1. UV and FIR properties of galaxies from combined GALEX-IRAS data
Jorge Iglesias Páramo
V. Buat
T. Takeuchi
K. Xu
& the GALEX team
This wok has been performed mainly in collaboration with V. Buat, Tsutomu
Takeuchi, Kevin Xu and the GALEX team.

2. Motivation
To understand the differential properties and observational biases of UV and FIR selections.
To consistently calibrate physical quantities estimated from the samples.
To provide observables for models of statistical properties of galaxies.
The motivations for this work are first to understand the differential properties and the
Observational biases of FIR and UV selections, since as we will se later on some of
The observed properties of galaxies are strongly affected by the selection criterion.
We intend also to calibrate physical quantities, like the dust attenuation and the star
formation rate, which span different ranges of values for each sample.
Finally, to provide observational quantities that will put constraints to models of
Statistical properties of galaxies.

3. The data NUV selected sample: from GALEX AIS (615 deg2).
mNUV < 16 ABmag
FIR counterparts from IRAS FSC (SCANPI)
Detection at 60µm of all galaxies with ANUV > 0.3 mag
94 galaxies
FIR selected sample: from IRAS PSCz (509 deg2).
f60 > 0.6 Jy
UV counterparts from GALEX AIS
Detection at NUV of all galaxies with ANUV < 4.4 mag
163 galaxies
Two sets of data have been used to carry out this work, selected from the same region
of the sky on the basis of their NUV and FIR fluxes.
The NUV selected sample, from 615 squared degrees of the GALEX AIS, containing
all galaxies brighter than NUV=16ABmag. The FIR counterparts were extracted from
the IRAS FSC or SCANPI. With these constraints we can detected at 60microns all the
galaxies with A_NUV > 0.3mag. The resulting catalog contains 94 galaxies.
The FIR selected sample was built from the region of the sky of the PSCz catalog
coincident with the GALEX AIS survey. The limiting constraint being f_60>0.6Jy.
The UV counterparts were extracted from the GALEX AIS. With these constraints we
have detections at both wavelengths for all galaxies with A_NUV<4.4 mag. The final
catalog contains 163 galaxies.
The limiting fluxes of both samples were imposed in order to obtain detections at both
wavelengths for almost all the galaxies and/or upper limits highly significant.

4. FNUV vs. F60µm log F60 (erg s-1 cm-2) log FNUV (erg s-1 cm-2) mNUV=16
IRAS PSCz
IRAS FSC
GALEX AIS
mNUV=16
log F60 (erg s-1 cm-2)
This figure shows the F_NUV vs. the F_60. Blue and red symbols correspond to NUV
and FIR selected galaxies respectively. The limiting fluxes of both samples at both
wavelengths are indicated by straight lines. The few upper limits are indicated with
arrows. It can be seen that besides the fact that both samples were extracted from the
same regions of the sky, most of the galaxies belong to only one of the samples.
log FNUV (erg s-1 cm-2)

5. The FIR selected sample is drawn from a larger volume than the NUV selected one.
The figure shows the histograms of the radial velocities for both samples. The mean
radial velocity is almost double for the FIR selected galaxies than for the NUV
selected ones. This large difference is due to the different selection criterion imposed
to each of the samples.
Vel. (km s-1)

6. Representativity of the Samples
Takeuchi et al. (2003)
Wyder et al. (2005)
An important concen in order for the validity of the results derived from these samples
is the representativity with respect to the galaxtic populations in the local Universe. As
it is shown the 60micron luminosity function constructed from the FIR selected sample
is consistent with the one derived by Takeuchi et al. (2003) for a large sample of
galaxies from the PSCz. A similar result is obtained for the NUV luminosity function
of the NUV selected sample when compared to the local Universe LF from Wyder et
Al. (2005) derived from GALEX data.

