Radial Star Formation Histories in 32 Nearby Galaxies

Staudaher et al. (2019) describe in detail how the near-infrared mosaics were constructed for EDGES galaxies. The mosaics are quite large, tracing the 3.6 μm emission out to at least five times the optical radius a₂₅.⁸ Additionally, the EDGES near-infrared mosaics probe to fainter surface brightness levels than other Spitzer/IRAC imaging campaigns of nearby galaxies. The EDGES program achieved 1800 s integration per position on the sky, 7.5 times longer than the integrations for the Spitzer Infrared Nearby Galaxies Survey (SINGS; Kennicutt et al. 2003), Local Volume Legacy (LVL; Dale et al. 2009), and S⁴G (Sheth et al. 2010) surveys, and 12–30 times longer than the Infrared Array Camera Guaranteed Time Observations (IRAC GTO) project (Pahre et al. 2004). We reach a 1σ per pixel sensitivity of 2 kJy sr⁻¹, and averaging over annuli spanning several square arcminutes can reach down to sensitivities below 0.4 kJy sr⁻¹ (fainter than 29 mag arcsec⁻² AB), a level necessary for securely detecting the faint outer substructures associated with nearby galaxies (Purcell et al. 2007; Krick et al. 2011; Barnes et al. 2014; Staudaher et al. 2015). For comparison, Wide-field Infrared Survey Explorer (WISE) achieves a 3.4 μm diffuse sensitivity of at least 1 kJy sr⁻¹ over a 5' × 5' area (Wright et al. 2010; Jarrett et al. 2019).

Archival ultraviolet and infrared space-based data were gathered from the Galaxy Evolution Explorer (GALEX; 0.15 and 0.23 μm), Spitzer (8.0 and 24 μm), WISE (12 and 22 μm), and Herschel Space Observatory (70 μm) archives. The ultraviolet observations with GALEX primarily trace continuum emission from massive stars. The majority of the far- and near-ultraviolet images utilized here arises from integrations longer than 1 ks; we only rely on GALEX All-Sky Imaging Survey data, which were obtained with integrations shorter than 1 ks, for NGC 5523, NGC 5608, and UGC 8303. The infrared images employed here for the energy-balanced SED fitting (Section 4.2) trace polycyclic aromatic hydrocarbon emission and the underlying dust grain continuum (e.g., Smith et al. 2007) in the case of 8.0 and 12 μm, or chiefly warm dust in the case of 22, 24, and 70 μm emission. Dale et al. (2016) review the surface brightness sensitivities for these archival data. These ancillary/archival data sets have native angular resolutions of ∼5''–6'', or they were smoothed to this resolution.

New deep ugr imaging was obtained for 17 EDGES galaxies on the WIRO 2.3 m telescope with the WIRO DoublePrime camera (Findlay et al. 2016) over the course of the summer of 2018. The ground-based WIRO data previously obtained for 15 additional EDGES galaxies are described in Dale et al. (2016). DoublePrime is a four-amplifier 4096 × 4096 camera with ∼058 pixels, for an overall field of view of 39' × 39'. For each galaxy and each filter 12 individual 300 s frames were taken. Individual frames were randomly dithered with small offsets for enhanced pixel sampling. The typical atmospheric seeing was 15–20. Each night a series of zero second bias frames were obtained in addition to a series of twilight sky flats within each filter.

The optical images were processed with standard procedures, including subtraction of a master bias image and removal of pixel-to-pixel sensitivity variations through flat-field corrections. The 12 dithered 300 s frames for a galaxy taken in one filter were aligned and stacked (summed), resulting in images with integrations equivalent to one hour. The stacked images are flat to 1% or better on 20' scales. The limiting surface brightnesses are ∼28 mag arcsec⁻² AB. The astrometric solutions and flux zero-points were calibrated using positions and photometry extracted from Sloan Digital Sky Survey (SDSS; York et al. 2000) imaging on several (N ≳ 15) foreground stars spread across each image stack. The uncertainties in the zero-point calibrations were typically 2%. Similar to what was done for the imaging at all other wavelengths studied here, foreground stars and background galaxies were removed from each optical image using IRAF/IMEDIT and a local sky interpolation. This editing typically reached down to sources with flux densities of several microJanskys (∼21–22 mag AB).

