Star formation and dust extinction properties of local galaxies from AKARI-GALEX All-Sky Surveys: First results from most secure multiband sample from FUV to FIR
T. T. Takeuchi, V. Buat, S. Heinis, E. Giovannoli, F.-T. Yuan, J. Iglesias-Paramo, K. L. Murata, D. Burgarella
aa r X i v : . [ a s t r o - ph . C O ] D ec Astronomy&Astrophysicsmanuscript no. tttakeuchi˙accepted c (cid:13)
ESO 2018September 4, 2018
Star formation and dust extinction properties of local galaxies fromAKARI-GALEX All-Sky Surveys:
First results from most secure multiband sample from FUV to FIR
T. T. Takeuchi , V. Buat , S. Heinis , E. Giovannoli , F.-T. Yuan , J. Iglesias-P´aramo , K. L. Murata , and D.Burgarella Institute for Advanced Research, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464–8601, JAPANe-mail: [email protected] Laboratoire d’Astrophysique de Marseille, OAMP, Universit´e Aix-Marseille, CNRS, 38 rue Fr´ed´eric Joliot-Curie, 13388 Marseillecedex 13, FRANCEe-mail: veronique.buat, elodie.giovannoli, [email protected] Department of Physics & Astronomy, The Johns Hopkins University, 3701 San Martin Drive, Baltimore, MD 21218, USAe-mail: [email protected] Division of Particle and Astrophysical Sciences, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464–8602, JAPANe-mail: yuan.fangting, [email protected] Instituto de Astrof´ısica de Andaluc´ıa (IAA, CSIC), Camino Bajo de Hu´etor 50, 18008 Granada, SPAINe-mail: [email protected]
ABSTRACT
Aims.
We explore spectral energy distributions (SEDs), star formation, and dust extinction properties of galaxies in the Local Universe.
Methods.
The AKARI All-Sky Survey provided the first bright point source catalog detected at 90 µ m. Starting from this catalog, weselected galaxies by matching AKARI sources with those in the IRAS PSC z . Next, we have measured total GALEX FUV and NUVflux densities by a photometry software we have specifically developed for this purpose. Then, we have matched this sample withSDSS and 2MASS galaxies. By this procedure, we obtained the basic sample which consists of 776 galaxies. After removing objectswith photometry contaminated by foreground sources (mainly in SDSS), we have defined the “secure sample” which contains 607galaxies. Using this galaxy sample, we have explored various properties of galaxies related to star formation and dust extinction. Results.
The sample galaxies have redshifts < ∼ .
15, and their 90- µ m luminosities range from 10 to 10 L ⊙ , with a peak at 10 L ⊙ .The SEDs display a large variety, especially more than four orders of magnitude at M-FIR , but if we sort the sample with respect to90 µ m, their average SED shows a coherent trend: the more luminous at 90 µ m, the redder the global SED becomes. The M r -NUV − r color-magnitude relation of our sample does not show bimodality, and the distribution is centered on the green valley between theblue cloud and red sequence seen in optical surveys. We have established formulae to convert FIR luminosity from AKARI bands tothe total infrared (IR) luminosity L TIR . With these formulae, we calculated the star formation directly visible with FUV and hiddenby dust. The luminosity related to star formation activity ( L SF ) is dominated by L TIR even if we take into account the far-infrared(FIR) emission from dust heated by old stars. At high star formation rate (SFR) ( >
20 M ⊙ yr − ), the fraction of directly visible SFR,SFR FUV , decreases. We also estimated the FUV attenuation A FUV from FUV-to-total IR (TIR) luminosity ratio. We also examined the L TIR / L FUV -UV slope (FUV − NUV) relation. The majority of the sample has L TIR / L FUV ratios 5 to 10 times lower than expected fromthe local starburst relation, while some LIRGs and all the ULIRGs of this sample have higher L TIR / L FUV ratios. We found that theattenuation indicator L TIR / L FUV is correlated to the stellar mass of galaxies, M ∗ , but there is no correlation with specific SFR (SSFR),SFR / M ∗ , and dust attenuation L TIR / L FUV . Conclusions.
Together, these results show that the AKARI FIS All-Sky Survey gives a representative sample of SF galaxies in theLocal Universe. This sample will be a comprehensive standard of various properties of SF galaxies to be compared with, e.g., high- z SF galaxies.