7. UV and FIR luminosities
log LNUV (Lsun)
The plot shows L_NUV vs. L_60 for both samples. The first interesting point is that
there is a tight relation between both luminosities for the NUV selected galaxies over
several orders of magnitude in luminosity. This trend is not seen for the FIR selected
galaxies which show a very dispersed relation. L_NUV accounts for a non negligible
fraction of the total luminosity for NUV selected galaxies. For these galaxies, L_NUV
>L_60 up to L ~ 10^10 and the opposite for more luminous galaxies. The FIR selected
galaxies show a large range of L_60 for a given L_NUV and are in general more
Luminous at UV than at FIR.
log L60 (Lsun)

8. In order to improve the quality of the FIR/UV flux association some objets of both samples were discarded...
NUV
IRAS 60µ
IRAS 60µ
NUV
NVSS
NVSS
Once we have seen the representativity of our samples and the effect of the selection
criteria on the luminosities of the selected galaxies we perform a study on some
properties of our galaxies based on the combination of the FIR and UV fluxes. With
the purpose of minimizing the observational uncertainties in our FIR and NUV fluxes
we impose some extra criteria to our galaxies:
Multiple galaxies resolved in UV but not resolved in FIR are excluded for this
subsequent analysis, as is the case of the one in the figure.
Objects for which the FIR emission is not clearly localized in the NUV frames are also
excluded as the one shown in this figure.

9. NUV selected subsample: 62 galaxies
Also discarded for the FIR/UV analysis:
Ellipticals and S0s
AGNs
Galaxies for which Cirrus > 2
Finally we end up with:
NUV selected subsample: 62 galaxies
FIR selected subsample: 118 galaxies
We also discard galaxies for which the UV or FIR does not come undoubtedly from
recent star formation like ellipticals, S0s and active galaxies.
Also galaxies for which the associated cirrus as defined in the PSCz and FSC is larger
than 2 are excluded since the FIR photometry for these objects could have rather low
quality.
At the end we end up with two subsamples of galaxies composed by 62 and 118
galaxies.

10. ANUV directly from FFIR/FNUV (Buat et al. 2005)
Dust attenuation
ANUV directly from FFIR/FNUV (Buat et al. 2005)
〈ANUV〉 ~ 0.8 mag
〈ANUV〉 ~ 2.1 mag
We begin with the dust attenuation at NUV, which is estimated directly from the
FIR/UV ratio.
The figure shows the distribution of the attenuation for both samples. The average
values are 0.8 and 2.1 respectively. Not only the NUV selected galaxies are less
attenuated but also the dispersion in the attenuation is much lower for NUV than for
FIR selected galaxies. This is expected and results directly from the selection
criteria.
ANUV (mag)

11. ANUV correlated with LH for NUV selected galaxies.
ANUV (mag)
No such relation is observed for the FIR selected galaxies.
The attenuation is related to the galaxy mass, traced by the H-band luminosity for the
NUV selected galaxies.
However, this does not happen for the FIR selected galaxies where the dispersion is
very large.
This result holds also if we use FIR or NUV luminosities. This is a consequence of the
fact that the extinction for NUV selected galaxies is almost constant.
log LH (Lsun)

12. Star Formation Rates Comparison between SFRNUV and SFRdust Scenario:
Constant SFR over the last 108 yr
Salpeter IMF with 0.1M⊙ < M✮ < 100M⊙
Solar metallicity
We move now to the estimation of the SFRs for both samples. Both the NUV and the
FIR fluxes are related to the number of young stars, the NUV flux emitted directly by
The young stars and the FIR flux which is related to the dust emission which is the
result of the reprocess of the stellar light of youg stars.
In order to relate the NUV and FIR fluxes to the recent SFR we assume the following
scenario: constant SF in the last 10^8 years, a Salpeter IMF from 0.1 to 100 M_sun and
solar metallicity. Under these assumptions we derive the following formulae using
SB99. We note that L_NUV must be extinction corrected.
We stress the fact that our SFRs are averaged over the last 10^8 yr.
From Starburst99:
log SFRNUV (M⊙ yr-1) = log LNUV,corr (L⊙) – 9.33
log SFRdust (M⊙ yr-1) = log Ldust (L⊙)

