literature. For instance, in Gallazzi et al. (2005) metallicities are averaged logarithmically, whereas Asari et al. (2007) work with arithmetic averages.

As shown in González Delgado et al. (2014a) (see also Fig. 4), our metallicities span about 1 dex for galaxy masses ranging from to , with an MZR which matches well the stellar metallicities of both Milky Way and LMC-like galaxies.

4.3.2 Galaxy averaged stellar metallicity

González Delgado et al. (2014c) obtained the important result that galaxy-averaged stellar ages, mass surface density, and extinction are well matched by the corresponding values of these properties at HLR and also with the values obtained from the analysis of the integrated spectrum (i.e, the one obtained by collapsing the datacube to a single spectrum). The general pattern therefore is that galaxy averaged properties match both the values at 1 HLR and those obtained from integrated spectra. Do our stellar metallicities comply with this rule?

To answer this question we first define the galaxy-wide average stellar metallicity following eq. 2 in González Delgado et al. (2014a). which gives the mass weighted mean value of as


where is the stellar mass in spaxel .

Fig. 4 compares our results for with the mass weighted mean values obtained at HLR (, bottom panels) and those derived from the integrated spectrum (, top panels), analyzed in the exact same way as the individual zone spectra. Results are shown for both base GMe (left panels) and CBe (right).

The agreement is remarkable. The galaxy averaged metallicity and the one at 1 HLR are the same to within a dispersion of 0.1 dex. The integrated metallicity also matches the galaxy averaged value, with only slightly larger dispersions. The largest deviations occur for low metallicity systems. Similar conclusions are reached if the comparison in Fig. 4 is done using the light weighted version of eq. (4),


where is the luminosity (corrected by stellar extinction) in each spaxel evaluated at a reference wavelength (5635 Å in our case).

The stellar metallicities behave as expected, in the sense that, like other properties, their galaxy-wide averages match the values at HLR, and also the values derived from integrated spatially unresolved spectroscopy (González Delgado et al. 2014c). We thus conclude that galaxy-wide spatially averaged stellar population properties (stellar mass, mass surface density, age, metallicity, and extinction) match those obtained from the integrated spectrum, and that these spatially averaged properties match those at R= 1 HLR, proving that effective radii are really effective (González Delgado et al. 2014b).

Figure 4: Upper panels: Comparison of the galaxy-wide average stellar metallicity (weighted in mass) derived from the spatially resolved spectral analysis () and the integrated metallicity derived from fitting the total (integrated) galaxy spectrum (). Lower panel: Comparison of with the value measured HLR (). Left and right panels show results obtained with base GMe and CBe SSPs, respectively. All panels include 300 galaxies. The difference between the y-axis and x-axis is labeled in each panel as , and the dispersion as .

5 Spatially resolved stellar population properties as a function of morphology and mass

This section presents a series of results derived from our spatially resolved spectral synthesis analysis of CALIFA galaxies. We focus on the following four stellar populations properties: mass surface density (, §5.1), mean ages (, §5.2), metallicities (, §5.3), and extinction (, §5.4). Each of these properties is studied by means of (i) 2D maps of the individual galaxies, (ii) radial profiles, and (iii) radial gradients. Throughout the section, the emphasis is on evaluating and comparing the roles of morphology and total stellar mass in shaping the observed behavior of these four properties.

Before discussing the results, we briefly explain how these quantities are obtained and how they are presented.

2D maps in the CMD: Using pycasso we obtain, for each galaxy, 2D maps of each of the four properties. The results for all the galaxies are presented in the framework of the color-magnitude diagram, where each map is placed at the galaxy’s coordinates in the vs.  CMD. Because absolute magnitude is related to and redder galaxies are (usually) older and more metal rich, these plots show the correlations -, –age, and –metallicity in a 2D fashion. Because in our sample the galaxy Hubble type is correlated with color and luminosity, these plots not only show how the galaxy averaged properties and their radial structure change with the galaxy stellar mass, but also with the morphological type. These maps are shown in the Appendix C (Figs. 2427).

Radial profiles: Each 2D map is azimuthally averaged in order to study the radial variations of each of the four stellar population properties. Elliptical apertures 0.1 HLR in width are used to extract the radial profiles, with ellipticity and position angle obtained from the moments of the 5635 Å flux image. Expressing radial distances in units of HLR allows the profiles of individual galaxies to be compared on a common metric, as well as averaging (“stacking”) them as a function of Hubble type or stellar mass. Radial profiles expressed in units of the HMR were also analyzed and lead to similar shapes, so all profiles presented below use HLR as the unit for radius.

Radial gradients: Inner and outer gradients are defined as differences between the values at and 0 (), and and 1 (), respectively. For instance,


for , and similarly for ,  and . Defined in this way, the gradients have units of dex/HLR (mag/HLR for ). Since the stellar population properties of galaxies at 1 HLR represent very well the galaxy-wide average, () effectively measures how the bulge (disk) properties change with respect to those of the galaxy as a whole.333Based on the exponential fit analysis developed by (Sánchez et al. 2013) and (Sánchez-Blázquez et al. 2014), we conclude that the regions between 1 and 2 HLR are dominated by the disk component; thus, measures the disk gradient. However, is not measuring the bulge gradient. The reason is that the effective radius (R) of the spheroidal component can be smaller than 1 HLR, and it shows a dependence with the morphological type. Thus, R HLR for E, but is significantly smaller in late type spirals.

Unless otherwise noted, all results reported below are for the GMe base, although the whole analysis was carried out with properties derived with both sets of SSP models discussed in 3.1. Differences between GMe and CBe SSPs go in the following way: (a) The stellar mass surface density is lower with CBe than with GMe by 0.27 dex on average, mostly due to the different IMFs (Salpeter in GMe versus Chabrier in CBe). (b) Variations in stellar extinction are negligible. (c) CBe yields somewhat younger ages and higher metallicities than GMe, by an average of 0.14 dex in  and 0.12 dex in . These shifts reflect the age-metallicity degeneracy, and are mainly a consequence of the different sets of metallicities available in these bases. However, radial gradients are not affected by this degeneracy. A detailed comparison of properties derived with the two bases is given in Appendix B.

5.1 Stellar mass surface density

2D maps of the stellar mass surface density for the 300 individual galaxies of our sample are presented in the Appendix C (Fig. 24). Here we discuss the radial structure of as a function of Hubble type and .

5.1.1 –morphology and –mass relations

Fig. 5 shows how measured at 1 HLR changes with Hubble type (left panel), and with the galaxy stellar mass (right). Recall from González Delgado et al. (2014c) that properties measured at 1 HLR match very well the corresponding galaxy-wide average value, so these plots ultimately show how the global depends on the morphology and on .

The plot shows increasing from late spirals to spheroids, with average and dispersion values of , , , , , , , for E, S0, Sa, Sb, Sbc, Sc and Sd, respectively. Note that E and S0 are remarkably similar.

Surface densities also increase with , as seen in the right panel of Fig. 5. The overall - relation is relatively smooth, with no evidence of an abrupt change of behavior as that discussed by Kauffmann et al. (2003) for SDSS galaxies. Fig. 5, however, reveals that morphology is also behind the dispersion in the - relation. The black line shows the relation for the full sample, obtained by averaging in 0.4 dex-wide bins in mass, while the big circles break this general relation into different (color-coded) morphological types for the same mass bins. Despite the reduced statistics, it is evident that: (a) for the same stellar mass, early type galaxies are denser than late type ones, and (b) Sa and earlier type galaxies exhibit a much flatter - relation than later types. The overall impression from these results is that morphology, and not only stellar mass, plays a fundamental role in defining stellar surface densities, and it is responsible for the change of slope in the SDSS - relation.

