arXiv:0909.2658v1 [astro-ph.CO] 14 Sep 2009
Radial distribution of stars, gas and dust in SINGS galaxies: II.
Derived dust properties
J.C. Mu˜ noz-Mateos1, A. Gil de Paz1, S. Boissier2, J. Zamorano1, D.A. Dale3, P.G.
P´ erez-Gonz´ alez1, J. Gallego1, B.F. Madore4, G. Bendo5, M. D. Thornley6, B. T. Draine7,
A. Boselli2, V. Buat2, D. Calzetti8, J. Moustakas9, R. C. Kennicutt, Jr.10,11
We present a detailed analysis of the radial distribution of dust properties
in the SINGS sample, performed on a set of UV, IR and HI surface brightness
profiles, combined with published molecular gas profiles and metallicity gradients.
The internal extinction, derived from the TIR-to-FUV luminosity ratio, decreases
with radius, and is larger in Sb-Sbc galaxies. The TIR-to-FUV ratio correlates
with the UV spectral slope β, following a sequence shifted to redder UV colors
with respect to that of starbursts. The star formation history (SFH) is identified
as the main driver of this departure. Both LTIR/LFUVand β correlate well with
metallicity, especially in moderately face-on galaxies. The relation shifts to redder
1Departamento de Astrof´ ısica y CC. de la Atm´ osfera, Universidad Complutense de Madrid, Avda. de la
Complutense, s/n, E-28040 Madrid, Spain; jcmunoz, agpaz, jaz, pgperez, email@example.com
2Laboratoire d’Astrophysique de Marseille, OAMP, Universit´ e Aix-Marseille & CNRS UMR 6110,
38 rue Fr´ ed´ eric Joliot-Curie, 13388 Marseille cedex 13, France;
3Department of Physics and Astronomy, University of Wyoming, Laramie, WY; firstname.lastname@example.org
4Observatories of the Carnegie Institution of Washington, 813 Santa Barbara Street, Pasadena, CA 91101;
5Astrophysics Group, Imperial College, Blackett Laboratory, Prince Consort Road, London SW7 2AZ;
7Princeton University Observatory, Princeton, NJ 08544-1001; email@example.com
8Department of Astronomy, University of Massachusetts, Amherst, MA 01003; firstname.lastname@example.org
9Department of Physics, New York University, 4 Washington Place, New York, NY 10003, USA
10Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK
11Steward Observatory, University of Arizona, Tucson, AZ 85721
– 2 –
colors with increased scatter in more edge-on objects. By applying physical dust
models to our radial SEDs, we have derived radial profiles of the total dust
mass surface density, the fraction of the total dust mass contributed by PAHs
and the intensity of the radiation field heating the grains. The dust profiles are
exponential, their radial scale-length being constant from Sb to Sd galaxies (only
∼ 10% larger than the stellar scale-length). Many S0/a-Sab galaxies have central
depressions in their dust radial distributions. The PAH abundance increases
with metallicity for 12 + log(O/H) < 9, and at larger metallicities the trend
flattens and even reverses, with the SFH being a plausible underlying driver
for this behavior. The dust-to-gas ratio is also well correlated with metallicity
and therefore decreases with galactocentric radius. Although most of the total
emitted IR power (especially in the outer regions of disks) is contributed by
dust grains heated by diffuse starlight with a similar intensity as the local Milky
Way radiation field, a small amount of the dust mass (∼ 1%) is required to be
exposed to very intense starlight in order to reproduce the observed fluxes at
24µm, accounting for ∼ 10% of the total integrated IR power.
Subject headings: dust,extinction — galaxies: ISM — infrared: galaxies — ul-
Understanding the spatial distribution of interstellar dust is of particular importance
for two reasons: it affects our view of galaxies at different wavelengths, by absorbing UV and
optical light and reemitting it in the infrared, and it also constitutes an important element
in the chemical evolution of the interstellar medium (ISM). As for the first issue, correcting
for dust extinction is usually the main source of uncertainty when deriving properties such
as the star formation rate (SFR), age or metallicity (Calzetti et al. 1994; Buat & Xu 1996;
Calzetti 2001; P´ erez-Gonz´ alez et al. 2003). This limitation not only applies to integrated
data, but also to surface brightness profiles and color gradients. For instance, the so-called
inside-out scenario for the formation of galactic disks predicts that the timescale of gas
infall and conversion into stars increases with radius, leading to radial variations of the star
formation history (SFH), which are in turn observationally translated into color gradients
(de Jong 1996; Bell & de Jong 2000; MacArthur et al. 2004; Taylor et al. 2005). Given that
the dust content also changes with radius, it is key to properly quantify that variation with
direct tracers of dust extinction; otherwise, derived parameters such as the radial growth
rate of disks might be biased (see Mu˜ noz-Mateos et al. 2007 and references therein).
