Online material

Appendix A: Improvements resulting from the new reduction pipeline

The quality of the starlight fits to CALIFA version 1.3c spectra was assessed in Cid Fernandes et al. by averaging R_λ = O_λ − M_λ residual spectra of 107 galaxies (~ 10⁵ spectra). Inspection of these residuals revealed low amplitude (a few %) but systematic features related to unmasked weak emission lines, SSP deficiencies, and data calibration imperfections. This exercise needs updating now that the reduction pipeline has changed to version 1.5.

Figure A.1 summarizes the results of this re-evaluation. The plots show stacked R_λ = O_λ − M_λ residual spectra, in units of the median flux in the 5635 ± 45 Å window. The top panel shows results for the nuclear extractions, while the middle and bottom panels are built using spectra from zones within radial distances R = 0−1 and 1−2 half light radius (HLR, computed in the same wavelength range), respectively. Residuals are colored according with the Hubble type of the galaxies. When all galaxies are stacked, the residuals are colored according with the spatial zone extracted. These subdivisions are presented to get a sense of how the residuals relate to position within a galaxy and its Hubble type, which are two central aspects of this paper.

No matter which panel one looks at, the improvement with respect to version 1.3c is evident to the eye when compared

to figure 13 of Cid Fernandes et al. . The broad trough around Hβ present in the 1.3c spectra, for instance, is much shallower now. In fact, it is confined to late types (compare blue and red lines in the lower panel in Fig. A.1), indicating that its origin is related to calibration, and also to the SSP spectra of young stellar populations (as previously reported by Cid Fernandes et al. for SDSS data). Residuals are also visibly smaller towards the blue, including the CaII K line, which is now well fitted whereas in version 1.3c a small systematic residual subsisted¹². The humps around 5800 Å, on the other hand, are still present in version 1.5, particularly noticeable for outer regions, indicating that further refinement of the sky subtraction are warranted.

In short, the spectral fits have improved substantially with the new reduction pipeline. We attribute this to the updated sensitivity curve used in version 1.5. A more extended discussion of these and other aspects of the data reduction are presented in García-Benito et al. .

Despite these changes, the stellar population properties derived from the spectral fits did not change much in comparison to those obtained for version 1.3c data. The most noticeable changes were in mean ages, which become 0.1 dex older, and extinction, which is now 0.2 mag smaller on a global average.

	Fig. A.1 Upper panel: residual spectra averaged for all the spectra (black line), nuclei (grey line) and spectra belonging to zones that are between 0−1 HLR (pink line) and between 1−2 HLR (yellow line). Middle panel: residual spectra averaged for zones inner to 1 HLR and for Hubble type. Bottom panel: as in the middle panel except for spectra of zones located between 1 and 2 HLR.
Open with DEXTER

Appendix B: Base experiments and uncertainties due to SSP templates

To derive the stellar population properties of these 300 CALIFA galaxies we have fitted ~253000 spectra with the GMe and CBe using the cluster Grid-CSIC at the Instituto de Astrofísica de Andalucía and the cluster Alphacrucis at IAG-USP Sao Paulo. Examples of the quality of the spectral fits as a function of the Hubble type and radial distance are presented in Fig. A.1.

Appendix B.1: Mass-metallicity relation

Here we want to find out how well our metallicity definitions follow a MZR which guarantees that galaxies like the MW or Andromeda (log M_⋆(M_⊙) ~ 11) have solar metallicity at the disk, while LMC-SMC-like galaxies (log M_⋆(M_⊙) ~ 9) have ~ 1 / 4 Z_⊙. We do this with mass-weighted and luminosity-weighted definitions of Eqs. (2) and (3), and with the two sets of SSP models (GMe or CBe).

Similarly, the correlation between the galaxy averaged stellar metallicities and the metallicity measured at 1 HLR, and the MZR, guarantee that the metallicity radial profiles scale with the galaxy stellar mass. However, the MZR in González Delgado et al. was derived using the galaxy averaged stellar metallicity instead of the metallicity measured at 1 HLR, and using the mass-weighted definition of the metallicity and only the results with the base GMe. For these reasons, we derive the MZR that results from using the mass-weighted and the light-weighted definition of the metallicity, and the GMe and CBe bases.

Figure B.1 shows the correlation of M_⋆ and (upper panels), and (lower panels) for the GMe (left panels) and CBe (right panels) SSP models. The mass-metallicity relation found by Panter et al. and Gallazzi et al. are the magenta and brown lines, respectively¹³. Note that in the four cases, the metallicities are well in the range of the dispersion given by Gallazzi et al. ; brown dashed line represent the 16th and 84th percentiles of their distribution. To compare the general trend of these values, we derived a smoothed mass-binned relation, represented by a solid black or grey-black line. As expected from the global MZR derived in González Delgado et al. , base GMe and ⟨ log Z_⋆⟩_M predict stellar metallicities for MW and LMC-SMC with the expected values. But ⟨ log Z_⋆⟩_L gives a MZR that predict higher metallicities. The opposite happens for the MZR using the base CBe, which gives mass-weighted metallicities higher on average than the SDSS metallicities. But the MZR with ⟨ log Z_⋆⟩_L goes close to the Gallazzi et al. relation, and also predicts stellar metallicities for MW-Andromeda-like and LMC-SMC galaxies with the expected values.