Figure 5: Left panel: stellar mass surface density measured at 1 HLR as a function of Hubble type. Small dots represent for each galaxy; the colored circles are the average for each Hubble type, and the error bars are the dispersion in for each morphological type. Right panel: - relation. Individual galaxies are represented by small dots colored by their morphological type. The black line is the average in galaxy stellar mass bins of 0.4 dex. Large colored circles are the average in each bin of mass for each Hubble type.

5.1.2 Radial Profiles

Figure 6: (left) Radial profiles (in units of HLR) of the stellar mass surface density obtained with base . The results are stacked in seven morphology bins. The error bar in the panel indicate the dispersion at one HLR distance in the galaxies of the Sa bin. It is similar for other Hubble types and radial distances. (right) Radial profiles stacked in seven bins of galaxy stellar mass, : 9.19.6, 9.610.1, 10.110.6, 10.610.9, 10.911.2, 11.211.5, 11.511.8.

Azimuthally averaged radial profiles of are shown in Fig. 6. Results are stacked by Hubble type (left panel) and mass (right). In the left panel galaxies are grouped in our seven morphological classes. The typical dispersion within these bins is illustrated by the error bar, which shows the standard deviation in for galaxies of the Sa class.

A clear trend with Hubble type is seen: The profiles scale with Hubble type from late to early spirals, and this modulation with morphology is preserved at any given distance. E and S0 have remarkably similar profiles, with core and extended envelope equally dense at any given distance, suggesting that the disk of S0 galaxies and the extended envelope of ellipticals have grown their mass by similar processes.

The right panel of Fig. 6 shows the radial profiles grouped in seven bins of stellar mass spanning the 9.1–11.8 range. These also show that the average of is modulated by . However, this - modulation breaks for early type galaxies (concentration index C () 2.8; see also Fig.13 in González Delgado et al. (2014c)), that in our sample are populated mainly by E and S0, and some Sa. On the other hand these early types are all massive, with 10 M.

5.1.3 Radial gradients

Figure 7: (left) Correlation between the inner (grey-red) and outer (grey-blue) gradient of and the morphological type. The results are shown for the (stars) and (circles) SSP models. The inner gradient is calculated between the galaxy nucleus and 1 HLR, and the outer gradient between 1 HLR and 2 HLR. (right) Correlation between the inner gradient of and the galaxy stellar mass. Small dots represent the results for each galaxy, and black crosses the average for each 0.3 dex mass bin. Large circles represent the averaged inner gradient in mass intervals of 0.3 dex for each color-coded morphological type. Black crosses show the average correlation between the inner gradient of and galaxy mass independently of the morphological type.

Inner (0–1 HLR) and outer (1–2 HLR) gradients in , as defined by equations (6) and (7), are plotted as a function of morphology and stellar mass in Fig. 7. values (corresponding to the core region) are plotted in grey-red, while (which trace the disks of spirals and S0 and the extended envelope of ellipticals) are plotted in grey-blue. Circles and stars show results for bases GMe and CBe respectively, illustrating that even though these bases yield different absolute values of the resulting gradients are nearly identical.

A clear correlation exists between and Hubble type. The gradient in the inner HLR increases (in absolute values) significantly from late to early spirals, converging to a constant value for E and S0. This relation reflects the variation of the bulge to disk ratio in spirals, and the dominance of the bulge component in spheroids (S0 and E). The outer gradient is weaker (smaller in absolute value) than the inner one, as expected if a disk component dominates the mass outwards of 1 HLR.

The right panel of Fig. 7 shows the relation between the inner gradient and the stellar mass. There is a clear increase (in absolute values) of with , with the more massive galaxies having a steeper increase of the central density. The dispersion with respect to the average values (black cross) within -bins is significant. To check the effect of morphology on this dispersion we have averaged in mass intervals for each Hubble type and plotted the resulting averages (large colored circles). The general trend that emerges is that, for galaxies of the same mass, early type galaxies tend to be overall centrally denser than later types, in agreement with Fig. 5; although, there are a few intervals of stellar mass (e.g. = 11.4 M), in which the variations in with Hubble type are not significant.

It is also worth mentioning that in Sa and Sb is very close to that in S0 and E, and in this sense it would be easy to fade early type spirals into S0’s.

5.2 Ages of the stellar populations

2D maps of the luminosity weighted mean log stellar ages (eq. 1) for the 300 galaxies are presented in Fig. 25. Here we discuss the radial structure of  and its relation to Hubble type and . The presentation follows the same script used in the presentation of -related results in §5.1.

5.2.1 Age–morphology and age–mass relations

Fig. 8 shows how the mean age of the stellar populations at 1 HLR changes along the Hubble sequence (left panel), and with the galaxy stellar mass (right). Similarly to , represents well the galaxy-wide averaged stellar population age (, González Delgado et al. (2014c)).

Clearly, scales with Hubble type, increasing steadily from Sd to Sa. S0 and ellipticals have stellar populations of similar mean age, and older than spirals. The average and dispersion values of (yr) are , , , , , , and , for Sd, Sc, Sbc, Sb, Sa, S0 and E, respectively.

Mean ages also increase with the galaxy mass (right panel of Fig. 8), a “downsizing” behavior that has been confirmed with widely different samples and methods. For instance, our age-mass relation is similar to that derived for SDSS galaxies by Gallazzi et al. (2005) (their figure 8). They found that there is a transition at 444Equivalent to for our IMF., below which galaxies are typically young and above which they are old. This is the same mass at which Kauffmann et al. (2003) find the - relation to flatten.

Unlike in these SDSS-based works, we do not see sharp transitions as a function of in neither nor , although differences in sample selection and statistics prevent a proper comparison. We do, however, note a common behavior in the right panels of Figs. 5 and 8, in the sense that the dispersion above is strongly related to morphology.

Like in Fig. 5 (right panel), the black line in the right panel of Fig. 8 shows the age-mass relation for the whole sample, obtained by averaging values in bins. Small dots show individual galaxies, while the large colored circles represent the mass-binned average for each Hubble type. As with the - relation, breaking the age-mass relation into morphological types reveals clean trends. In this case, we see that, for a fixed , earlier type galaxies are older than later types. The corollary is that mass is not the sole property controlling the SFH of a galaxy. In fact, given the flat age-mass relations for Sa, S0 and E, morphology seems to a more relevant factor, at least in these cases.

Figure 8:  measured at 1 HLR as a function of Hubble type (left) or galaxy stellar mass (right). Symbols and colors are as in Fig. 5. The black line is the average  in galaxy stellar mass bins of 0.4 dex.

5.2.2 Radial profiles

Figure 9: Radial profiles of  as a function of Hubble type (left) and in seven bins of galaxy stellar mass (right). These bins are = 9.19.6, 9.610.1, 10.110.6, 10.610.9, 10.911.2, 11.211.5, 11.511.8. Symbols and colors are as in Fig. 6. These results are obtained with base .

Fig. 9 shows the age radial profiles obtained by stacking galaxies as a function of Hubble type and mass. The profiles scale with Hubble type, but by different amounts at the center than at 1 HLR. At any radial distance, however, the early type galaxies are older than later type ones. E and S0 are again very similar at all radii. This suggests that E and S0 have similar histories not only on average, but also spatially resolved, at least in the inner 2 HLR. Negative age gradients are detected in all galaxies (except perhaps in Sd, whose ages profiles are flatter than in the other spirals555The small drop of  toward the center of Sd galaxies is caused by a couple of galaxies with young nuclear regions. Given that this group is the least populated in our analysis (only 15 galaxies), better statistics is needed to evaluate the reality of this feature.). These negative gradients reflect the inside-out growth of galaxies. Furthermore, the decrease of  with indicates that quenching happens earlier at the galaxy center; and also earlier in early type galaxies (spheroids and Sa) than in later type spirals (Sbc–Sc).