– 3 –
While being an inconvenience when studying star formation, dust itself is also a key
ingredient in the chemical enrichment of the ISM. Metals resulting from stellar nucleosyn-
thesis are returned to the ISM, where they condense to form dust grains, some of which
are later destroyed and incorporated into new generations of stars. Elements that later con-
dense to form dust grains are injected into the ISM at different rates (see e.g. Dwek et al.
2009 and references therein). For instance, carbon now locked up into Polycyclic Aromatic
Hydrocarbons (PAHs) might have been originally produced in AGB stars, whose typical
lifetimes are a few Gyr. Other elements now constituting larger grains may have been syn-
thesized in more massive stars, which die as supernovae in shorter timescales. Thus, the
PAH abundance and the dust-to-gas ratio are expected to vary with the age of the stellar
populations and correlate with the metal (oxygen) abundance of the gas (see e.g. Galliano
et al. 2008 and references therein). Note, however, that dust formation results from a long
chain of poorly-understood processes, of which stellar lifetimes and yields are only the first
link. Grain growth and destruction in the ISM must be also considered. Indeed, it is thought
that only 10% of interstellar dust is directly formed in stellar sources, with the remaining
90% being later condensed in the ISM (see Draine 2009 for a recent review on the subject).
Addressing these important issues requires a multi-wavelength, multi-object spatially
resolved analysis, which is now possible for nearby galaxies thanks to the Spitzer Infrared
Nearby Galaxies Survey (SINGS, Kennicutt et al. 2003). The SINGS project has made
use of the Spitzer Space Telescope (Werner et al. 2004) to collect IR data, as well as an-
cillary data from other facilities, for a sample of 75 nearby galaxies, representative of the
galaxy population in the local universe. Together with UV images from the Galaxy Evo-
lution Explorer (GALEX, Martin et al. 2005) and HI maps from The HI Nearby Galaxies
Survey (THINGS, Walter et al. 2008), this data-set provides the scientific community with
unprecedented spectral and spatial coverage of the most representative nearby galaxies.
In this paper we exploit the unique data-set resulting from the GALEX-SINGS-THINGS
collaboration by obtaining multi-wavelength radial profiles for the 57 galaxies in the SINGS
sample which are detected and resolved at the longest wavelengths. Our subsample still in-
cludes galaxies representative of different morphological types (ellipticals, lenticulars, spirals
and irregulars). The profiles themselves, along with other observational parameters such as
asymptotic magnitudes, concentration indexes and asymmetries, can be found in the accom-
panying paper Mu˜ noz-Mateos et al. (2009, Paper I hereafter). The present paper focuses on
the radial distribution of dust properties. Finally, by comparing the profiles of the spirals in
the sample with models for the chemical and spectro-photometric evolution of spiral galaxies
(Boissier & Prantzos 1999), we are studying the radial variation of the star formation history
in these galaxies in a self-consistent frame, where radial changes in the gas-infall and chemi-
cal enrichment are considered. The attenuation profiles turn out to be essential to properly
– 4 –
accommodate the model predictions with the observed profiles. Ignoring extinction not only
biases the overall flux level in the UV and optical bands, but also the different radial scale-
lengths at these wavelengths, which are a direct result of the inside-out assembly of disks.
The results of this comparison with theoretical models will be presented in a forthcoming
paper and, combined with the present study of the dust properties, should help to better
understand how the different components of spiral galaxies get assembled together.
This paper is organized as follows. In Section 2 we briefly outline the main properties of
the sample, describe the different data used in this work, and explain how the radial profiles
were obtained. Section 3 deals with the radial distribution of internal extinction, and how
it relates with other properties such as the UV color, inclination or metallicity. In Section 4
we compare our profiles with the dust models of Draine & Li (2007) (DL07 hereafter), and
analyze the radial variation of the PAH abundance, the dust mass and luminosity surface
densities, the intensity of the heating radiation field and the dust-to-gas ratio. Finally, we
summarize our main conclusions in Section 5. In Appendix A we provide some empirical
relationships between the dust properties derived from the model and the observed flux
densities at the near- to far-infrared bands. In Appendix B we discuss the possible systematic
factors that might affect our results in deriving the parameters of the dust models of DL07.
2.The sample, data and procedure
The SINGS sample (Kennicutt et al. 2003) consists of 75 nearby galaxies which span
the range in morphological type, luminosity and FIR/optical luminosity observed in the
local universe. The SINGS galaxies were also selected to cover a reasonably wide range in
other additional properties, like nuclear activity, spiral and bar structure, inclination, surface
brightness and environment. Note, however, that no significant luminous or ultra-luminous
infrared galaxy (that is, with LIR> 1011L⊙) is included the sample. The median distance of
the SINGS galaxies is 10Mpc, with all objects being closer than 30Mpc. Since the sample
is neither flux- nor volume-limited, its statistical power as a whole is limited. Nevertheless,
the wealth of panchromatic data available for these galaxies, together with their proximity,
makes it possible to carry out detailed studies on the physics of star formation at kpc scales,
including the interplay between star formation and the ISM.