In summary, ⟨ log Z_⋆⟩_M with GMe and ⟨ log Z_⋆⟩_L with CBe provide a mass-metallicity relation similar to the SDSS MZR, and predict metallicities between −0.7 and −0.4 dex for galaxies with mass between ~10⁹ and 10¹⁰M_⊙, the expected values for LMC and SMC-like galaxies, and solar for MW-like galaxies.

Fig. B.1 The global stellar MZR for 300 CALIFA galaxies is shown as dots, color coded by the morphological type. The metallicity is derived using (upper panels) and (lower panels), and the total stellar mass, M⋆, obtained with the GMe SSP models (left panels) and CBe (right panels). A mass-binned smooth mean relation is shown as a solid black or grey-black line. The MZRs obtained for SDSS galaxies by Gallazzi et al. and Panter et al. are plotted as brown and magenta lines, respectively, with dashed brown indicating the 16 and 84 percentiles of Gallazzi et al. .

Open with DEXTER

Appendix B.2: Variations over bases GMe and CBe

	Fig. B.2 Correlation between SP obtained with the GMe (x-axis) and CBe (y-axis bottom and middle panels; and GMe (x-axis) and GMd (y-axis). The average difference between the property in the y-axis and x-axis is labeled in each panel as Δ, and the dispersion as (σ).
Open with DEXTER

Although the two sets of models are built with the same stellar libraries (MILES and granada), base GMe stops at 1.5 Z_⊙, while CBe goes up to 2.5 Z_⊙. Because MILES is built with stars in the solar neighborhood, it does not contain stars as metal rich as 2.5 Z_⊙, so CBe results in over solar metallicity should be interpreted with care. On the other hand, the central parts of galaxies can be as metal rich as 2−3 Z_⊙, and the base GMe may be too low to fit spectra of these regions, leading to saturation effects. To avoid these problems, we also fitted the spectra with another two sets of SSP, are identical to GMe and CBe, but where for each metallicity bin, two SSPs of age 16 and 18 Gyr are added to our “standard” bases. These extra bases, which we name GMd and CBd, allow galaxies older than the age of the universe if their bulges are very metal rich. Furthermore, results at very low metallicity also must be taken with care. MILES contain only few stars of metallicity below 1/100 Z_⊙. For this reason, Vazdekis et al. provide a safe age range for each metallicity bin, being the models with log Z_⋆(Z_⊙) ≤ −1.7 only valid between 10 and 18 Gyr. This safety margin is provided to avoid the cases when, because of age-metallicity degeneracy, these old metal poor models fit young metal-rich populations, may happen if the base does not include SSP younger than 100 Myr. Our fits do not suffer this problem because bases GMe and CBe both have spectra of ages as young of 1 Myr.

Appendix B.3: Global results and uncertainties associated with SSP models

To evaluate to which extent the spectral synthesis results depend on the choice of SSP models, we now compare the global properties derived with bases GMe and CBe. Using our pipeline pycasso we obtained the radial distribution of the stellar population properties for each galaxy with a spatial sampling of 0.1 HLR. Here we compare the stellar population properties of the 0.1 HLR radially sampled points, instead of comparing the results obtained from the individual 253 418 fitted spectra. Figure B.2 shows the results for a total of 6000 points corresponding to a maximum of 20 radial points (from nucleus to 2 HLR) for each of the 300 galaxies analyzed in this work. The figure compares the results for base GMe in the x-axis, with CBe in the y-axis, in the bottom and middle panels. The upper panels compare the results of GMe with GMd. Each panel quotes the mean Δ and its standard deviation, where Δ = property(CBe) − property(GMe) or Δ = property(GMd) − property(GMe).

GMe-based μ_⋆-values are higher than CBe by 0.27 dex on average, reflecting the different IMF used. Apart from this offset, the two stellar mass surface density agree to within 0.08 dex. Mean extinction is also in good agreement with a dispersion of 0.05 mag. Ages are higher in GMe than CBe by 0.14 dex for ⟨log age⟩_L and 0.08 dex for ⟨log age⟩_M, with dispersion 0.18 dex and 0.12 dex, respectively. This result is expected since base GMe also differs from CBe in IMF and isochrones. The differences in opacities in the equation of state between Padova 2000 (GMe) and 1994 (CBe) tracks produce somewhat warmer stars in the red giant branch in the former. Thus, older ages are expected with GMe than with CBe. However, the metallicities are lower in GMe than in CBe by 0.13 dex for ⟨log (Z_⋆/Z_⊙)⟩_M and very similar (on average) for ⟨log (Z/Z_⊙)⟩_L. In both cases, the dispersion is similar, 0.11 and 0.13 dex, respectively. Note that for Z_⋆ ≥ Z_⊙, the metallicities (weighted in light or in mass) are always higher with CBe than GMe, reflecting the saturation effects in the base GMe due its limitation to Z ≤ 1.5 Z_⊙. The shift at under-solar metallicities may be reflecting the age-metallicity degeneracy, CBe giving higher metallicity and younger ages.