The radial profiles also show a clear trend with (Fig. 9, right), with the more massive galaxies being older everywhere, hence preserving the downsizing pattern at all radial distances. Comparing the left and right panels in Fig.  9, one sees that grouping galaxies by their stellar mass leads to a reduced vertical stretch in their profiles than when the averaging is done by morphology. But the profiles expand similar vertical scale if galaxies earlier than Sd and more massive than 10 M are considered; indicating that the effect of morphology and stellar mass are not easily disentangled here. However, in §6.3, Fig. 20 shows that the dispersion in the profiles between galaxies of the same and different Hubble type is significant, and larger than between the profiles of galaxies of different but the same Hubble type. These results in agreement with Fig. 8 indicate that the age profiles are more related to morphology than to M. Since is essentially a first moment of the spatially resolved SFH, we can conclude that the SFH and its radial variation are modulated primarily by the galaxy morphology, with mass playing a secondary role.

5.2.3 Radial gradients

Figure 10: (left) As Fig. 7 but for . The inner gradient shows a clear dependence with Hubble type, that seems to be stronger than with the galaxy mass. Sb-Sbc-Sc galaxies have larger inner gradients with than with , but both sets of models show a similar dependence with Hubble type. (right) The inner gradient of  as a function of galaxy mass. Colors and symbols are as in Fig. 7.

Gradients in , computed as indicated in eqs. (6) and (7), are plotted in Fig. 10 against Hubble type (left panel) and stellar mass (right). The figure layout is exactly as in Fig. 7. Whilst in that plot results obtained with bases GMe and CBe (circles and stars in the left panel, respectively) could hardly be distinguished, here the results for these two sets of SSPs do not overlap so precisely, although the differences in are clearly very small (see §B.2).

A clear relation exists between and morphology: The inner age gradient increases from early type galaxies to Sb-Sbc spirals, which are the galaxies with the largest variation between the age of the stellar population at the bulge and the disk. Spirals of later type (Sc and Sd) have flatter radial profiles than Sb-Sbc. The outer (between 1 and 2 HLR) age gradient shows a similar bimodal behavior as , but with a smaller amplitude.

The right panel of Fig. 10 shows the behavior of with . The gradient tends to increase (become more negative) from low mass galaxies (which have roughly flat profiles) up to about , at which point the trend reverses and decreases with increasing . This is best seen following the black crosses, that trace the mass-binned mean relation. The dispersion with respect to this relation is significant and is related to the morphology, as seen through the large colored circles. The tendency is that, at a given mass, S0 and early type spirals have weaker than Sb-Sbc. This dependence of age gradients with the Hubble type at a fixed indicates again that the spatial variation of the SFH is mainly driven by the morphology and not by the stellar mass.

However, the morphology (understood as the B/D ratio (Graham & Worley 2008)) can not be the only driver of the spatial variation of the SFH along all the Hubble sequence. Fig. 10 shows that there is not a monotonic relation between the B/D ratio and , with galaxies with the smaller B/D ratio having the largest variations in  between the central core and the disk. This bimodal behavior seen in Fig. 10 suggests that other physical properties are also important in establishing the spatial variation of the SFH, which on the other hand is reflecting the different bulge formation processes along the Hubble sequence.

5.3 Stellar metallicity

Fig. 26 presents the images of the mass weighted mean (logarithmic) stellar metallicity (cf. eq. 2). Here we discuss the radial structure of  as a function of Hubble type and .

5.3.1 Metallicity-morphology and mass-metallicity relations

Fig. 11 shows how the stellar metallicity measured at 1 HLR changes with the Hubble type (left panel) and with the galaxy stellar mass (right).

Stellar metallicities grow systematically from late to early type galaxies. The statistics within each Hubble type are , , , , , , and for Sd, Sc, Sbc, Sb, Sa, S0, and E, respectively.

Not surprisingly, metallicities also grow with , as shown in the right panel of Fig. 11. Since we have shown in §4.3.2 that the galaxy-wide average stellar metallicity is well represented by the metallicity at 1 HLR, this plot is in fact equivalent to the global mass-stellar metallicity relation (MZR). We have previously found that this relation is steeper than the one derived from HII regions, which is similar to the flatter stellar MZR obtained when we consider only young stars (González Delgado et al. 2014a). As in Fig. 5, the smoothed black curve is obtained by averaging in 0.4 dex bins of . The dispersion in the MZR is significant, and larger than the dispersion produced by the galaxy morphology as shown by the distribution of large colored circles. These circles are the average in each mass bin for each Hubble type, and show the tendency of earlier type galaxies to be more metal rich than late type galaxies of the same stellar mass.

Figure 11:  measured at 1HLR as function of Hubble type (left) and galaxy stellar mass (right). Symbols and colors are as in Fig. 5. The black line is the average  obtained in 0.4 de bins of .

5.3.2 Radial profiles

Figure 12: Radial profiles of  as a function of Hubble type (left) and of galaxy stellar mass (right). Mass bins are = 9.19.6, 9.610.1, 10.110.6, 10.610.9, 10.911.2, 11.211.5, 11.511.8. Symbols and colors are as Fig. 6. These results are obtained with base .

Fig. 12 shows the results of stacking the radial profiles of  as a function of Hubble type and . Outwards decreasing  is detected for most morphological classes, but flat profiles are found for Sc-Sd galaxies. Intermediate type spirals (Sb-Sbc) stand out as the ones with the largest variations in stellar metallicity.

The behavior of the radial variation of the stellar metallicity with (right panel in Fig. 12) is similar to the behavior with morphology. Most galaxies have  that decreases with , except for the two lowest mass bins, which show flat profiles. The largest spatial variations are also found in galaxies in the intermediate mass bins ().

These negative radial gradients of the metallicity are also an indicator of the inside-out formation processes in galaxies. The inversion of the gradient in late type spirals and in low mass spirals may be an indicator of the secular processes or the outside-in formation scenario in these galaxies (Pérez et al. 2013).

5.3.3 Radial gradients

Figure 13: Left: As Fig. 7 but for . Right: The inner gradient as a function of the galaxy stellar mass. Symbols and colors are as in Fig. 7.

Fig. 13 clones Figs. 7 and 10 for  gradients. On the left panel, one sees that, as for stellar densities and ages, results for bases GMe and CBe are very similar. On average, galaxies have  gradients dex/HLR, similar to the value obtained from nebular oxygen abundances (Sánchez et al. 2013). Outer and inner gradients are not significantly different. Despite the large scatter, there is a hint of a bimodal distribution as that found for stellar ages, also with intermediate type spirals in a pivotal position and late type spirals with the flattest gradients, at least in a statistical sense.

The right panel of Fig. 13 shows as a function of . The dispersion is significant, but on average there is a tendency to turn flat profiles into negative gradient ones as increases from to . The largest gradients are found between and . More massive galaxies tend to have weaker stellar metallicity gradients. The dispersion is significant throughout this relation. A trend with morphology is seen in the sense that, for a given mass, early types are the ones with weaker gradients.

5.4 Stellar extinction

starlight models the stellar extinction as a foreground screen, parametrized by and following the Galactic reddening law. Images showing the spatial distribution of for our 300 galaxies are presented in Fig. 27. Here we present related results as a function of Hubble type and , following the same script adopted in the discussion of , , and  in the previous subsections, thus completing the list of stellar population properties studied in this work. Unlike masses, ages, and metallicities, extinction is more easily affected by inclination effects, so the results reported below should be interpreted with caution. Section 5.5 explores this issue in depth.