Eighteen SINGS galaxies were not suitable for our purposes and were excluded from the
1. Seven galaxies, DDO 154, DDO 165, Holmberg IX, M81 Dw a, M81 Dw b, NGC 0584
and NGC 4552, are not (or marginally) detected in at least one of the MIPS bands.
– 5 –
2. Nine galaxies, Mrk 33, NGC 1266, NGC 1377, NGC 1482, NGC 2798, NGC 3265,
NGC 3773, NGC 5195 and NGC 7552, are unresolved in the MIPS bands. Thus,
radial profiles for these galaxies would reflect the shape of the PSF rather than real
structures. In NGC 7552 some spatially extended emission can be discerned, but it is
too faint compared to the extremely bright nucleus.
3. The only evident IR source within the optical extent of NGC 1404 at 70 and 160µm
is an off-center object at the northeast of the frame. Since it might not be related to
NGC 1404, we also excluded this galaxy.
4. The bright center of NGC 3034 saturates the MIPS 24µm detector, thus precluding
Therefore, the final subsample consists of 57 galaxies, whose main properties are sum-
marized in Table 1.
Nearly all SINGS galaxies have been observed in the FUV (λeff= 151.6nm) and NUV
(λeff = 226.7nm) by GALEX (Martin et al. 2005) (see Table 1). The observations are
performed simultaneously at both bands thanks to a beam splitter, but some galaxies lack
FUV images since the corresponding detector had to be turned off due to intense solar activity
or overcurrent events. The delivered images have a final pixel-scale of 1.5′′. Although the
size of the PSF depends slightly on the position on the detector and the brightness of the
source, the typical FWHM is 5-6′′. This resolution is similar to that of the MIPS 24µm
images, and corresponds to a physical scale of ∼300pc at 10Mpc, the median distance of
the SINGS galaxies. The flux calibration is based on white dwarf standard stars. For the
pipeline version used here (the same as in Gil de Paz et al. 2007), the estimated zero-point
uncertainty is 0.15mag.
The reader is referred to Paper I for a more detailed description of the Spitzer data used
here. Mid- and far-infrared observations of the SINGS sample were carried out using the
Spitzer Space Telescope (Werner et al. 2004). The Infrared Array Camera (IRAC, Fazio et
al. 2004) was used to image the SINGS galaxies at 3.6, 4.5, 5.8 and 8.0µm. We made use of
the images provided in the SINGS Fourth Data Delivery, which are based on the Version 13
Basic Calibrated Data produced by the Spitzer Science Center. The delivered images have
– 6 –
a pixel scale of 0.75′′, and the FWHM of the PSF at each band are 1.7′′, 1.7′′, 1.9′′and 2.0′′
respectively. These correspond to spatial scales of 80-100pc at the median distance of the
sample. The estimated photometric uncertainty is ∼ 2% (Reach et al. 2005). However, the
photometry needs to be corrected for the diffuse scattering of incoming photons throughout
the IRAC array1, and these corrections have an associated uncertainty of ∼ 10%.
The Multi-band Imaging Photometer (MIPS, Rieke et al. 2004) was used in scan-
mapping mode to observe the SINGS galaxies at 24, 70 and 160µm. The final frames are
delivered with pixel scales of 1.5′′, 4.5′′and 9.0′′, respectively, thus being integer multiples
of the pixel-scale of the IRAC frames while still properly mapping the MIPS PSF. The cor-
responding FHWM are 5.7′′, 16′′and 38′′at each band, probing physical scales of 0.28, 0.78
and 1.84kpc at 10Mpc. The estimated zero-point errors are 4%, 5% and 12% at 24, 70 and
160µm, respectively (Engelbracht et al. 2007; Gordon et al. 2007; Stansberry et al. 2007).
The HI Nearby Galaxy Survey (THINGS, Walter et al. 2008) used the Very Large Array
(VLA) to map HI 21-cm line emission from 34 nearby (D < 15Mpc) galaxies, most of which
were also targets of SINGS and the GALEX Nearby Galaxies Survey. The observations
were done using the B-, C- and D-array configurations. For details of data reduction and
processing, see Walter et al. (2008). Here, we use profiles derived from natural-weighted
moment-zero (integrated intensity) maps. These have a typical angular resolution of 11′′and
sensitivity to surface densities as low as 4 × 1019cm−2once convolved to our 38′′working
resolutions (see below). These radial profiles were kindly provided by A. Leroy and F.
Walter. Because THINGS includes data from the compact B array configuration, the maps
comfortably recover extended structure in our sources. The column densities in the THINGS
maps are estimated to be correct to within ±10%.
2.4. Surface brightness profiles
The reader is referred to Paper I for a more in-depth description of the procedure
followed to obtain the surface brightness radial profiles; here we will just briefly describe the
most important steps relevant to this paper. In our study of the spatial distribution of the
dust properties, we are limited by the resolution of the MIPS 160µm images (FWHM of 38′′).