Key words. galaxies: evolution-galaxies: stellar content-infrared: galaxies-ultraviolet: galaxies
1. Introduction
Star formation history of galaxies is one of the most importantand exciting topics in extragalactic astrophysics and observa-tional cosmology. Especially, exploring the “true” absolute valueof the cosmic star formation rate (hereafter SFR) has been of acentral importance.However, it remained di ffi cult for a long time because ofdust extinction. Even in the Local Universe, there aresome prob- Send o ff print requests to : T. T. Takeuchi lems to estimate the absolute value of SFR density because ofdi ff erent dependence of various SFR estimators on dust atten-uation (e.g., Hopkins & Beacom, 2006, and references therein).Active star formation (SF) always comes along with dust pro-duction, because of various dust grain formation processes re-lated to the final stage of stellar evolution (e.g., Dwek & Scalo,1980; Dwek, 1998; Nozawa et al., 2003; Takeuchi et al., 2005c).Observationally, SFR of galaxies is measured by the ultravio-let (UV) luminosity from massive stars because of their shortlifetime ( ∼ yr) compared with the age of galaxies or theUniverse. However, the UV photons are easily scattered and ab- T. T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX sorbed by dust grains. Hence the SFR of galaxies is always in-evitably a ff ected by dust which is produced by their SF activ-ity. Since the absorbed energy is re-emitted at far-infrared (FIR)wavelengths, it is essential to observe galaxies both at UV andFIR to have an unbiased view of their SF (e.g., Buat et al., 2005;Seibert et al., 2005; Cortese et al., 2006; Takeuchi et al., 2005a).To know the history of SFR in the Universe, we must knowthe starting reference value, i.e., the SFR density in the LocalUniverse: otherwise, we cannot define how much larger the SFRwas in the past. However, since the volume of the Local Universeis limited by definition, an all-sky survey is the only viable wayto refine our knowledge of local galaxies. On the “directly visibleSF” side, the advent of the UV surveyor-type satellite GALEX(Martin et al., 2005) has changed the situation of UV astronomydrastically. GALEX is performing an all-sky survey (All SkyImaging Survey: AIS) at FUV (1530 Å) and NUV (2300 Å) withdetection limits of 19.9 and 20.8 mag (Morrissey et al., 2007),as well as deep surveys in some selected regions. In a previousstudy, we have shown that GALEX AIS provides us with an un-precedented opportunity to explore the visible SF in the LocalUniverse (Buat et al., 2007a).As for the “hidden” side of SF, the Infrared AstronomicalSatellite (IRAS; Neugebauer et al., 1984) has brought a vastamount of statistics of dusty galaxies in the Local Universeby IRAS Point Source Catalog (PSC) (see, e.g., Soifer et al.,1987). Subsequently, FIR facilities with much higher sensitivityhave been launched, like ISO (e.g., Genzel & Cesarsky, 2000;Verma et al., 2005) and Spitzer (e.g., Soifer et al., 2008), and re-vealed the deep IR universe, but the latter two were observatoriesdedicated to pointed observations.In contrast to the latter two facilities, the Japanese IR satelliteAKARI has performed various large-area surveys at IR wave-lengths (Murakami et al., 2007) after IRAS, especially includingall-sky surveys at FIR and MIR . In particular, with the aid of theFar-Infrared Surveyor (FIS: Kawada et al., 2007) onboard, var-ious IR surveys were performed by AKARI. AKARI FIS hasfour FIR wavebands centered on 65 , , µ m, and aFIR all-sky survey was completed by this instrument. Since thelatter two bands are longer than 100 µ m which was the longestwavelength band of IRAS, the obtained sample of dusty galax-ies is less biased than the IRAS sample, i.e., thanks to the bettersensitivity to cooler dust than IRAS, AKARI can detect galax-ies with dust emission with lower temperatures. Thus, AKARIis a very promising facility to bring new knowledge of galaxieswith cold dust. In addition to these SF-related wavelengths, weneed other various wavelength bands to examine physical prop-erties of galaxies, e.g., stellar mass [closely related to near-IR(NIR)], and intermediate stellar population (related to optical).For the former, we have a set of all-sky data from 2-Micron All-Sky Survey (2MASS: Skrutskie et al., 2006), and for the latter,the SDSS final data (DR7) are publicly available , even if SDSSis not an all-sky survey.In this work, we constructed a multiband galaxy catalogbased on AKARI All-Sky Survey 90- µ m selected sources asso-ciated with IRAS PSC z galaxies (Saunders et al., 2000). Then,we measured GALEX FUV and NUV flux densities, and asso-ciated SDSS and 2MASS photometries. For this initial study,we have only selected “secure” galaxies with good photometricmeasurements for most of the bands. We present the sample con-struction in Section 2. We describe basic properties of galaxiesin the sample in Section 3. Results on the SF and dust attenuation URL: . Fig. 1.