5.4.1 Extinction–morphology and extinction–mass relations

Figure 14: A measured at 1HLR as function of Hubble type (left) and galaxy stellar mass (right). Symbols and colors are as in Fig. 5. The black line is the average  obtained in 0.4 dex bins of .
Figure 15: Radial profiles of A as a function of Hubble type (left), and in seven bins of galaxy stellar mass (right). These bins are = 9.19.6, 9.610.1, 10.110.6, 10.610.9, 10.911.2, 11.211.5, 11.511.8. Symbols and colors are as Fig. 6. These results are obtained with base .

Fig. 14 shows how the stellar extinction at 1 HLR changes with Hubble type (left panel), and with stellar mass (right panel). As with other properties, represents well the mean extinction of the galaxy666The galaxy average extinction for each galaxy is calculated as the mean of all the 20 radial values obtained for each galaxy between the center and 2 HLR. as well as the value derived from spectral fits of the integrated spectra.777The difference between and is , while between A and it is . The left panel in Fig. 14 shows as a function of morphology. Ellipticals and S0s have almost no extinction, with mean , and mag, respectively. Sa, Sb and Sc galaxies have around 0.25 mag, and somewhat smaller ( mag) in Sd’s.

There is no clear behavior of stellar extinction with galaxy stellar mass. In general, galaxies with have –0.3 mag. More massive galaxies are less extinguished, and for fixed mass early types tend to have smaller , but the dispersion is large.

5.4.2 Radial profiles

Fig. 15 shows profiles stacked by Hubble type (left panel), and mass (right). Spirals have mag in the disk and up to 0.6 mag in the center. Their profiles are similar for all Hubble types except for Sd’s, where does not change much from disk to center. Ellipticals and S0’s also show negative gradients, although at distances larger than 1 HLR they are almost dust-free. The radial profiles in different bins of (right panel) show a similar behavior to that with morphology. Except for the most massive bins, shifted to lower extinction values, all other mass-binned profiles are similar.

Figure 16: Left: As Fig. 7 but for A. Right: The inner gradient as a function of galaxy stellar mass. Symbols and colors are as in Fig. 7.

5.4.3 Radial gradients

gradients are shown in Fig. 16, which is formatted as Figs. 7, 10 and 13. As for the previous properties, results for bases GMe and CBe are nearly indistinguishable, as illustrated by the overlap of circles and stars in the left panel. and show similar behavior with morphology, although the inner gradient is always higher than the outer one. In Ellipticals the gradient of exists only in the central region. With the exception of Sd galaxies, spirals have mag/HLR.

On average, gets stronger with increasing up to (Fig. 16, right) and weakens towards higher mass, spheroid dominated systems. The dispersion with respect to the mass-binned relation (traced by the black crosses) is large, and not clearly related to morphology (coded by the colored circles).

As a whole, and despite the general trends summarized above, of the four properties analyzed in this section, is the one for which tendencies with Hubble type and stellar mass are less clear. A major reason for this is that, unlike for , , and , estimates are sensitive to inclination effects. This is explained next.

5.5 Effect of inclination on the radial profiles

Figure 17: Radial profiles of , , , and (from the upper to the bottom panels) for the different Hubble types (from E to Sd from left to right panel) and three different bins of the ratio of minor to major photometric radius of the galaxy: solid line (, face on), dashed line (0.39 b/a 0.63), and dotted line (, edge on).

An implicit hypothesis throughout the analysis presented so far is that galaxy inclination does not affect our estimates of the stellar population properties and their radial distributions. One expects this assumption to break down in the case of , which should increase from face on to edge on galaxies, although it is not unreasonable to conceive that inclination effects propagate to the spectral synthesis-based estimates of stellar mass surface densities, mean ages, and metallicities. It is therefore relevant to evaluate if and how inclination affects our results.

In order to do so, we have divided the 300 galaxies in three subsamples on the basis of the ratio (minor to major isophotal axes), as measured in SDSS -band images. The three subsamples, each containing 100 galaxies, cover (i) , edge on, (ii) , and (iii) , face on. Galaxies in each sub-sample were grouped by Hubble type, and their radial profiles of , , , and averaged as previously done for the whole sample in the left panels of Figs. 6, 9, 12, and 15.

Fig. 17 shows the resulting stacked profiles of , , , and . Solid, dashed and dotted lines show profiles for the “face-on” (), intermediate inclination (), and “edge-on” () samples respectively, and each column is for one of the seven Hubble types used throughout the paper. Average profiles were only computed for morphology-inclination bins containing at least 4 galaxies.

Stellar mass surface density, age, and metallicity profiles show a negligible variation among the -based subsamples. This result indicates that inclination does not affect the estimates of these properties in any systematic way. Any difference is at a level not significant insofar as this paper is concerned. The exception is the  profiles for “edge-on” Sc’s, which differ substantially from the profiles of less inclined Sc’s. It so happens, however, that the sub-group of Sc’s has a mean stellar mass 0.4 dex lower than other Sc’s, which could explain their lower metallicities without implying inclination effects.

The one property which does vary systematically with is , and it does so in the expected sense: Spirals with lower have larger extinction. This is particularly evident in Sb’s. This dependence hinders the interpretation of the stacking results presented in §5.4, and explains why no clean tendencies of the values and profiles with morphology and stellar mass were identified.

6 Discussion

This section is divided into four main parts. First we summarize our results in the context of related studies. We then discuss our findings in the context of the growth of galaxies – theoretical expectations, high redshift observations, and previous results of inside-out growth for CALIFA galaxies. In the third part we explore what the results tell us about the quenching of star formation in galaxies. Finally, we discuss the theoretical predictions for the radial variations of age and metallicity in early types and in spirals from different simulations of galaxy formation. We compare our results for variations of the radial structure in the inner (R HLR) and outer (R HLR) parts with other observational results in the literature.

6.1 Age and metallicity of bulges and disks

The analysis of SDSS data has generated a general knowledge of how the age and metallicity of galaxies change with , color, or concentration index (e.g Kauffmann et al. 2003; Gallazzi et al. 2005; Mateus et al. 2006). These studies have confirmed that, in general, early type galaxies are old and metal rich, while late type galaxies are younger and more metal poor. Numerous (single spectra or longslit) studies have reported also that ellipticals are metal rich, and have a range of ages, 2–10 Gyr, that depend on stellar velocity dispersion (e.g. Trager et al. 2000; Thomas et al. 2005; Sánchez-Blázquez et al. 2006; Graves et al. 2009; Johansson et al. 2012).

Our spatially resolved analysis introduces a significant improvement in the study of the structure of galaxies. For example, we compute ages and metallicities of bulges in disk galaxies and compare them with elliptical galaxies in a systematic way, avoiding problems derived from the lack of spatial resolution that some of the previous studies have.

We compute the luminosity-weighted and the mass-weighted age and metallicity: (i) in the central part of galaxies (values averaged within 0.25 HLR) as representative of the stellar population properties of bulges and central core of ellipticals; and (ii) in the outer part of galaxies, values averaged in a ring at 1.50.1 HLR, as representative of disks. Fig. 18 plots the individual results as small dots; large dots represent the average values for each (color-coded) Hubble type, and the error bars show the dispersion. While  gives information about the globally ‘averaged’ star formation history,  informs when most of the stellar mass was formed.

Fig. 18 shows that the bulges of Sa-Sb and the cores of E-S0 formed at a similar epoch; they are very old (10 Gyr) and metal rich (1 Z). Thus, they probably formed by common processes, that occurred rapidly and early on. However, the bulges in Sc-Sd galaxies (shown as the two darkest shade of blue) are younger and have a more extended star formation history (both  and  are smaller), and have lower stellar metallicities. Thus, Sc-Sd galaxies formed in later times and/or by different processes.