– 7 –
We convolved the GALEX, IRAC, MIPS and HI images with different kernels (see Gordon
et al. 2008) in order to match the shapes and resolution of their PSFs to the MIPS 160µm
PSF. Prior to convolving the images, foreground stars, background galaxies and artifacts
were masked and interpolated over to avoid contamination when degrading the images.
The radial profiles were obtained using the IRAF2task ellipse, measuring the mean
intensity along elliptical isophotes with fixed ellipticity and position angle, equal to those of
the µB= 25 mag arcsec−2isophote from the RC3 catalog3(de Vaucouleurs et al. 1991). For
those objects for which these parameters were not included in the RC3 catalog, we used the
major and minor axis diameters and position angles available in NED. The centers of these
elliptical isophotes were set at the coordinates shown in Table 1. The semi-major axis of
these ellipses were successively incremented by 48′′(a step larger than the PSF FWHM), to a
final radius at least 1.5 times the R25 radius (depending on the extension of each particular
galaxy). While using radially-varying ellipticities and position angles is useful in detailed
studies of galactic structure at a specific wavelength, a panchromatic analysis requires using
the same set of fixed elliptical isophotes in all bands to measure the different fluxes in the
same regions of each galaxy, and for that matter a fixed PA and ellipticity was found to be
as appropriate as any other set of values obtained from a given band.
Uncertainties in the surface photometry include the error of the mean intensity within
each isophote, computed assuming Poisson statistics, and the uncertainty in the sky level.
The latter comes from high spatial frequency errors (Poisson noise, pixel-to-pixel variations)
and low spatial frequency ones (flat-fielding errors) (Gil de Paz & Madore 2005).
The radial profiles were corrected for Galactic extinction as in Dale et al. (2007), using
the color excesses from the maps of Schlegel et al. (1998) and the extinction curve of Li &
Draine (2001), assuming RV = 3.1. The final profiles are shown in Table 2.
Note that our radial profiles should be taken with caution in very inclined galaxies.
First, although the radial step along the major axis is larger than the FWHM of the PSF,
this is not the case along the minor axis in edge-on or close to edge-on galaxies. Moreover,
the outer regions might be contaminated by emission from the central ones when performing
the azimuthal average. Finally, the observed UV and IR radiation in these systems will likely
probe different spatial regions within each galaxy, due to the large amount of dust along the
line of sight.
2IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the As-
sociation of Universities for Research in Astronomy, Inc., under cooperative agreement with the National
3Except for NGC 5194, whose original values were highly affected by its companion galaxy, NGC 5195.
– 8 –
3.Radial distribution of dust attenuation
3.1. Radial extinction profiles
As part of our analysis of the radial distribution of dust properties, we will first determine
the radial variation of internal extinction, which is also necessary to recover the intrinsic
radial profiles from FUV to NIR. These kinds of studies have been addressed by several
authors following different methodologies. Boissier et al. (2004, 2005, 2007) used the radial
change in the total-infrared (TIR) to far-ultraviolet (FUV) ratio to derive radial extinction
profiles for nearby galaxies. A similar procedure was followed by Popescu et al. (2005) on a
pixel-to-pixel basis for M 101. Prescott et al. (2007) obtained extinction profiles in Hα for
the SINGS galaxies by comparing the Hα and 24µm fluxes of individual star-forming regions.
The number of distant galaxies seen through a spiral disk can also provide an independent
estimation of the extinction (e.g. Holwerda et al. 2005 and references therein).
Here we follow the first method to infer the radial distribution of dust attenuation for
the SINGS galaxies. Several studies (Buat & Xu 1996; Meurer et al. 1999; Gordon et al.
2000; Witt & Gordon 2000; Buat et al. 2005) have shown that the TIR-to-ultraviolet ratio
is a robust tracer of the internal extinction in star forming galaxies, in the sense that it
depends weakly on details such as the relative geometry of stars and dust, the shape of the
extinction curve, or the star formation history (SFH). Regarding this latter issue, Buat et
al. (2005) only found significant deviations in systems with very quiescent SFH at present-
day (that is, havig decayed with exponential time-scales ≤ 2Gyr). In these galaxies, the
general starlight radiation field would become an even more important source of dust-heating
than it already is in normal star-forming galaxies. This precludes computing the internal
extinction in ellipticals, lenticulars and also the bulges of spirals with the same calibration.