Comparison between AKARI FIS and IRAS PSC z fluxdensities. Upper-left, upper-right, and lower-left panels presentcomparisons of IRAS 60 µ m with AKARI 90 µ m, 65 µ m, and140 µ m flux densities of the AKARI-IRAS correlated sample.The vertical dotted lines in these panels represent the flux densitylimit of IRAS PSC z . Lower-right panel shows a comparison ofIRAS 100 µ m with AKARI 90 µ m flux densities. The horizontaldotted line in Upper-left panel represents the formally expecteddetection limit of FIS 90 µ m.are presented in Section 4. Section 5 is devoted to our summaryand conclusions.Throughout the paper we will assume Ω M0 = . Ω Λ = . H =
70 km s − Mpc − . The luminosities are defined as ν L ν and expressed in solar units assuming L ⊙ = . × erg s − .
2. Sample construction z The primary purpose of the AKARI mission is to providesecond-generation infrared (IR) catalogs with better spatial res-olution and wider spectral coverage than the IRAS catalog.AKARI is equipped with a cryogenically cooled telescopeof 68.5 cm aperture diameter and two scientific instruments,the Far-Infrared Surveyor (FIS; Kawada et al., 2007) and theInfrared Camera (IRC; Onaka et al., 2007). Among various as-tronomical observations performed by AKARI, as we have men-tioned in Introduction, an all sky survey with FIS has been car-ried out (AKARI All-Sky Survey). Since FIS is an instrumentdedicated to FIR λ = µ m, all the AKARI FIS bandsare in the FIR wavelengths: N60 (65 µ m), WIDE-S (90 µ m), WIDE-L (140 µ m), and N160 (160 µ m) (Kawada et al., 2007).Hereafter, we note as S , S , S and S the flux densitiesin these bands. Especially, since FIS has sensitivity at longer . T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX 3 wavelengths than IRAS, a new classification scheme is neededif we try to select a certain class of objects. Such a scheme isnot merely an empirical technique but also will provide us witha new understanding of objects with cool dust which were di ffi -cult to detect by IRAS bands.We use the AKARI FIS Bright Source Catalogue (BSC),the first primary catalog from the AKARI All-Sky Survey. Datafrom the β -1 version of this catalog are used in this work.AKARI BSC is supposed to have a uniform detection limit, cor-responding to per scan sensitivity, over the entire sky, except forvery bright sky parts where di ff erent data acquisition mode hadto be applied. A summary of the All-Sky Survey is presented inYamamura et al. (2009). The AKARI FIS BSC provides data for64311 sources. For each detected source, AKARI source identi-fier, equatorial coordinates of the source position and flux densi-ties in the four FIR bands are given. Errors are not estimated foreach individual source, but instead they are in total estimated tobe 35 %, 30 %, 60 %, and 60 % at N60 , WIDE-S , WIDE-L , and
N160 , respectively (Yamamura et al., 2008). AKARI IRC per-formed another all sky survey, but the data are not fully reducedat the time we have been preparing this paper. Hence we focuson the FIS data only. z The AKARI BSC contains many Galactic sources, like AGBs,H ii regions, planetary nebulae, etc. (e.g., Pollo et al., 2010). Inorder to construct a reliable catalog of galaxies, we should pickup FIS sources confirmed as galaxies. For this purpose, we haveperformed a cross identification of BSC sources with the IRASPSC z (Saunders et al., 2000). The PSC z is a redshift survey ofgalaxies selected at IRAS 60 µ m with a flux density limit of S > . . The PSC z contains ∼ v > − so that we canavoid the e ff ect of the peculiar velocity of galaxies. Then, wehave matched the AKARI BSC sources with PSC z galaxies witha search radius of 36 arcsec, which corresponds to the positionuncertainty of IRAS PSC. The number of matched sources was5890. To examine the e ff ect of the choice of search radius, wechanged the criterion from 20–60 arcsec. This change of radiusdoes not have a significant impact on the resulting catalog ( < ∼ .