Many bulges in late type spirals are in fact pseudo-bulges. Unlike true bulges, pseudo-bulges are thought to grow by secular processes, where material from the disk (by the loss of angular momentum) is the main source of star formation in the central parts of these galaxies (e.g. Kormendy & Kennicutt 2004). We may see this effect at work in Fig. 9 and Fig. 12, as a flattening of the radial profiles of  and , and the positive  gradient in the core of Sc galaxies. Some effects of the secular processes due to the disk may also be present in the bulges of Sa-Sb. For example, Fig. 18 shows that bulges of Sa-Sb have 6 Gyr, younger than the 10 Gyr epoch of formation derived from ; and this may be understood if some disk stars are rearranged into the bulges or if dissipation processes bring gas triggering new star formation in the center.

Fig. 18 also shows that disks are younger and more metal poor than bulges. Both  and  are lower in disks than in their respective bulges, indicating that disks formed later than bulges, and that star formation continues for a longer time in disks that in bulges, probably as a consequence of a continuing availability of gas in disks (Roberts & Haynes 1994). This indicates a general scenario of inside-out formation.

Figure 18: Left: Mass-weighted age at the galaxy center () and at 1.5 HLR () for the different Hubble types. Small dots are the individual radial points, while big colored dots represent mean values for each Hubble type, with the error bars indicating the dispersion in the mean. Middle: As in the left panel but for the light-weighted age (). Right: As in the left panel for . The top horizontal axes in the left and middle panels show the redshift scale. The diagonal line in the three panels is the bisector.

6.2 Inside-out growth of spheroids and spirals

6.2.1 Theoretical expectations and recent results from high redshift galaxies

Models of galaxy formation predict a common inside-out view for the mass assembly in galaxies (e.g. Kauffmann et al. 1993; Aumer & White 2013). First, the bulge formed at high redshift; then, the disk was built around the bulge in the case of spirals. In the case of ellipticals, the central core formed at z2, and the envelope grew later through minor mergers (Oser et al. 2010; Hilz et al. 2013, e.g). Observational evidences come from the significant size evolution in early type galaxies (ETG), that grow as size (van Dokkum et al. 2010; Patel et al. 2013).

More recently, van Dokkum et al. (2014) find evidence against the inside-out formation scenario for spirals. For a sample of MW-like spirals at redshift z=2.5, they estimate the dependence of the radius with , and find that their size– relation is similar to the size– of similar galaxies at z=0. They conclude that the mass growth took place in a fairly uniform way, with the galaxies increasing their mass at all radii, thus, their R barely grows. These results seem to be supported by numerical simulations by Elmegreen et al. (2008), that find that bulges can be formed by migration of unstable disks. Other observational evidence come from the detection of clumpy star forming disks in galaxies at z2 (Genzel et al. 2008; Förster Schreiber et al. 2011), that may indicate an early build up of bulges by secular evolution. Thus, studies at high redshift are providing new results that draw a complex landscape of galaxy build up. For example, Wuyts et al. (2011) also find clumpy disk star formation, but at the same time conclude that there is a Hubble sequence in place at least since z2.5. On the other hand, there is other evidence that galaxies rapidly assemble inside-out at z=1 (Nelson et al. 2012; Szomoru et al. 2010, 2012); while Hammer et al. (2005) find evidence that MW-like galaxies have rebuilt their disk at z in a major merger epoch that drastically reshapes their bulges and disks, and is consistent with earlier cumplier evolution.

In summary, there is mounting evidence of the major processes responsible for the assembly and shaping of galaxies at different epochs, and these are complemented with a variety of processes that modify the inside-out formation scenario: stellar migration, bar induced gas inflows, gas-rich minor merger, angular momentum loss due to reorientation of the disk, infall of gas with misaligned angular momentum, etc (Aumer et al. 2014).

6.2.2 CALIFA view of the inside-out growth of galaxies

The results from our studies favor an inside-out growth of spirals. Pérez et al. (2013) studied the stellar mass growth as a function of the radius and cosmic time in galaxies with , and showed that the nuclei grow faster than the inner 0.5 HLR, that, in turn, grow faster than the outer 1 HLR. This conclusion is supported by the stellar age radial profiles presented in González Delgado et al. (2014c), and confirmed here in Fig. 10 for most spirals and spheroidals. Further support comes from the ratio HMR/HLR (Fig. 3), a good probe of the spatial variation of the star formation history in galaxies (González Delgado et al. 2014c). This ratio is lower than 1 (Fig. 3), a fingerprint of the inside-out growth found by Pérez et al. (2013).

Fig. 19 shows how the radial profiles of  decrease outwards for all the Hubble types. Most of the stellar mass in the center has been formed 10 Gyr ago or earlier (z2). But at 1.5 HLR,  ranges from 7 Gyr (z1) in E–S0 to 4.5 Gyr (z0.4) in Sbc, suggesting that, both early type and MW-like, galaxies have continued accreting or forming in-situ stars in their disks until more recent times, thus supporting the inside-out scenario in these galaxies.

This trend, however, changes beyond 1.5-2 HLR, where  and  flatten. This may be interpreted as indicating that the mass was formed in a more uniformly distributed manner across the outer disk, or that stellar migration shuffles inner born stars to the outer disk, washing out the inside-out formation signs so clearly seen in the inner 1.5 HLR. In the case of E–S0 this may be understood if beyond 2 HLR most of the stellar mass in the galaxies was already accreted at .

Figure 19: Radial profiles (in units of HLR) of the mass weighted age, , obtained with GMe base. The results are stacked by morphological type as in Fig.7.

6.3 Quenching

Figure 20: Radial profiles of  (upper panel) in four galaxy stellar mass bins. From left to right: = 11.211.5 (continuum line), 10.911.2 (dashed line), 10.610.9 (dashed-point line), 10.110.6 (dotted line). In each panel, the average profile for each Hubble type is plotted if more than four galaxies have galaxy stellar mass in the bin. Bottom: each panel shows the radial profile of each Hubble type averaged in each of the four bins.

Several mechanisms have been proposed to explain the shutdown of star formation in galaxies. Halo mass quenching is one of the most popular ones that explains the bimodal distribution of the properties of galaxies, and it is required to explain the green valley as a pathway towards quenching of star formation in early and late type galaxies (e.g. Schawinski et al. 2014). In this scheme, galaxies with a halo mass below a critical value (a few 10 M) accrete cold gas conducive to star formation. Above this critical mass, the infalling gas reaches the sound speed and cannot form new stars (e.g. Cattaneo et al. 2006; Dekel & Birnboim 2006). The dependence with environment and clustering strongly supports this quenching mechanism (e.g. Weinmann et al. 2006; Peng et al. 2010).

The differential dependence of the stellar mass surface density (Fig. 5 and Fig. 6) with the galaxy stellar mass (a proxy of the halo mass) provides further evidence of the halo quenching (e.g. Behroozi et al. 2010). Estimating of the properties of the stellar populations in SDSS galaxies, Kauffmann et al. (2003) found that there is a critical mass ( M, equivalent to M for our Salpeter IMF) below which scales with , and above which is independent of the galaxy stellar mass. Right panels of Fig. 5 and Fig. 6 support this scenario because the radial profiles of scale with , and furthermore they do so all along their extent. Our results also show that saturates at high ; because the high mass end of the distribution is dominated by early type galaxies (Sa-S0-E), this suggests that the spheroidal component plays a significant role in the quenching of star formation in high mass galaxies.

The importance of morphology in the quenching of galaxies has also been reported in the literature (e.g. Bell 2008; Bell et al. 2012; Barro et al. 2013; Pan et al. 2014; Woo et al. 2015). Martig et al. (2009) found that the dependence of quenching with morphology is a consequence of the bulge-building mechanism. The steep potential well induced by the formation of a large spheroid component results in the stabilization of the disk, that cuts the supply of the gas, preventing its fragmentation into bound, star forming clumps. Our results support this scenario, as it is explained below, because the dependence of the SFH of galaxies with the morphology.