13. Quite good agreement on average but...
SFRNUV vs. SFRdust
Quite good agreement on average but...
log SFRdust (Msun yr-1)
In the figure we can see SFR_NUV vs. SFR_dust for both samples of galaxies. It can
be seen that the agreement is good along several orders of magnitude, however two
trends can be seen.
log SFRNUV (Msun yr-1)

14. Two different trends are observed:
At low values of ANUV, the dust emission underestimates the total SFR because of the non negligible NUV emission.
log SFRNUV/SFRdust
At high values of ANUV, the NUV emission underestimates the total SFR.
Problem with ANUV?
At low SFRs, SFR_dust underestimates the total SFR. Low SFR galaxies correspond to
those of low attenuation. We saw that for these galaxies, the L_NUV, and consequently
SFR_NUV, is larger than L_60 and it cannot be neglected, so the dust emission does
Not trace the SFR.
At high SFRs, we see the opposite trend: the total SFR is overestimated by SFR_dust.
SFR_dust should be a proper estimator of the total SFR for galaxies with very large
SFR since strong starbursts are known to be very attenuated. In fact we checked that
the discordant galaxies are the most attenuated of our samples. So we argue that
the dust attenuation for strong starbursts should probably to be treated in a different
way as for more quiescent galaxies, and thus the SFR_NUV must be corrected.
We will come back later to this point.
log SFRNUV (Msun yr-1)

15. Star Formation Activity: F60µm/F100µm
Now we move to the estimation of the recent star formation activity of our galaxies
through different estimators.
The first one is the ratio f60/f100 which is directly related to the temperature of the
dust which is taken to be higher for active star forming galaxies.
We show the distributions of this parameter for our samples and for comparison
for the ISO Key Project sample of Dale et al. (2001) and the starbursts sample of Wu
et al. (2002). The median values are indicated with vertical lines. It can be seen that
our samples present lower median values of f60/f100, even showing tails of values of
f60/100 corresponding to quiescent galaxies.
log F60/F100

16. Birthrate parameter Following Boselli et al. (2001): b = SFR0/〈SFR〉
b ∝ SFRNUV/LH
b = SFR0/〈SFR〉
We obtain values typical of Sb - Sbc galaxies (Kennicutt et al. 1994).
Finally we use the birthrate parameter as an indicator of recent star formation activity,
which is defined as the recent SFR divided by the average SFR over the lifetime of the
galaxy. We estimate b from the SFR_NUV, proportional to the recent SFR, and L_H,
Proportional to the dynamical mass within the 25mag arcsec^-2 isophote.
We stress that given that, our b
is averaged over the last 10^8 yr and is not exactly instanteneous.
The values we obtain for our galaxies are similar for both samples and correspond to
Sb-Sbc galaxies.
log b

17. Fdust/FFUV vs. β
To the left of the starburst sequence of Meurer et al. (1995).
Starbursts
log Fdust/FFUV
The distance to the sequence is not related to the birthrate parameter.
Here we show the IRX vs. Beta diagram for the NUV selected galaxies. The sequence
of the starbursts of Meurer et al. (1995) is indicated by the dashed line. We see that
most galaxies are located below this sequence. The size of the symbol increases with
The birthrate parameter. No trend between b and the distance to the starbursts sequence
is seen. We stress however that our estimation of b takes into account the average SF
in the last 10^8 yr and it is not instantaneous.
beta

18. Fdust/FFUV vs. β
Some of them follow the same trend as the NUV selected.
Starbursts
The rest, to the left of the starburst sequence, below the ULIRGS of (Goldader et al ).
log Fdust/FFUV
Now we show the same figure for the FIR selected galaxies. Some of them follow the
Same trend as the NUV selected galaxies, although spanning a larger range in beta.
The remaining ones are located to the left of the starbursts sequence, slightly below the
locus occupied by the ULIRGs of Goldader et al. (2002). In fact none of these galaxies
are ULIRGs.
beta