015 deg .This restriction to this region of the sky has one advantage: theGalactic di ff use FIR emission is strong in some areas of the sky.In such regions, measured FIR flux densities of point sources arecontaminated by the Galactic emission and not very accurate.Since the SDSS region is selected so that the Galactic extinctionis small, the selected area automatically avoids such FIR-brightregions. Then, by this selection, our sample automatically ex-cludes strongly contaminated sources. The resulting parent cat-alog contains 1186 galaxies. We compared AKARI and IRAS flux densities to examine oursample selection. The correlation is shown in Figure 1. The hori- Because of this step, we should note that we do not make a maximaluse of the advantage of the long wavelength bands of AKARI FIS, sincethe sample is limited by IRAS bands ( λ < µ m). zontal dotted lines in upper-left, upper-right, and lower-left pan-els represent the flux density limit of IRAS PSC z . AKARI BSCsources are selected at WIDE-S , i.e., 90 µ m. It is important tosee which selection controls the sample selection. As seen in theupper-left panel, the IRAS PSC z limit and AKARI limit are bothwell-defined, and neither of them strongly restricts the sample.The e ff ective 90 µ m flux density limit of our parent sample is ∼ . GALEX AIS now observed 25000 deg at FUV and NUV. Thelatest version of the public imaging is GR4 / GR5 . We havemeasured the FUV and NUV photometry of the parent AKARIgalaxies as follows:1. Cut out a 30 ′ × ′ square subimage from GALEX AIS im-ages around each AKARI galaxy.2. Select a subimage with the largest exposure time when mul-tiple observations were available.3. Measure FUV and NUV flux densities. The NUV observa-tion is taken as our reference.Since the sky coverage of GALEX AIS is not complete, in somecases we do not have a proper GALEX image for an AKARIgalaxy. In such a case we omit the galaxy because we do nothave any UV information.Almost all of the sources are resolved by GALEX. Theyare thus very often separated into small bright patchy regions,and the GALEX pipeline misidentifies these fragments as indi-vidual objects. This is referred to as shredding . We must dealwith the shredding to obtain sensible flux density measurementsfor nearby extended galaxies. For this purpose, we have usedan IDL software package developed by ourselves. This softwareperforms aperture photometry in the NUV sub-image using aset of elliptical apertures. Total flux density is calculated withinthe aperture corresponding to the convergence of the growthcurve. The sky background is measured by combining severalindividual regions around the source. NUV and FUV flux den-sities are corrected for Galactic extinction using the Schlegelmap (Schlegel et al., 1998) and the Galactic extinction curve ofCardelli et al. (1989). A detailed description of the photometryprocess can be found in Iglesias-P´aramo et al. (2006).This software was already used for our previous IRAS-GALEX based studies (Buat et al., 2007a), and its performanceis carefully checked and established. During the procedure, wealso excluded all the sources contaminated by stars or too closeto be disentangled by the photometry software. By this proce-dure, the number of remaining galaxies is 776. We further matched the AKARI-IRAS PSC z -GALEX sample(776 galaxies, hereafter abbreviated as the AKARI-GALEXsample) with SDSS and 2MASS.The AKARI-GALEX sample was matched with the 2MASSAll-Sky Extended Source (XSC) catalog in order to obtain theNIR ( J , H , and Ks ) flux densities. A matching radius of 20 ′′ wasinitially adopted for the cross correlation. All but 7 out of the 776galaxies in the AKARI-GALEX sample showed 2MASS coun-terparts at distances closer than the matching radius. In case ofmultiple candidates, we always selected the brightest one at Ks band as the most plausible one. A further cross correlation with URL: http://galex.stsci.edu/GR4/
T. T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX
Fig. 2.
Number counts of our AKARI multiband sample atAKARI FIS four bands. Triangles, diamonds, squares, andcrosses represent AKARI
N60 , WIDE-S , WIDE-L , and
N160 galaxy counts, respectively.a matching radius of 30 ′′ was attempted for these seven galax-ies with no 2MASS counterpart within the 20 ′′ radius. However,again, no 2MASS counterparts were found, and we adopted thestandard 2MASS XSC upper limits for these galaxies at J , H ,and Ks bands. The standard 2MASS limiting magnitudes at J , H ,and Ks are 14.7, 13.9 and 13.1 mag, respectively (Jarrett et al.,2000).We also matched the AKARI-GALEX sample to SDSS us-ing the GALEX coordinates for the AKARI-GALEX objects,and a search radius of 15 ′′ . In the case of SDSS, large resolvedgalaxies such as those we are dealing with here are shredded inmultiple detections during the deblending step of the pipeline.We used the closest SDSS match to the AKARI object to obtainthe SDSS photometry of the parent object, namely the objectdetected by the SDSS pipeline before deblending. We inspectedall sources to check that the actual flux measured for the parentobject is not contaminated by nearby bright stars, artifacts etc.For most of the sample galaxies, SDSS galaxies were as-sociated. However, stars superposed on the SDSS galaxy im-ages often hamper accurate photometry. It requires a very care-ful masking of the SDSS image, and in the current analysis wesimply omitted such galaxies. After this selection, 607 galax-ies remained. We call this sample “the final sample”. Almost allgalaxies have very secure photometric data at UV, optical, NIR,MIR, and FIR, as well as the redshift information.