If the halo mass is the main property responsible for quenching, we should expect that the radial structure of  (both, the age values and the gradients) to change more with than with Hubble type. On the contrary, if quenching is driven by morphology, galaxies of similar stellar mass would have very different  structure depending on Hubble type. We explore the relevance of morphology versus in Fig. 20: age radial profiles are shown as a function of and of morphology, in four mass bins (=11.511.2, 11.210.9, 10.910.6, 10.610.1). Clearly, morphology is the main driver: it can account for up to 0.75 dex change in age at a given mass (top panels); conversely, at a fixed morphology, mass accounts for less than 0.25 dex (bottom panels). Further, morphology accounts not only for changes in absolute values, but also for changes in the gradients at a given galaxy mass.

This confirms the similar result obtained above with , and it implies that galaxies of similar (equivalent to have similar ) and with a large spheroid have shutdown their star formation (outside their central core) earlier than galaxies of later morphology. These results indicate that the SFH and their radial variations are modulated primarily by galaxy morphology, and only secondarily by the galaxy mass, suggesting that the bulge formation has a relevant role in quenching the star formation in galaxies.

6.4 Radial structure of the stellar population properties in ETG and their relation with galaxy formation models

6.4.1 Theoretical predictions from cosmological simulations

Classical chemical evolution models of the formation of early type galaxies (ETG) are based in two possible scenarios: 1) dissipative formation, the well known monolithic collapse; and 2) the non-dissipative collapse. These scenarios produce very different radial gradients of ages and abundances, being very steep in the first case, with values of [Fe/H] 0.5 to 1.0 [dex/dex] (Larson 1974, 1975; Carlberg 1984)888The metallicity gradient measured in spheroids is traditionally calculated as [Fe/H]/ and expressed in [dex/dex] units., but (almost) flat when there are pure stellar mergers. This second case may even erase a previously existing radial gradient.

The most recent cosmological simulations propose a two phase formation scenario for ETG’s in which the central core formed at z 2, and the envelope grows after this through minor mergers (e.g. Naab et al. 2009; Oser et al. 2012; Hilz et al. 2012; Navarro-González et al. 2013). Thus: 1) Galaxies assemble their mass through dissipative processes and star formation occurs in-situ. Starbursts formed at the center as a consequence, for example, of large major mergers or monolithic collapse. The star formation is induced by cold flow of accretion or by gas-rich mergers. 2) Galaxies grow in size by mass assembly through the external accretion of satellites; ex-situ star formation formed by dry mergers of galaxies towards the central most massive one.

Observationally, there is evidence of a significant size evolution in ETGs. The growth of the galaxy size with M supports this proposal. A transition region is expected between the in-situ central core of ETG and ex-situ outer regions. Since the central core of these ETG is enriched very quickly due to major mergers at high redshift ( 2), and the satellites that are accreted are less metal rich than the central massive one, a negative radial gradient of the metallicity is expected, even with a change of the slope in the transition region. Thus, values as [Fe/H] = [dex/dex] (Pipino et al. 2010) or [Fe/H] = [dex/dex] (Kawata & Gibson 2003) are predicted.

However, the merger history may change an existing radial gradient: while dry major mergers can flatten the pre-existing gradient (Kobayashi 2004; Di Matteo et al. 2009; Hopkins et al. 2009), dry minor mergers can steepen the metallicity gradient. Thus, Kobayashi (2004) SPH chemodynamical simulations of elliptical galaxies that include radiative cooling, star formation and feedback from SNII-Ia, and chemical enrichment, (but do not include kinematic feedback), found that the steep negative radial metallicity gradient, established during the initial starburst at , can flatten significantly by later major-mergers in the galaxy. Following these simulations, the average gradient at the present time is [Fe/H]= [dex/dex], but it may decrease to a value of [dex/dex] when major mergers appear.

Beside the merger history, feedback can change the inner and outer metallicity gradients. Thus, a strong AGN feedback can stop the star formation in the central core of massive galaxies, flattening the inner gradients. Feedback from in-situ star formation can alter the outer metallicity gradient. Also, the existence of galactic winds may modify the composition of the ISM in a galaxy. Hirschmann et al. (2015) performed cosmological simulations that include an empirical model for the momentum driven galactic winds, to investigate the dependence of the age and metallicity outer gradients with metal cooling and galactic winds, (in principle required to explain the mass-metallicity relation, MZR). These simulations including winds predict [Fe/H] = [dex/dex], steeper than the simulations without winds that predict [Fe/H] = [dex/dex]. The main explanation is that in wind models the stars accreted are of lower metallicity than in the simulations with no winds. In both cases, however, they predict a positive age gradient of 0.030.04 [dex/dex].

6.4.2 Implications from this work and comparison with other results from the literature

Following our own results in this work, E and S0 have formed their central core at similar cosmic time since they have similar central ages (see Fig. 18). Further they must have formed through similar processes since their radial profiles of , , and  are remarkably similar. They both show small but negative , and  gradients in the central core. In the central 1 HLR, E and S0 in our sample have [dex/dex] (std = 0.15)999Our gradients, that are measured in a linear scale, are converted here to a logarithmic scale to be compared with predictions from simulations and other works in the literature.. (Sligthly steeper, = [dex/dex], when CBe models are used.) These are within the range of values found in other studies based on long-slit or IFS data up to one effective radius (e.g. Mehlert et al. 2003; Sánchez-Blázquez et al. 2007; Annibali et al. 2007; Rawle et al. 2008; Spolaor et al. 2010; Kuntschner et al. 2010). However they are shallow compared with theoretical expectations if minor mergers are relevant in growing up the central core of E and S0 galaxies. This may indicate that major mergers are more likely the main process building the central regions (up 1 HLR) of ETGs.

Between 1 and 3 HLR, the radial profile of  is of similar slope or slightly shallower than in the inner 1 HLR. We do not find any evidence of a transition region where the metallicity radial profile steepens to metallicities below solar. If the 1 to 3 HLR envelope of ETG had grown through the accretion of low mass satellites a steepening of metallicity would be expected, as explained before, because the mass-metallicity relation implies that low mass satellites would be of low metallicity. In our results there is no evidence either of an inversion of the age radial profile toward older ages beyond HLR, as expected if these satellites were formed very early on like the core of E and S0 (see Fig. 9 and Fig. 18). These results are in contrast with recent ones by Greene et al. (2012, 2013): for a sample of 30 early type galaxies they find at 2 R an old (10 Gyr) stellar population with [Fe/H], and interpret this as the stellar outskirts of these galaxies being built up through minor mergers. Also Coccato et al. (2010); La Barbera et al. (2012); Montes et al. (2014) observing a few massive ellipticals have reported a decline of the metallicity to under solar in an old stellar population in their outskirts ( 10 R) suggesting that these galaxies are formed in two phases, the central core through major mergers, and through minor mergers farther out. However, other recent works show examples of ETGs with an old and metal rich stellar population and a very shallow metallicity gradient up to 3 R (Trujillo et al. 2014), in contrast with the results by Greene et al. (2012, 2013).

Our results do not support the minor merger scenario for the size growth of ETGs. Thus, the ages, 9.7 (yr), and metallicity, Z, at HLR, and the shallow metallicity gradient, [dex/dex], that we obtain are more consistent with the growth of the HLR envelope of ETGs through major mergers.