Multiwavelength photometry was carried out for each galaxy using a series of six concentric elliptical annuli. The annuli cover semimajor axis a ranges extending to 1.5 times the de Vaucouleurs radius a₂₅, namely 0 < $\tfrac{a}{{a}_{25}}$ < 0.25, 0.25 < $\tfrac{a}{{a}_{25}}$ < 0.5, 0.5 < $\tfrac{a}{{a}_{25}}$ < 0.75, 0.75 < $\tfrac{a}{{a}_{25}}$ < 1, 1 < $\tfrac{a}{{a}_{25}}$ < 1.25, and 1.25 < $\tfrac{a}{{a}_{25}}$ < 1.5. The results presented in Section 5 are effectively insensitive to the choice of annular widths; similar results are obtained when using annular widths that are narrower or wider by a factor of 1.5. All annuli for a given galaxy used the same (NED-based) centroids, position angles, and ellipticities. Photometric uncertainties _total are computed by summing in quadrature the calibration error and the measurement uncertainty as described in Dale et al. (2016). All fluxes were corrected for Galactic extinction adopting the results of Schlafly & Finkbeiner (2011) by assuming ${A}_{V}/E(B-V)\approx 3.1$ and the reddening curve of Draine (2003).

Fitting theoretical models to observed SEDs has become a common technique for estimating the physical properties of galaxies (Hunt et al. 2019). We fit the broadband SEDs for our 32 galaxies using the CIGALE software package and its large grid of models (Boquien et al. 2019). Though CIGALE allows for the traditional approach of estimating physical parameters via the single best-fit model (χ² minimization), we opt for the Bayesian-like approach that weights all the models in the chosen grid (the priors) based on their likelihood values ($\exp \left(-{\chi }^{2}/2\right)$). The distributions of likelihoods are then used for estimating the physical parameters and their associated uncertainties. CIGALE also invokes an energy-balanced approach in which the amount of energy that is absorbed by dust in the ultraviolet/optical regime reappears in the infrared as dust emission. We utilize the stellar and dust emission libraries of Bruzual & Charlot (2003) and Dale et al. (2014), respectively, the Chabrier (2003) stellar initial mass function, and a dust attenuation curve (dustatt_modified_starburst) based on the work of Calzetti et al. (2000) and Leitherer et al. (2002). The fit parameters, listed in Table 2, include metallicity Z, extinction $E(B-V)$, a power-law exponent δ that modifies the slope of the attenuation curve, and the e-folding decline rate τ of the galaxy's star formation history. The main physical outputs are stellar mass M_*, stellar mass-weighted age, and star formation rate. As described in Dale et al. (2016), we explored single and double exponential star formation histories, but we have opted to present results using the so-called delayed star formation history (delayed-τ) model (Lee et al. 2010, 2011; Schaerer et al. 2013), i.e.,

$$\begin{eqnarray}&&\mathrm{SFR}(t)\propto {{ \mathcal A }}_{0}{{te}}^{-(t-{t}_{0})/\tau },\ \ {{ \mathcal A }}_{0}(t-{t}_{0}\lt 0)=0,\end{eqnarray} \tag{ 1 }$$

where the maximum star formation rate occurs at the value of τ after the onset of star formation: $t-{t}_{0}=\tau$. We prefer the delayed star formation history model since it is a common, simple prescription that relies on a small number of parameters. Moreover, CIGALE-based simulations show that the delayed star formation model provides superior accuracy in recovering galaxy stellar masses and star formation rates (Buat et al. 2014; Ciesla et al. 2015). However, it should be noted that our main conclusions do not change if we opt instead for a single or double exponential star formation history. We fix t₀ = 11 Gyr since the results of the fits proved to be insensitive to reasonable values of t₀. As a result of fixing the onset of star formation t₀, differences in τ directly correspond to differences in the timing of the peak of the star formation histories.