The upper panels compare the results of GMe with the GMd. Here, we see two relevant effects. The results of GMd also differ from GMe in the range of extinctions allowed to starlight. While with GMe and CBe starlight always assumes A_V ≥ 0, with GMd, starlight can bluer the SSP spectra by up to A_V = −0.5 mag. This is allowed to avoid the effect of saturation at A_V = 0. The global effect is that ages can be 0.11 dex younger with GMe than GMd. Metallicity is not affected by this choice of A_V; but the extension to ages older than the age of the Universe in GMd has some effect on the metallicity above Z_⊙, so metallicities are slightly lower and ages slightly older.

Appendix C: 2D maps and tables

Appendix C.1: 2D maps of logμ_⋆

	Fig. C.1 2D maps of stellar mass surface density, μ⋆. Each galaxy is placed in its location in the u − r vs. Mr diagram, where color and magnitude correspond to its global values. The 2D maps are shown with north up and east to the left.
Open with DEXTER

Figure C.1 shows the M_r vs. u − r CMD for the 300 CALIFA galaxies of our sample. Each galaxy is represented by its 2D map of the log μ_⋆ located at the position of its integrated M_r and u − r values. In this plot, the SFH is compressed into the present-day stellar mass surface density, which measures the end product of the SFH. Because our analysis accounts for extinction, these log μ_⋆ values and their radial variations are free from extinction effects. Figure C.1 clearly shows¹⁴ that log μ_⋆ correlates with M_r, and spheroids are significantly denser than late-type galaxies by one to two orders of magnitude at the center, and by one order of magnitude at distances 1−2 HLR. At the center, 2.0 ≤log μ_⋆ (M_⊙ pc^-2) ≤ 4.7, while ≤log μ_⋆ (M_⊙ pc^-2) ≤ 3.4 at 1 HLR, and 1.0 ≤log μ_⋆ (M_⊙ pc^-2) ≤ 2.9 at 2 HLR.

Appendix C.2: 2D maps of ⟨log age⟩_L

Similarly to Figs. C.1, C.2 shows the 2D maps of ⟨log age⟩_L. It portrays the correlation between the average age of the stellar populations and M_r and colors, with the most luminous and red galaxies being older, while the bluest galaxies are the youngest. Gradients of the stellar population ages are also clearly detected within each galaxy in these 2D maps, and more remarkably in galaxies located in the green valley. At the center, 7.3 ≤⟨log age⟩_L (yr) ≤ 10.1, while 8.3 ≤⟨log age⟩_L (yr) ≤ 10.1 at 1 HLR, and 7.5 ≤⟨log age⟩_L≤ 9.9 at 2 HLR.

	Fig. C.2 As Fig. C.1 except for images of the luminosity-weighted mean age, ⟨log age⟩L.
Open with DEXTER

Appendix C.3: 2D maps of ⟨log Z_⋆⟩_M

Similar to Figs. C.1, C.3 presents the 2D maps of ⟨log Z_⋆⟩_M. It clearly shows the correlation between the stellar metallicity and galaxy luminosity, equivalent to the mass metallicity

relation. Gradients of the stellar metallicities are more clearly seen in these 2D maps in galaxies with intermediate luminosity (−22 ≤ M_r ≤ −20). The most luminous galaxies have solar or over solar metallicity producing a visual saturation in the 2D maps. The stellar metallicities range from ⟨log Z_⋆⟩_M = –1.4 to 0.22.

	Fig. C.3 As Fig. C.1 except for images of the mass-weighted mean metallicity, ⟨log Z⋆⟩M.
Open with DEXTER

Appendix C.4: 2D maps of A_V

Similarly to Figs. C.1, C.4 presents 2D maps of A_V. Effects of spatial binning are visible in the A_V maps, where all the pixels within a Voronoi zone have the same value. These effects are not noticeable in the μ_⋆ images (Fig. C.1) because μ_⋆ is an extensive property, and the zoning effect was softened by scaling the value at each pixel by its fractional contribution to the total flux in the zone; this is not possible for A_V (Fig. C.4), ⟨log age⟩_L (Fig. C.2), or ⟨log Z_⋆⟩_M (Fig. C.3), because these are intensive properties. Figure C.4 shows how A_V changes across the CMD: the most luminous galaxies are little affected by extinction, while A_V is higher in spirals of intermediate type and with blue colors. The mean (dispersion) A_V values at the nuclei, 1 HLR, and 2 HLR are 0.47 (0.37), 0.19 (0.16), and 0.13 (0.13), respectively. 2D maps and the difference in the mean values at different distance indicate that stellar extinction shows radial gradients.

	Fig. C.4 As Fig. C.1 except for images of the stellar extinction, AV [mag].
Open with DEXTER

© ESO, 2015