Other interesting result reported in the literature that can be compared with ours is the correlation between the metallicity gradient and the galaxy mass (or stellar velocity dispersion, ) found for E and S0. Spolaor et al. (2010) have found that the relation between and the metallicity gradient shows two regimes of behavior: (i) for galaxies with 2.2 km s, the metallicity gradient steepens with ; (ii) galaxies with 2.2 km s (the most massive ellipticals), have a metallicity gradient that does not correlate with (or galaxy mass), with a mean value [dex/dex]. On the other hand, Pastorello et al. (2014) derive the gradient in the outer R of a sample of ellipticals, and they find that the gradient covers a wide range of values, from negative very steep () to flat or even positive; these values correlate with the galaxy stellar mass and stellar velocity dispersion, with the galaxies of lower (or ) having the steeper metallicity gradient. However, the most massive galaxies exhibit the flattest gradients and an average value of [dex/dex] (std = 0.38) for galaxies with . Both works show a significant scatter in the relation for , and reasonable doubts of the existence of the correlation for high mass ellipticals.

Our results (Fig. 13) indicate that there is no correlation between the metallicity gradient and for the CALIFA early type galaxies (E and S0). Even so, our results are compatible with Spolaor et al. (2010) and Pastorello et al. (2014), because E and S0 in our sample are all above 10 (and 100 km s), for which no correlation is found between the metallicity gradient and in Spolaor et al. (2010) or Pastorello et al. (2014). We find that CALIFA ellipticals have a shallow gradient. This behavior is also in agreement with Hirschmann et al. (2013) simulations and their interpretation of the lack of correlation between the metallicity gradient and in massive ellipticals. Following these authors, massive galaxies accrete higher mass satellites, and because of their deeper potential well they retain their own gas against stellar winds, producing a shallower metallicity gradient in the outer regions of massive ellipticals.

Because the metallicity at 23 HLR in E and S0 are similar to the metallicity in the bulge of early spirals, and the stars at these distances are as old as the bulges of Sa-Sb galaxies (see Fig. 9), the 13 HLR envelope of early type galaxies might have built from the centers of early type spirals. In summary, the negative but shallow gradients of the metallicity and ages suggest that massive (M ) early type galaxies built their inner 3 HLR through mergers with massive and metal rich spirals.

6.5 Radial structure of the stellar population properties in spirals and their relation with galaxy formation models

New insights on the structure of the Milky Way disk, in particular through the measurements of chemical abundances of large sample of stars, are provided by the spectroscopic surveys undertaken in recent times (e.g., SEGUE, RAVE, Gaia-ESO survey, HERMES, APOGEE, LAMOST, etc; Yanny et al. 2009; Steinmetz et al. 2006; Gilmore et al. 2012; Zucker et al. 2012; Majewski et al. 2010; Zhao et al. 2012). RAVE (Radial Velocity Experiment) (Steinmetz et al. 2006; Boeche et al. 2014) is studying the radial and vertical chemical gradients using a very large sample of dwarf stars. Close to the Galactic plane, RAVE shows a negative radial gradient of Fe abundance, dex/kpc 101010The metallicity gradient in disks is traditionally calculated as [Fe/H]/r, and expressed in dex/kpc., that becomes flatter or even positive when measured above the disk. So, the [Fe/H] gradient ranges from to dex/kpc when measured at heights kpc, or kpc above the Galactic plane, respectively.

The radial gradient of abundances in the different regions of a spiral galaxy are important because they are directly related with the formation process. Obviously, not all scenarios of disk/spiral formation are valid, since it is necessary that they produce a radial gradient of abundances in the disk but not in the halo, as observed in the MW and in M31. Thus, the formation of the halo from different fragments or minor mergers with very short free fall times does not create a radial gradient but a dispersion of abundances, and therefore it was early concluded that the MW halo may be formed from mergers or from the accretion of low mass galaxies (or part of them). However, disks are more likely formed from a single cloud falling on and from inside-out.

6.5.1 Theoretical predictions from ”classical” chemical evolution models

Most classical chemical evolution models claim that infall of gas with a radial dependence, implying an inside-out scenario for the disk formation, is essential to reproduce the observed radial gradient of abundances. The key ingredient is the dependence of the disk infall time scale with the radial distance, that makes the gas to accumulate faster in the inner disk. Since the SFR depends on the gas density, these assumptions produce a radial dependence of the star formation rate and a negative radial metallicity gradient (Ferrini et al. 1994; Molla et al. 1996; Chiappini et al. 2001; Mollá & Díaz 2005). Thus Molla et al. (1996) give a value for a MW like galaxy, reproducing the value found by Shaver et al. (1983) and other Hii regions studies. Chiappini et al. (2001) models predict a gradient of dex/kpc for a MW-like galaxy. In fact, as theoretical equations show (Goetz & Koeppen 1992), the radial gradient of abundances appears in the disk when there is an adequate ratio between star formation rate to infall rate. It also implies, therefore, that a dependence of the radial gradient on the morphological type of galaxies may exist. Molla et al. (1996) models already predicted radial gradients for galaxies of different morphological types, with values in the range (for a M31-like galaxy) to (for a late type galaxy like NGC 300). More recent works (Mollá & Díaz 2005) calculate models where the infall rate was a function of the mass distribution (or rotation curve) of the galaxy, assuming a stronger radial dependence of the infall timescales than in Chiappini et al. (2001). Moreover Mollá & Díaz (2005) models also depend on an efficiency factor to condense the molecular gas, and to convert the gas reservoir into stars. The metallicity gradients range to with flat gradients for galaxies with the largest efficiency factor, or the most massive ones, although in the extreme end the low mass and lowest efficiencies models also show flat radial distributions. Thus, the steepest gradients appear in the intermediate mass or intermediate type galaxies. However, there is no dependence on the morphological type when the gradient is normalized to a characteristic value, such as the effective radius as recent results by CALIFA have found based on HII regions abundances (Sánchez et al. 2013).

6.5.2 Theoretical predictions from cosmological simulations

Recently, hydrodynamical cosmological simulations have provided evidences in support of the imposed inside-out disk growth scenario adopted within the ”classical” chemical evolution models. Like spheroidals, spirals are formed in two phases. In the first phase the bulge formed in a similar way as the core of E-S0. In the second phase, the disk grows by star formation in-situ from the infalling gas (Kauffmann et al. 1993; Aumer & White 2013). Metal poor gas with higher angular momentum at lower redshifts is turned into stars at larger radii. Negative radial metallicity gradients are expected, as the classical models predict. This assumption is a natural outcome of the mass, momentum, and energy conservation laws, imposed in the simulations of disks in a cosmological context (Brook et al. 2011, 2012; Few et al. 2012; Pilkington et al. 2012a, b; Gibson et al. 2013).

Pilkington et al. (2012a) have examined a set of 25 simulations, from several groups, using different codes and initial conditions (Stinson et al. 2010; Rahimi et al. 2011; Kobayashi & Nakasato 2011; Few et al. 2012) to predict the present-day metallicity gradient in MW-like galaxies and its evolution. Although the evolution of the simulated metallicity gradients depends strongly on the choice of the sub-grid physics employed, most of the simulated galaxies tend to a similar present-day gradient of , in agreement with the Chiappini et al. (2001) and Mollá & Díaz (2005) models for normal galaxies as the MW.

6.5.3 Implications from this work

Our findings show that spiral galaxies (excluding Sd) have negative radial gradients as indicative of the inside-out growth of the disk (see Fig. 12). The average  for spirals (excluding Sd and later type) is dex/HLR or dex/kpc. These values are compatible with the results obtained by Sánchez-Blázquez et al. (2014), that have already derived the metallicity gradients for 62 CALIFA face-on spiral galaxies to study the effect of bars on the properties of the stellar populations. For these galaxies, they find a metallicity gradient of dex/kpc (std = 0.05), equal ( dex/kpc) to the gradient that we derive for the same group of galaxies.