3. Basic Properties of the Sample Galaxies
Figure 2 presents the cumulative AKARI flux density distribu-tion (number counts) of our final sample. We see that the fluxdensity limit of the sample at 90 µ m is ∼ ∼ . Fig. 3.
Redshift distribution of the sample. The distribution isnormalized so that we obtain the number of galaxies per unitsolid angle if we integrate it over redshifts.
Fig. 4.
Luminosity distribution of the sample at 90 µ m. The redshift distribution of our final sample is shown in Figure 3.Since AKARI FIS BSC is rather shallow, most of the sam-ple locate at low redshifts z < ∼ .
05. This is quite consistentwith predictions of various IR galaxy evolution models (e.g.,Takeuchi et al., 2001a,b; Chary & Elbaz, 2001).Figure 4 presents the 90- µ m luminosity distribution of thesample. Since this is a raw luminosity distribution of IR galaxies,it decreases toward the faint end. The 90- µ m luminosity rangesfrom 10 L ⊙ –10 L ⊙ . Only a few galaxies are classified as ul-traluminous IR galaxies (ULIRGs). The peak of the luminositydistribution is around 10 L ⊙ , which is more luminous than theknee of the 60- µ m luminosity function (Takeuchi et al., 2003a).Since our selection procedure is multifold and complicated, esti-mating a reliable luminosity function is not straightforward. Wewill try this task in future works. . T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX 5 Fig. 5.
Luminosity distribution of the sample at several wave-lengths. Triangles, diamonds, squares, and crosses representgalaxies AKARI
WIDE-S , GALEX FUV, SDSS r , and 2MASS K -band, respectively.We can compare the distribution of galaxies on the redshift–luminosity plane to see the e ff ect of the shape of luminosityfunction. We show the z - ν L ν relation in Figure 5. Some rep-resentative wavelengths are shown: AKARI WIDE-S , GALEXFUV, SDSS r , and 2MASS Ks . We see that the luminosity at90 µ m increases monotonically toward higher redshifts, whilethe optical and near-IR (NIR) luminosities of the same galax-ies saturate at certain values. This is related to the di ff erence inthe shape of their luminosity functions (see, e.g., Takeuchi et al.,2005a; Iglesias-P´aramo et al., 2006); i.e., the luminous end ofthe function exponentially declines at optical-NIR and UV, whileit shows a power-law decline at FIR. This is also related to thefact that the more luminous galaxies are more strongly extin-guished by dust. We will revisit this issue in a much more directway in Section 4.4 Since we have the monochromatic luminosities from FUV toFIR, we construct the spectral energy distributions (SEDs) of oursample. We show all the SEDs of the sample in Figure 6. Largeempty squares represent the AKARI FIS measurements, whilethe small squares represent all the other data in Figure 6. We ob-serve a very large variety of SEDs among the sample galaxies.To see a global trend of the SEDs more clearly, we sortedthe sample by their 90- µ m luminosities and subdivided the sam-ple into six logarithmic bins from 10 L ⊙ to 10 L ⊙ (with abin width ∆ log L = µ m luminosity,shown in Figure 7. Because of a large dispersion and asymmet-ric distribution of the SEDs in each bin, sometimes the averageand median SEDs do not agree very well. Even so, now the trendis more clearly seen: low 90- µ m luminosity galaxies have coolerdust emission and bluer UV-optical continuum, while high 90- µ m luminosity galaxies have hotter dust emission and redderUV-optical continuum.The 90- µ m luminosity dependence of the dust emission tem-perature is more clearly seen if we plot a flux density ratio Fig. 6.
Spectral energy distributions (SEDs) of the whole sample.Large empty squares represent the AKARI measurements, whilethe small squares are data taken from GALEX FUV, GALEXNUV, SDSS u , g , r , i , z , 2MASS J , H , Ks , and IRAS 12, 25, 60,and 100 µ m from left to right. Dotted lines connect data pointsof each object to guide the eye. S / S as a function of L . This is shown in Figure 8. Again,the agreement between average and median values is not excel-lent, we see a clear monotonically increasing trend of the fluxdensity ratio along with L .The lowest luminosity galaxies have the largest uncertaintymainly because of poor statistics. This problem will be solved byconstructing deeper and larger sample, possibly by the next gen-eration AKARI FIS catalog. The most 90- µ m luminous galaxiesseem to have an upturn at UV. This may be because of an AGNcomponent in these galaxies. We will examine this issue in ourfuture work. In both Figures 6 and 7, it is clear that AKARI FISmeasurements at N60 and
WIDE-S do not agree with IRAS ones.This discrepancy is mostly due to the better angular resolutionof AKARI FIS compared with IRAS. Thus, the measured valuesare much less contaminated by Galactic cirrus emission. Alsobecause of better angular resolution, the source confusion e ff ectis much smaller than for IRAS. As both cirrus and source confu-sion e ff ects cause flux boosting for IR galaxy number counts, theIRAS flux densities could be overestimated (Takeuchi & Ishii,2004). Jeong et al. (2007) also examined the same problem withan early AKARI sample and concluded that the di ff erence be-tween IRAS and AKARI flux densities are due to the confusione ff ects. We see a strange bump at 12 µ m for luminous galaxies.This may not be a physical feature in the SEDs of our samplebut because of a poor measurement at this band by IRAS. Inthe redshift–luminosity diagram at 12 µ m (see Fig. A.1), we seethat many galaxies locate on the limiting luminosity line. Thismeans that, even if they are classified as measurements, actuallythey are upper limits of the flux density. We plan to study thispoint further by AKARI IRC all-sky survey in the future. T. T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX
Fig. 7.