Zoom In Zoom Out Reset image size

Table 2.  Fit Parameters

Parameter	Notation	Allowed Values
Metallicity	Z	0.008, 0.02, 0.05
Stellar library		Bruzual & Charlot (2003)
Initial mass function		Chabrier (2003)
Color excess: young stars	$E(B-V{)}_{* }^{{\rm{y}}}$	0.0, 0.025, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.4
Color excess: old stars	$E(B-V{)}_{* }^{{\rm{o}}}$	$0.44E(B-V{)}_{* }^{{\rm{y}}}$
Dust emission template	α	0.5, 1.0, 1.25, 1.50, 1.75, 2.0, 2.25, 2.50, 3.00
Slope of the power law that modifies attenuation curve	δ	−0.5, −0.4, −0.3, −0.2, −0.1, 0
Delayed Star Formation History
SFR e-folding time (Gyr)	τ	0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8, 10
Age of oldest stars (Gyr ago)	t0	11

Download table as:  ASCIITypeset image

Representative SED fit results are provided in Figure 4 for UGC 7699, the same galaxy portrayed in Figure 3 with our approach to the location of photometric apertures. The fits are systematically low compared to the observations for the GALEX NUV filter and mostly systematically high compared to the Spitzer 8 μm fluxes, but otherwise the fits for all six annular regions are generally good, with reduced χ² values ranging from 0.5 to 1.9.

Zoom In Zoom Out Reset image size

Sample-wide, the flux signal-to-noise values naturally are smaller for the outermost annuli. The median signal-to-noise values, in order of increasing filter wavelength, are [13, 11, 4, 13, 13, 23, 3, 3] for the penultimate annulus and [7, 3, 3, 4, 5, 8, 3, 3] for the outermost annulus. There is no minimum recommended signal-to-noise for CIGALE. In CIGALE, each band is weighted by the inverse variance, and bands with lower signal-to-noise will thus have a lower weight in the computation of the likelihood. In certain situations this weakens the constraints on some of the physical properties, and this will be reflected in larger uncertainties. For example, a lower signal-to-noise in the ultraviolet will primarily increase the star formation rate uncertainties.

Figure 2 was briefly mentioned in Section 2 in the sample description. This figure provides a selection of physical parameters and the ranges spanned by the sample of 32 galaxies studied here—proceeding from the top to bottom: stellar mass, star formation rate surface density, g − r color, and stellar mass-weighted age (see also Figures 2 and 1, respectively, of López Fernández et al. 2018; Casasola et al. 2020). The values displayed in Figure 2 are either obtained directly from the observations in the case of g − r color or they are inferred from the SED fits. These data are based on the integrated fluxes arising from within semimajor × semiminor 1.5a₂₅ × 1.5b₂₅ elliptical apertures (Table 3). Though the statistics suffer from small numbers when the galaxies are binned by morphology, the general trends are unsurprising. The stellar mass is, on average, highest for the S0 and earlier-type spirals and generally decreases for later-type spirals and irregulars. The S0 and early-type spirals are globally redder and older whereas the later-type spirals are bluer and younger.