Using synthetic SEDs of galaxies with different SFHs and attenuations, Cortese et al. (2008)
further investigated the dependence of the TIR-to-FUV ratio on the mean age of the stellar
populations, and confirmed that quiescent systems exhibit larger TIR-to-FUV ratios than
more actively star-forming ones with the same extinction. To account for the extra dust
heating contributed by older stars, they provide a SFH-dependent calibration to estimate
the UV attenuation. Note that, as pointed out by these authors, this recipe should not
be blindly applied to ellipticals, since their FUV emission does not seem to be linked with
recent star formation activity. We include these galaxies here just for completeness, but any
further interpretation of their observed TIR-to-FUV in terms of attenuation should be done
Using the low (48′′-step) resolution profiles at 8, 24, 70 and 160µm, we built TIR
(3-1100µm) profiles using the weighted sum proposed by DL07, after having subtracted
– 9 –
the stellar emission at 8 and 24µm (see Section 4.2). This estimator of the total-infrared
luminosity constitutes a slight improvement over that of Dale & Helou (2002) using the
MIPS bands alone, to the degree that including the 8µm flux seems to reduce the scatter
associated with differences in PAH abundances4.
We then measured low resolution profiles on the convolved GALEX FUV and NUV
images and estimated AFUVand ANUVfrom the TIR-to-FUV and TIR-to-NUV ratios, re-
spectively. The attenuation at both wavelengths was computed using both the fits of Buat
et al. (2005) and those of Cortese et al. (2008). Note that at small subgalactic scales, ra-
diative transfer could limit the usefulness of the TIR-to-UV ratio as an extinction tracer, in
the sense that UV photons emerging from a given region could heat dust in another region.
Considering that our radial sampling is relatively coarse and that we average our data within
elliptical annuli, this effect −if present− should be small (except maybe in extreme objects
with very different UV and IR distributions, which are not common in the SINGS sample)
The extinction profiles we derive are shown in Table 3. In Fig. 1 we show all the FUV
extinction profiles for the galaxies in the sample, normalized to the optical radius R25. A
general trend of AFUV decreasing with radius is clearly seen. The NUV profiles are not
shown, since they exhibit similar behavior as the FUV ones.
The extinction in the top panels has been computed with the calibration of Buat et al.
(2005), whereas in the bottom ones we have used the SFH-dependent recipe of Cortese et
al. (2008). While the equation relating LTIR/LFUVand AFUVis unique in the case of Buat
et al. (2005), Cortese et al. (2008) parameterized the coefficients of such a conversion as a
function of observable colors that depend on the overall SFH. We followed the prescriptions
given by these authors and determined the particular conversion between LTIR/LFUVand
AFUVfor each annulus depending on its observed (FUV−H) color. The latter was estimated
assuming that (H − 3.6µm) ∼ −1.03 which, according to the stellar populations models of
Bruzual & Charlot (2003), is the typical color exhibited by star-forming galaxies, and it is
quite independent of their SFH.
In S0/a-Sab galaxies and the bulges of later-type galaxies, the attenuation derived from
the age-dependent calibration is ∼ 0.5mag lower than the one yielded by the age-independent
4We have checked that the particular choice of this recipe is not critical, since both estimators yield
almost equal TIR luminosities (on average, the TIR values derived from the DL07 calibration are 5% larger,
with a rms of 5%). The DL07 recipe is a proxy for the actual TIR luminosity derived from the DL07
models, designed to be valid over a wide range of starlight intensities and PAH abundances. In Section 4 we
will derive LTIR from the model fitting, along with the remaining model parameters. On average, for the
particular range of starlight intensities and PAH abundances in our sample, the TIR luminosities yielded by
the estimator are only 6% larger than those from the model, with a rms of 3%.
– 10 –
recipe. The difference is negligible in the disk-dominated regions. Given that the SINGS
galaxies span a wide range in star formation histories, both among them and within different
regions of the same object, the results discussed in the rest of the paper refer to the extinction
derived following the method of Cortese et al. (2008), except when mentioned otherwise.
The overall level of extinction varies along the Hubble sequence, reaching a maximum
in Sb-Sbc galaxies. On average, in these spirals the attenuation in the FUV ranges from
∼ 2.5mag in the central regions to ∼ 1.5mag in the outer ones, although with large scatter.
The extinction is ∼ 1mag lower in earlier spirals, and it goes below 0.5mag in Sdm spirals and
irregulars. Note again that the attenuation in ellipticals and lenticulars is highly uncertain,
since even the age-dependent recipes of Cortese et al. (2008) might fail in these galaxies.
In order to quantify the attenuation radial gradients, we performed a linear fit to the
AFUV(r) profiles, without including the bulges of spirals for the reasons mentioned above.
This exclusion was done visually: bulges produce a steep central rise of the 3.6µm luminosity
above the exponential disk, and the FUV emission is significantly reduced in the central
regions. This yields a sharp change in the (FUV−3.6µm) that, together with a visual
inspection of the image, can be used to roughly delimit the bulge- and disk-dominated
regions of the profiles (see also Paper I). The results are shown in Fig. 2, where the gradients
are also expressed in terms of the R25 radius and the radial exponential scale-length of
the 3.6µm profiles, defined so that I3.6µm ∝ e−r/α3.6µm.
computed by fitting the profiles measured on the convolved 3.6µm images (again, after
excluding the bulges). Since at this wavelength the luminosity traces the stellar mass, this
can be considered as the stellar mass scalelength. Most galaxies exhibit negative attenuation
gradients, and the dispersion is larger in spirals of intermediate types.