19. Lower than local starbursts for both samples.
SFR per unit area
Lower than local starbursts for both samples.
Meurer et al. (1999)
log SFRNUV/Area (Msun yr-1 kpc-2)
Another indicator of star formation activity is the density of star formation per unit
area, or the intensity of star formation. We show in this figure the density of star
Formation within the effective area, and for comparison the starbursts of Meurer et al.
(1999). Our galaxies show a trend between SFR and SFR/Area, whereas the starbursts
show a flat relation. Our galaxies do not reach the values spanned by the starbursts.
We can see also that the FIR selected galaxies show higher values than the NUV
selected ones.
log SFRNUV (Msun yr-1)

20. log SFRNUV/Area (Msun yr-1 kpc-2)
The SFRNUV vs. SFRdust discrepancy is mainly related to the star formation activity.
log SFRNUV/SFRdust
We come back now to the discrepancy between the SFR_NUV and SFR_dust. As seen
In the Figure, those galaxies for which the SFR_NUV underestimates the total SFR are
also the most actives, as indicated by the intensity of star formation .
This confirms that we cannot use the same correction for dust attenuation for quiescent
and for starburst galaxies.
log SFRNUV/Area (Msun yr-1 kpc-2)

21. Starburst
Normal SF
In fact this makes sense because our correction for dust attenuation was averaged over several models with SFHs similar to those of normal SF galaxies and starbursts.
In fact the derived correction for starbursts is approximately 0.5 mag higher than for
normal star forming galaxies, for a given F_dust/F_NUV ratio.

22. UV Structural Properties
Concentration index: 5 × log r80/r20
We have also studied the UV morphology of galaxies. For this we used the
concentration index, defined this way.
The figure shows the distributions of the concentration index for the galaxies of both
samples. We have indicated the typical values for exponential, de Vaucouleurs and
flat profiles.
The median values of both samples are between the flat profile and the exponential
profile.
Concentration index

23. No obvious correlation between concentration index and morphological type.
There is no clear correlation between the concentration index and the morphological
type. In fact, the UV morphology of galaxies follows different morphological patterns
Without any relation with properties such as SFR, SF activity, mass or so...

24. Different morphological patterns...
Here I show some examples of the different morphological patterns which are
representative of our galaxies.
NGC0010, with a concentration index of 1.58, and showing a typical flat profile.
MRK0544, concentration index and profile typical of non resolved objects.
NGC0723, exponential disk profile.
IC1516, profile consistent with a de Vaucouleurs profile.

25. UV Bright Galaxies
We find 8 galaxies with LFUV ≥ 2 × 1010 L⊙, all of them FIR selected and 3 also NUV selected.
Redshift range: z < 0.10
Co-moving spatial density: ~ 5.3 ± 3.1 × 10-6 Mpc-3
Dust attenuation: 1.3 < AFUV < 3.5 mag
Finally, I give the properties of the UV bright galaxies found in our samples, as these
galaxies at low redshift are supposed to be the low z counterparts of LBGs.
8 such galaxies were found in our samples, all of them in the FIRsel and 3 also in the
NUVsel.
All of them are located at z<0.1
The co-moving spacial density was found to be lower than the estimated by Heckman
et al. (2005) for a larger sample of galaxies.
The dust attenuation was estimated to be between 1.3 and 3.5.
They show a low surface brightness compared to other samples of UV bright galaxies.
The most important point is that they are really luminous from a bolometric point of
view since for all of them L_60 > L_NUV.
Low surface brightness: ∑FUV < 1.32 × 108 L⊙ kpc-2
For all of them: L60µm > LFUV ⇒ Very luminous objects!