Averaged spectral energy distributions of our sample asa function of 90 µ m luminosity. Luminosity bins used here arethe same as the bins in Figure 4. The central luminosities of thebins are 10 , 10 , 10 , 10 , 10 , and 10 L ⊙ , respectively, andthe obtained average SEDs are represented with di ff erent sizeof the symbols: from the smallest corresponding to 10 L ⊙ tothe largest corresponding to 10 L ⊙ . Filled and empty squaresare averaged SEDs, while filled and empty triangles are medianSEDs. Fig. 8.
Dependence of the flux density ratio S / S on aver-age 90- µ m luminosity. Same as Fig. 7, the squares are estimatedfrom averaged SEDs, while triangles are from median SEDs. − r distribution The NUV − r restframe color is very e ffi cient to separatethe galaxies into blue and red populations (Salim et al., 2007;Martin et al., 2007). At high- z , IR selected galaxies observedwith Spitzer were found to populate the so-called green valley Fig. 9.
The absolute r -magnitude-NUV − r color distribution.Upper and lower thick solid lines represent the average red se-quence and blue cloud taken from Salim et al. (2007), with itsenvelope indicated by dotted lines.between the blue cloud and red sequence, first mentioned byBell et al. (2005) and further identified in the COSMOS fields(Kartaltepe et al., 2009; Vergani et al., 2009).However, in contrast, no systematic study of IR selectedgalaxies has been performed in the nearby universe until now.Figure 9 shows the distribution of the NUV − r color for oursample galaxies against the absolute magnitude at r -band, M r .The distribution in NUV − r is found to be large and mono-lithic in contrast to the bimodal distribution found in optical sur-veys. From Figure 1 of Salim et al. (2007) (based on an SDSS-GALEX selected sample), we approximate the red sequence onthis diagram with a linear relationNUV − r = − . M r + . , (1)with ± . − r = − . M r − . , (2)with ± .
7. Clearly, we see that our galaxies lie in the bluecloud and populate the green valley to produce a continuousdistribution, and only a few objects are located in the red se-quence. Therefore, the color distribution of FIR-selected galax-ies is di ff erent than the one obtained from an optical selection,as they preferentially populate the green valley between the redsequence and blue cloud. An interpretation of this behavior interms of dust attenuation, SF history or AGN activity will bedeveloped in a future paper.
4. Star Formation and Dust Extinction of the µ m-Selected Galaxies To calculate SF- and attenuation-related physical propertiesof the galaxies, a total IR (TIR) luminosity is required.Various estimators of L TIR were proposed by previous au-thors (Helou et al., 1988; Dale et al., 2001; Dale & Helou, 2002;Sanders & Mirabel, 1996). Takeuchi et al. (2005b) has shown . T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX 7
Fig. 10.