In order to compare with our results, the RaDES simulated galaxies (Few et al. 2012), 19 galaxies of the Pilkington et al. (2012a) sample, have been analyzed in a similar way as we have done here, i.e. measuring the gradients in a similar way and using the HLR values of the simulated galaxies (Ruiz-Lara in prep., and Ruiz-Lara et al. private communication). The simulated galaxies analyzed in this work cover a narrow range of morphologies, mainly Sbc-Sc. Therefore, these results cannot be extrapolated to the full work presented here, but they are representative of the state-of-art of cosmological simulations of disk galaxies, and can be used to compare them to similar disks from our observations. Mock B-band images are used to derive the HLR and to perform a bulge-disk decomposition used as a proxy for the morphology. Metallicities are calculated for disk particles using Eq.(2) and the gradient is derived between HLR and also between 1 and 2 HLR. The stellar metallicity and age gradients of the simulated galaxies are compatible with the results presented here. Keeping in mind that the morphological range covered by these simulations is rather narrow and that they use B/D as a proxy for a morphological classification, the results show a slight dependence of  with B/D ratio, with a steeper slope for B/D=1 (Sbc galaxies) for which   dex/HLR. Later type spirals have a flatter gradient of   dex/HLR. These results go in line with those found here, namely, that the metallicity radial gradient of spirals shows a dependence on morphology, with the steepest gradient found in the intermediate Sb-Sbc spirals (see Fig. 12 and Fig. 13). However, a larger set of cosmological simulations is required covering from early Sa to late Sd, and a large range of galaxy masses (from 10 to 10 M), in order to confirm the general trend found here. On the other hand, these results indicate that the feedback recipes used in these simulations are able to recover realistic galaxies with small bulges and are fully in agreement with the work presented here.

Furthermore, our results are also compatible with classical chemical models, and certainly, the CALIFA Sb-Sbc galaxies have stellar metallicity gradients ( dex/kpc) in the range observed in the MW disk, but somewhat shallower than the [Fe/H] gradient measured in the Galactic disk. However, it is necessary to take into account that the gradient usually given in the literature is obtained for young stars or HII regions, while here it is an average value obtained for all stellar populations existing in the studied galaxy or region. Besides that, the number of objects is increased compared with the old studies and, more important, all of them have been self-consistently analyzed using the same reduction technique and spectral models.

In any case our results favor an inside-out growth of spirals. This conclusion is supported by the stellar age radial profiles presented here: the age decreases outwards for all Hubble types studied111111Sd galaxies, however, show a much flatter age gradient. Beyond HLR the radial distribution of ages flattens, suggesting that the mass forms more uniformly in those regions, or that the stellar mixing brings stars born in the inner disk to the outskirts. This last possibility has been recently investigated by Minchev et al. (2013, 2014), who have performed N-body hydrodynamical models with the chemical evolution implementation (Minchev et al. 2013, 2014). They have simulated MW-like galaxies with the aim to investigate whether the Galactic disk can be understood as a single structure with kinematic and chemical features that are continuously distributed, being the thin and thick disks two extreme cases of these structures. Furthermore, they investigate the effect of stellar migration and kinematic heating in the scatter of the age-metallicity relation, and how it changes with the Galactic radius. In fact, an increase of the scatter in the age-metallicity relation and a flattening of the stellar metallicity gradient is produced by the stellar radial migration, that causes a radial mixing in the older stellar population, creating the appearance of a flatter gradient in early times, and leading to a decoupling of the stelar population from their birth interstellar medium (Roškar et al. 2008). These results also indicate that even though radial mixing has a significant effect in flattening the metallicity gradient, it can not destroy it.

7 Summary and conclusions

We have analyzed the stellar population properties of 300 galaxies, observed by CALIFA with the V500 and V1200 gratings and IFU PPak at the 3.5m telescope of Calar Alto, to investigate the trends in the stellar populations properties with radial distance as a function of Hubble type and galaxy stellar mass. The sample includes ellipticals, S0 and spirals from early (Sa-Sb) to late types (Sc-Sd). They cover a stellar mass range from 0.710 to 710 if Salpeter IMF is assumed, and a factor 1.78 (0.25 dex) lower for a Chabrier IMF. A full spectral fitting analysis was performed using the starlight code and a combination of SSP spectra from González Delgado et al. (2005), Vazdekis et al. (2010), or Charlot & Bruzual (2007, private communication). Our pipeline pycasso is used to process the spectral fitting results to produce present day maps of the spatial distribution of the stellar population properties. For each galaxy, these maps are azimuthally averaged to produce radial profiles (in units of the half light radius, HLR: ) of the stellar mass surface density (), stellar ages (light weighted, , and mass weighted, ), metallicity (), and extinction (). The radial profiles are stacked as a function of Hubble type and of galaxy mass. Radial gradients of these properties measured within the inner 1 HLR and between 1 and 2 HLR are also obtained.

Our main results are:

  1. Spatially averaged vs. integrated galaxy properties: the metallicity, , galaxy-wide spatially averaged matches the metallicity obtained from the integrated spectrum, and the metallicity at R=1 HLR. This result is equivalent to that obtained for the other stellar population properties, , , and , as reported by González Delgado et al. (2014c, b), proving that effective radii are indeed effective.

  2. Mass weighted size: We confirm our earlier finding (González Delgado et al. 2014c) that galaxies are more compact in mass than in light by 20. The HMR/HLR ratio shows a dual distribution with Hubble type, that breaks in the Sb-Sbc, the galaxies with the smaller HMR/HLR. This ratio also shows a dual dependence with : it decreases with increasing mass for disk galaxies, and becomes almost constant in spheroidal galaxies. These results are a signpost of the inside-out growth previously found by Pérez et al. (2013).

  3. Stellar mass surface density: shows declining profiles that scale with morphology and with ; this behavior is preserved at any given distance. At constant , is higher in early type than in late type spirals. E’s and S0’s show equal profiles, independently of . The inner gradient, , correlates with Hubble type. The negative gradients steepen from late type spirals to spheroids, as well as with galaxy total mass in galaxies with 10 . At a constant , steepens with morphology, with E’s and S0’s having the steepest gradients. These results indicate that morphology, and not only , plays a relevant role in defining , and the relation.

  4. Stellar ages: (R) shows declining profiles that scale with morphology; this behavior is preserved at any given distance. Early type spirals are always older than late spirals. E’s and S0’s, although older than spirals, have both similar (R) profiles, indicating that these galaxies have similar star formation histories. The more massive galaxies are also the older ones; this “downsizing” behavior is always preserved at any given distance. The negative  depends on Hubble type in different ways: steeper from E and S0 to Sbc, and shallower from Sbc to Sd. Thus, Milky Way like galaxies have the steepest age gradient. A relation exists, increasing the gradient from the low mass galaxies (which have roughly flat profiles) up to about 10, at this point the trend reverses and  decreases with increasing . However, the dispersion in the relation and is significant and it is strongly related with the morphology. Even more, the dispersion of the (R) profiles of galaxies of equal mass is significant and larger than between the (R) profiles of galaxies of different but the same Hubble type. Thus, the SFHs and their radial variations are modulated primarily by the Hubble type, with mass playing a secondary role.

  5. Stellar metallicity: (R) shows mildly decreasing profiles for most Hubble types, except Sd’s that show little, if any, radial dependence. Milky Way like galaxies (Sbc) stand out as the ones with the steepest radial profiles. (R) scales with in a similar way as it does with morphology. This can be understood as a consequence of the global mass metallicity relation –a primary dependence of the metallicity with . The metallicity gradients are negative but shallow on average, with   dex/HLR, and show a small dependence with