The scale-length α3.6µm was
3.2. The IRX-β relation in normal disks
3.2.1. Estimating the extinction from UV data alone
In spite of the importance of correcting UV data for internal extinction, when FIR
data are not available it is not possible to apply such corrections following the methods
described in Section 3.1. In this regard, the slope of the UV spectrum β −or, equivalently,
the (FUV−NUV) color− has been proposed as an indirect tracer of dust attenuation in
starburst galaxies (Calzetti et al. 1994; Heckman et al. 1995; Meurer et al.1995, 1999).
While the infrared excess (IRX) relative to the UV seems to be tightly correlated with β
in starbursts, later studies (Bell 2002; Buat et al. 2005; Seibert et al. 2005; Cortese et al.
2006; Gil de Paz et al. 2007; Dale et al. 2007) have shown that the so-called IRX-β relation
– 11 –
shows rather large scatter in normal star-forming galaxies. A similar behavior has also been
observed for star-forming regions within galaxies (Calzetti et al. 2005). The correlation shows
a wider spread, and is also systematically shifted to redder (FUV−NUV) colors. Using radial
profiles from GALEX and IRAS, Boissier et al. (2007) confirmed that the IRX-β relation
differs between starburst and normal star-forming galaxies. The relation is clearly shifted
with respect to that for starburst galaxies, but the dispersion in their case seems to be
reduced when using radial profiles instead of integrated photometry, possibly due to not
mixing different stellar populations from the bulge and the disk.
Here we extend the work carried out in that study by deriving a new radially resolved
IRX-β diagram using data from Spitzer, which has better angular resolution than IRAS, and
also probes colder dust by reaching a bit further into the FIR. An analysis of the IRX-β
plot for the SINGS galaxies, derived from integrated photometry, can be found in Dale et
Throughout this paper we assume the definition of the UV spectral slope βGLXgiven by
Kong et al. (2004):
βGLX=log(fλ,FUV) − log(fλ,NUV)
log(λFUV) − log(λNUV)
= 2.201(FUV − NUV) − 2 (1)
In Fig. 3a we show the IRX-β plot for each of the low resolution radial profiles for the
SINGS galaxies, classified according to their morphological type. Each data-point represents
a given radial bin, and those belonging to spiral galaxies are also color-coded according to
the (FUV−3.6µm) color of that bin (corrected for internal extinction), which can be used as
a measure of the specific star formation rate (SFR per unit of stellar mass, sSFR hereafter)
or, equivalently, the present to past-averaged star formation rate, usually referred to as the
birthrate parameter b (Scalo 1986). This color scheme cannot be applied to ellipticals, since
their FUV flux is not necessarily linked to star formation (Burstein et al. 1988; O’Connell
1999; Boselli et al. 2005). As for the irregulars, the relative distribution of dust and stars
is usually patchy (and not necessarily axisymmetric), so their LTIR/LFUVprofiles must be
considered with caution. As a reference, the right vertical axis shows the extinction in the
FUV derived from LTIR/LFUVwith the fit of Buat et al. (2005), although the color-coding
is based on the (FUV−3.6µm) color corrected for attenuation using the SFH-dependent
calibration of Cortese et al. (2008). We also show the mean IRX-β for the starburst galaxies
studied by Meurer et al. (1999), as given by Kong et al. (2004), and the fit provided by
Boissier et al. (2007) for the radial profiles of normal spirals.
Although differences in SFH can introduce systematic deviations in this plot (see next
section), it is desirable to get at least a rough estimate of the internal extinction for normal
– 12 –
spirals when FIR data are not available, as has been traditionally done for starbursts. Since
we are interested in obtaining a fit valid for normal star-forming spirals, we have excluded
three galaxies with intense starburst activity (NGC 4536, NGC 4631 and NGC 5713). More-
over, we have only considered regions with (FUV − NUV) < 0.9, thus avoiding the large
scatter in the TIR-to-FUV ratio at redder UV colors. We applied a non-linear least-squares
algorithm to the remaining data-points, giving:
LTIR/LFUV= 100.30+1.15(FUV−NUV)− 1.64(2)
The resulting fit is shown in Fig. 3a, and allows for the recovery of log(LTIR/LFUV) for
normal star-forming galaxies with a residual rms uncertainty of ±0.27dex. Instead of im-
posing a UV-color condition, we could have used the extinction-corrected (FUV−3.6µm)corr
color to exclude the most quiescent systems. However, such a criterion cannot be applied by
an observer lacking FIR data, which are necessary to correct that color for internal extinction.