Relation between the AKARI FIR luminosity and total IR (TIR) luminosity of the sample. Left panel is for the AKARIFIR luminosity defined by eq. (4), while right panel is for the AKARI FIR luminosity from 2 bands, defined by eq. (8). Black solidlines are the best least-square fits to the data, and the dotted curves represent the 95 % confidence levels of the lines. Triangles aregalaxies with IRAS 100 µ m quality flag larger than 3, i.e., those with insecure flux density measurement.that we can safely estimate the TIR luminosity by a formula pro-posed by Sanders & Mirabel (1996): L TIR ≡ . × − [13 . L ν (12 µ m) + . L ν (24 µ m) + . L ν (60 µ m) + L ν (100 µ m)] [ L ⊙ ] . (3)Since our sample contains all IRAS band flux densities, we canuse the estimator of Sanders & Mirabel to examine the AKARIFIR flux. In this subsection, we make an attempt to establish a re-liable formula to convert a FIR luminosity measured by AKARIto the TIR luminosity.Since AKARI FIS has continuous bands from ∼ µ m to ∼ µ m, we can easily define FIR flux simply by addingthe flux densities multiplied with their bandwidths (in [Hz]).Hirashita et al. (2008) has defined the AKARI FIR luminosityas L AKARI = ∆ ν ( N60 ) L ν (65 µ m) + ∆ ν ( WIDE-S ) L ν (90 µ m) +∆ ν ( WIDE-L ) L ν (140 µ m) , (4)where ∆ ν ( N60 ) = . × [Hz] (5) ∆ ν ( WIDE-S ) = . × [Hz] (6) ∆ ν ( WIDE-L ) = . × [Hz] . (7)However, since the sensitivity of AKARI N60 is not as goodas other two wide bands, if we can use FIR luminosity definedonly by
WIDE-S and
WIDE-L , it will be useful because we canhave larger number of galaxies. We define L by omittingthe term of N60 as L = ∆ ν ( WIDE-S ) L ν (90 µ m) +∆ ν ( WIDE-L ) L ν (140 µ m) . (8)Figure 10 presents correlations between L AKARI and L TIR . Left panel shows the correlation between L AKARI ofHirashita et al. (2008) and L TIR , while right panel shows the onebetween L and L TIR . The fitting results are as follows:log L TIR = .
940 log L AKARI + . , (9) r = . , (10) andlog L TIR = .
964 log L + . , (11) r = . , (12)where r is the correlation coe ffi cient. The solid lines in Figure 10depict these equations. The envelopes delineated by dotted linesin Figure 10 represent the 95 % confidence intervals.In some cases the quality of IRAS 100 µ m flux density mea-surement is poor; we examine the impact of these objects inFigure 10. Triangles represent galaxies with IRAS 100 µ m qual-ity flag larger than 3, i.e., the ones for which the measurementwas di ffi cult and not secure. However, even so, clearly they liewell within the distribution of the whole sample and do not havea significant impact on the statistical analysis.Since these plots present luminosity-luminosity correla-tions, the very tight correlations are not extremely surprising.However, uncertainties of these equations are within a factor of ∼ L TIR obtained fromAKARI
N60 , WIDE-S , and
WIDE-L [equation (9)] for all thefollowing analysis.
Namely, when we mention L
TIR , it is alwaysAKARI-based total IR luminosity.
Now we can compare the FUV ( L FUV ) and total IR (TIR) ( L TIR )luminosities for our final sample. This is shown in Figure 11.It is striking that the luminosity is dominated by L TIR for thevast majority of our sample, even though we take into accountthat they are FIR-selected. Also, it is worth mentioning that theluminosity at FUV does not exceed 10 L ⊙ . In contrast, L TIR can be much higher.By combining L FUV and L TIR , we can define the luminositycontribution from newly formed stars, L SF : L SF ≡ L FUV + (1 − η ) L TIR , (13) T. T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX
Fig. 11.
Comparison between TIR and FUV luminosities. Thediagonal dotted line represents the case if L TIR equals L FUV .where η is the fraction of the dust emission due to the heatingof grains by old stars which is not directly related to the recentSFR. We adopt a value of 30 % for this fraction (Hirashita et al.,2003). The contribution of L FUV and L TIR is shown in Figure 12.Naturally, the contribution of (1 − η ) L TIR dominates L SF . In con-trast, the contribution of L FUV has a very large scatter, and thecorrelation is very poor. Hence, it is not surprising that it is verydi ffi cult to estimate the total energy emitted by newly formedstars only from FUV information, even in an average sense.The UV contribution becomes significant at lower luminosities L SF < L ⊙ , as seen in the left panel of Figure 12. Here, we interpret the data in terms of SFR. Assuming a con-stant SFR over 10 yr, and Salpeter initial mass function (IMF)(Salpeter, 1955, mass range: 0 . M ⊙ )), we have the relationbetween the SFR and L FUV log SFR
FUV = log L FUV − . . (14)For the IR, to transform the dust emission to the SFR, we assumethat all the stellar light is absorbed by dust. Then, we obtain thefollowing formula under the same assumption for both the SFRhistory and the IMF as those of the FUV,log SFR dust = log L TIR − . + log(1 − η ) . (15)Here, again, η is the fraction of the dust emission by old stars.Thus, the total SFR is simplySFR = SFR
FUV + SFR dust . (16)The obtained SFR is shown as a function of the fraction ofthe contribution of SFR FUV in Figure 13. Reflecting the largescatter of L FUV / L SF , the scatter of SFR FUV / SFR is very large atSFR <
20 M ⊙ yr − . However, quite sharply, no galaxies have alarge contribution of SFR UV at SFR >
20 M ⊙ yr − . The verticaldotted line shows this “threshold” SFR in Figure 13. Galaxies selected in IR are expected to have a quite large dustattenuation. Here we examine the extinction properties of galax-ies in the sample. For this study, a good observational indicatorof dust attenuation is required. The L TIR / L FUV ratio is widelyrecognized to be a robust measure of dust attenuation. This ra-tio was found to increase with the star formation luminosityin a similar way from z = z = . A FUV = − . log L TIR L FUV ! + . log L TIR L FUV ! + . log L TIR L FUV ! + . . (17)Figure 14 presents the variation of L TIR / L FUV as a function of thestar formation luminosity. A clear increase of L TIR / L FUV with L SF is seen. The solid line is the mean trend of local galaxiesfound in our IRAS study (Buat et al., 2007a):log L TIR L FUV ! = .