A primary goal of this effort is to quantify and compare the spatially resolved trends for the inner and outer disks. Figure 5 shows two example trends for the six photometric annuli of UGC 7699 and NGC 4460: the g − r colors and stellar (mass-weighted) ages. The lines provided within the figure represent linear fits to the inner (0 < a < 0.75a₂₅) and outer (0.75a₂₅ < a < 1.50a₂₅) disks; the slope values are provided next to each fitted line in the figure. For Scd galaxy UGC 7699, both the g − r colors and fitted stellar ages decrease over the radial range 0 < a < 0.75a₂₅ and increase over 0.75a₂₅ < a < 1.50a₂₅, consistent with inside-out disk formation for the inner disk and a reversal for the outer disk. The opposite trends are seen for S0 galaxy NGC 4460. The sample-wide statistics for these two parameters are provided in Figure 6 as a function of morphology. Negative (positive) gradients indicate younger (older) with radius (for reference, see the fitted lines and reported slopes in Figure 5). The general trend for the spiral galaxies in our 32 galaxy subset of the EDGES full sample is similar to that seen for UGC 7699 highlighted in Figure 5: negative gradients for the inner disks and positive gradients for the outer disks. The differences between the inner and outer disks for the later-type spiral galaxies can be more clearly seen in the histograms shown in Figure 7.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The picture is less clear for S0 and irregular galaxies. The g − r color trends for S0s and irregulars span both positive and negative slopes for both inner and outer disks. The average inner age gradient for irregulars is negative and the average outer age gradient is positive. At face value this result contradicts the findings that irregular galaxies differ from spirals in that irregulars are more likely to exhibit features consistent with forming in an outside-in fashion (Gallart et al. 2008; Zhang et al. 2012; Meschin et al. 2014; Pan et al. 2015; Sacchi et al. 2018), where their shallow gravitational potentials lead to being more easily influenced by external factors like ram pressure stripping and internal factors such as feedback from stellar winds and supernovae. However, there is significant overlap in the inner and outer age gradient distributions for the irregulars in Figure 6. Interestingly, for the S0s in our sample the age gradient is essentially reversed compared to the spirals: S0s appear to be increasingly older (younger) with increasing radius for the inner (outer) disk. A complicating factor here is that four of the six S0 galaxies host Seyfert (NGC 4138 and NGC 5273) or Low-Ionization Nuclear Emission-Line Region (LINER; NGC 4143 and NGC 4203) nuclei, and thus the observed optical colors and fitted stellar ages for their innermost annuli do not derive purely from star formation. Only one spiral galaxy of the 16 studied here has such a nucleus; NGC 4102 is a LINER. The CALIFA sample is only ∼6% active galactic nuclei (AGN; Walcher et al. 2014).

In their multiwavelength study of 461 DustPedia galaxies, Davies et al. (2019) caution that a delayed star formation history model does not apply well to S0 galaxies. Moreover, the formation histories of lenticulars in particular are thought to be quite diverse: in addition to a portion of S0s having been formed via mergers or the stripping of gas in clusters and groups, other S0 galaxies may have formed through internal secular processes and gas infall (Bellstedt et al. 2017; Eliche-Moral et al. 2018; Coccato et al. 2020). The main conclusion of the Davies et al. (2019) study is that a delayed-τ model and closed-box chemical evolution can be appropriate for late-type spirals but earlier-type galaxies have more complicated evolutionary histories. López Fernández et al. (2018) similarly analyzed CALIFA star formation radial trends according to morphology, also using a delayed star formation history model, though their analysis only extends to twice the half-light optical radius (we typically probe to 4–6 half-light radii). López Fernández et al. (2018) find that E/S0 galaxies are, on average, older (smaller τ) further from the nucleus whereas spiral galaxies are younger (larger τ) farther out. A different analysis of the CALIFA sample, focused on the inner half-light radius, shows that earlier-type galaxies have larger age gradients (García-Benito et al. 2017), echoing our result displayed in Figure 6. In short, the 16 spiral galaxies in our sample differ from the six S0 and 10 irregular galaxies in their radial gradients in optical color and characteristic stellar ages. These gradient differences in their star-forming properties may reflect their disparate formation histories. Though our project has been designed to probe more of the outer disk regions than integral field unit (IFU) surveys like CALIFA, the statistical significance of our results are much less robust. The López Fernández et al. (2018) and García-Benito et al. (2017) efforts, by comparison, respectively involve 48 and 78 S0 galaxies and 264 and 463 spiral galaxies. Note that the López Fernández et al. (2018) and García-Benito et al. (2017) samples exclude Type 1 Seyferts and galaxies that exhibit strong merger or interaction features.