While our UV condition includes a few ‘red’ systems [in terms of their (FUV − 3.6µm)corr
color], they lie well within the main relation delineated by the rest of points. The resulting fit
is not significantly sensitive to these few points, and this selection on the observed UV color
is more justified from a purely empirical point of view. Beyond (FUV − NUV) = 0.9mag,
the observed dispersion is too large to reliably estimate AFUVusing this fit.
Note that our fit differs from the one derived by Boissier et al. (2007) from GALEX and
IRAS data, the latter having a somewhat flatter slope at the reddest UV colors, possibly due
to data-points embedded within the bulges (no conditions on the UV color were imposed
when computing that fit). Nevertheless, both fits are in very good agreement in the regions
least affected by contamination from older stellar populations. By comparing the TIR-to-
FUV profiles of the galaxies we have in common with the sample of Boissier et al. (2007), we
have checked that the agreement between the TIR-to-FUV ratios is excellent. On average,
our values are 0.04dex higher, with a scatter of 0.13dex.
In brief, in order to determine the attenuation in the FUV one must first estimate
the log(LTIR/LFUV) ratio. If FIR data are available, they should be used to directly com-
pute log(LTIRusing, for instance, the calibrations of Dale & Helou (2002) or DL07. Once
log(LTIR/LFUV) is known, it can be translated into AFUVusing any of the recipes available
in the literature (e.g. Buat et al. 2005). In order to prevent the attenuation in early-type
spirals from being overestimated, one can rely on age-dependent calibrations such as that of
Cortese et al. (2008), which requires additional constraints such as optical and/or near-IR
measurements. In the absence of FIR data, AFUVcan be estimated from the (FUV−NUV)
colour via Eq. 2, but such a relation should be only employed in a statistical sense for rela-
tively large samples of galaxies. Given the large scatter, the use of Eq. 2 is discouraged for
– 13 –
3.2.2.Dependence on the star formation history
In Fig. 3a, only a few data-points follow the starbursts relation, and they indeed cor-
respond to galaxies with intense star formation activity (e.g. NGC 4536, NGC 4631). Most
galaxies, however, lie to the right in the diagram: that is, for the same amount of attenua-
tion, they have redder UV colors than starbursts. Differences in the star formation history
are the most likely explanation for this broadening (Kong et al. 2004, Calzetti et al. 2005).
Besides the effects of dust, regions with more quiescent SFHs will be intrinsically redder in
the UV due to their more evolved stellar populations. Indeed, there is a clear trend with
the (FUV − 3.6µm)corr color, which, as noted above, is a proxy for the present to past-
averaged star formation rate. Note that the ‘present’ SFR, as derived from the FUV, is
not instantaneous, but represents an average over the last ∼ 100Myr. Data-points with
(FUV − 3.6µm)corr< 3mag follow a reasonably well-defined sequence, parallel to the star-
burst one; but redder regions (usually embedded within bulges) depart towards the zone
populated by ellipticals.
The three dotted curves in Fig. 3a show the IRX-β relation predicted by Kong et al.
(2004) for galaxies with different values of the birthrate parameter b. The comparison with
(FUV−3.6µm)corrshould be done with caution, since the birthrate parameter of the models
is the instantaneous one. The empirical relation for starbursts closely follows the model
predictions for b = 5, which falls in the range where starbursts are commonly found (b > 2-3;
see Brinchmann et al. 2004 and references therein). The bulk of our data-points are consistent
with lower values of b. These model predictions for AFUV, however, should be considered just
as average approximations, given that their uncertainties range from ±0.3mag for b ? 0.3
to ±1mag for lower values, owing to differences in the particular details of the SFH and the
dust content (Kong et al. 2004).
We further explore this trend in Fig. 4, where we have plotted the (FUV − 3.6µm)corr
color as a function of the perpendicular (i.e. shortest) distance from each data-point to the
starburst relation, dstarb(see Kong et al. 2004). Although with considerable dispersion, the
trend is rather evident: regions with intrinsically redder (FUV − 3.6µm)corrcolors −having
then lower current-to-past star formation activity− are clearly located further away from
the relation for starbursts. Interestingly, selection effects can blur or even make this trend
vanish. For instance, if we only consider regions bluer than (FUV − NUV) ∼ 0.7, no clear
correlation between SFH and dstarbcan be inferred. Therefore, one needs to explore a wide
range in SFHs in order to see this trend. This could imply that although SFH seems to be
driving this departure from the starburst relation, other factors such as dust geometry or the
extinction law might be also contributing in different ways. It should be noted as well that if
we choose another IRX-β curve as a reference instead of the starburst one (i.e. MW-type dust
– 14 –
with different geometries, for instance), all perperincular distances will change accordingly.
This could explain the different conclusions reached by authors studying integrated
properties of galaxies. Kong et al. (2004) found a correlation between dstarband different
indicators of the SFH, like the Dn(4000) break and the Hα equivalent width. Cortese et al.
(2006) estimated the birthrate parameter from Hα and H-band luminosities for an optically
selected sample of normal star-forming galaxies, and found a weak correlation between b
and dstarb, with considerable scatter. The fact that their galaxies belong to nearby clusters
might be in part responsible for the observed scatter. Interactions with the the intra-cluster
medium and the cluster potential well may likely affect their SFHs by removing gas from the
disks and quenching their SF activity, thus progressively turning them into anemic spirals
(see e.g. Boselli & Gavazzi 2006 and references therein). In their analysis of the integrated
properties of the SINGS galaxies, Dale et al. (2007) found that most of the scatter towards
redder UV colors in the IRX-β diagram was due to ellipticals and early-type spirals. Panuzzo
et al. (2007) studied a UV selected sample of galaxies, and did not find any systematic
deviation from the starburst relation that depended on b, computed from NUV and H-
band luminosities. However, their UV selected sample did not contain objects redder than
(FUV−NUV) ∼ 0.7, thus making it difficult to infer any correlation, as Fig. 4 demonstrates.
In order to further illustrate these issues, in Fig. 3b we have highlighted some tracks for
particular galaxies, connecting the data-points of annular regions at different galactocentric
distances within each particular galaxy. The innermost point of each profile is marked for
clarity, and the integrated colors of each galaxy are shown with open stars. ‘Normal’ spirals
follow their own IRX-β relation, which is quite similar in shape to the average one (modulo
global offsets in the overall UV color and/or extinction). In early type spirals with well
defined bulges, such as NGC 3031 (M 81), the track clearly deviates towards redder colors
as we move closer to the center, due to the contribution of more evolved stars in the bulge.
This does not happen in late-type galaxies, like NGC 2403, given their smaller bulge-to-disk
ratios, especially at UV wavelengths.
The SINGS sample also includes some peculiar objects that do not follow these smooth
trends. This is the case of NGC 4826, which is a clear example of an anemic spiral (van
den Bergh 1976). Although it has a global radial extent of ∼ 13kpc at 3.6µm, the bulk of
the star formation activity −as traced by the FUV or 24µm images, for instance− seems to
be restricted to the central 5kpc. Therefore, as we move away from the center towards the
outer and more quiescent regions, the track followed by this galaxy in the IRX-β diagram
heads towards the region populated by bulges. Something similar happens with NGC 4569,
another anemic galaxy in the Virgo cluster, where ram pressure stripping seems to have
quenched the star formation in the outer regions (Boselli et al. 2006).
– 15 – Download full-text
Systems hosting starburst activity, like NGC 4536, lie close to the empirical relation
given by Kong et al. (2004). In very edge-on systems like NGC 4631 (which also happens
to host starburst activity), geometry might also play an important role. The observed TIR-
to-FUV ratio is probably larger than the one we would measure if the galaxy was face-on
(besides the fact that this galaxy is probably more dust-rich than the average spiral galaxy).
However, we do not find any significant trend between the position in the IRX-β diagram
and inclination, apart from this extreme case. At fixed metallicity, however, inclination does
seem to play a role (see Section 3.3).
Kong et al. (2004) suggest that galaxies in their sample departing too much from the
starburst relation require very quiescent SFHs, with very short time-scales of star-formation,
not typical of spirals. They invoke an additional mechanism that can contribute to broaden-
ing the IRX-β relation, consisting of an extra burst of star formation at some point during
the galaxy’s lifetime, superimposed on top of an otherwise smooth SFH. While this can
certainly add more scatter, we note that in our radial analysis most regions lying at large
distances from the starburst relation are either bulges or regions of anemic spirals with clear
signs of star-formation quenching. Indeed, there are 18 galaxies in our sample having points
with (FUV−3.6µm)corr> 3. Five of them are ellipticals or lenticulars, and 11 are early-type
spirals (S0/a-Sb), including the already mentioned anemic ones. Therefore, at least for these
very quiescent systems, a rapidly declining SFH might be a more reasonable possibility.
From the above analysis we can conclude that the star-formation history is possibly
driving the departure of star-forming regions from the locus of starburst galaxies in the
IRX-β diagram. However, the trends with different tracers of SFH (both here and in other
studies) seem to be quite noisy and not always evident −they can actually disappear if the
range of explored SFHs is not wide enough. Several reasons might explain this. First of
all, observational errors might blur the offset between regions if their SFHs are not different
enough.Secondly, realistic SFHs cannot be parametrized with a single quantity like b;
indeed, the intermediate bursts proposed by Kong et al. (2004) were shown to contribute to
the observed scatter. Besides, a number of additional factors such as the relative geometry
of dust and stars, the shape of the internal extinction law and the IMF can also have a great
impact (Burgarella et al. 2005, Panuzzo et al. 2007).
3.3. Influence of metallicity and inclination
Attenuation and metallicity are known to be correlated in starburst galaxies (Calzetti
et al. 1994; Heckman et al. 1998), which can be interpreted in terms of increasing extinction
at larger dust-to-gas ratios (see the discussion and references given in Section 4.3.6). Such a