64 log L SF − . . (18)This line was estimated by a weighting of L TIR / L FUV with1 / V max to eliminate the flux selection e ff ect, i.e., the e ff ect thatmore luminous galaxies can be more easily detected. In contrast,we plot raw values of L TIR / L FUV . However, though the steep riseof the distribution of L TIR / L FUV with L SF is partially because ofthis selection e ff ect, the trend is still well consistent with ourIRAS study (Buat et al., 2007a), and the conclusion would re-main valid. The slope of the UV continuum is commonly used as a proxyto estimate dust attenuation and L TIR / L FUV when IR data arenot available or considered as unreliable (Daddi et al., 2007;Reddy et al., 2008; Reddy & Steidel, 2009). This method isbased on a calibration performed on local starburst galaxies(Meurer et al., 1995, 1999). GALEX observations brought largeamount of data in this field since the slope of the UV contin-uum can be safely deduced from the FUV − NUV color. Studiesbased on GALEX data have shown that the local starburst cal-ibration is not valid for the bulk of local star forming galaxies(e.g., Dale et al., 2005; Cortese et al., 2006; Boissier et al., 2007;Gil de Paz et al., 2007; Johnson et al., 2007): i.e., these galaxiesexhibit lower dust attenuation ( L TIR / L FUV ) than expected fromtheir FUV − NUV color. IR selected galaxies also depart from thestarburst law but in the opposite way: they are found to exhibita L TIR / L FUV ratio much larger than expected (Buat et al., 2005;Goldader et al., 2002). Here, we re-investigate this issue makinguse of a much larger sample than used in Buat et al. (2005).Figure 15 shows the FUV-NUV color against L TIR / L FUV (of-ten called the IR-excess (IRX)– β relation) for our sample galax-ies together with the local starburst relation. The local relationwas taken from Meurer et al. (1999) and converted into the re-lation between the FUV-NUV color and L TIR / L FUV by eqs. (1) . T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX 9
Fig. 12.
Contribution of L TIR and L FUV to the total star formation (SF) luminosity L SF , which is the luminosity produced by newlyforming stars defined by eq. (13). In the left panel, η is the fraction of IR emission produced by dust heated by old stars which is notrelated to the current SF (see main text). Fig. 13.
Contribution of the FUV-estimated (or “directly vis-ible”) SFR, SFR UV , to the total SFR as a function of totalSFR. Vertical dotted line represents the e ff ective boundary abovewhich almost all energy produced by newly forming stars isemitted at IR.and (2) of Kong et al. (2004): L TIR L FUV = [1 . − NUV) + . − . . (19)Most of the galaxies lie below the starburst relation by a fac-tor of 5 to 10, but some locate above the relation. A significantfraction of galaxies above the local starburst line are luminousIR galaxies (LIRGs and ULIRGs). Especially, all the ULIRGshave larger L TIR / L FUV ratios than expected from the relation.This general trend is quite consistent with Buat et al. (2005). Itmay be worth mentioning that LIRGs are consistent with the lo-cal starburst line. These trends are more clearly represented byfunctional fits in Right panel of Figure 15. We used a functional
Fig. 14.
Distribution of L TIR - L FUV ratio as a function of thestar formation luminosity L SF . Solid line is the relation foundby IRAS-GALEX analysis (Buat et al., 2007a), represented byeq. (18).form of eq. (19) for “normal” galaxies with L < L ⊙ , L TIR L FUV = [ a (FUV − NUV) + b ] − c . (20)with letting three parameters, a , b and c free. The fitted relationis well below the local starburst relation originally claimed byMeurer et al. (1999). For more luminous galaxies, eq. (20) doesnot give a reasonable fit, hence we adopted a simple linear fit.Even if LIRGs are roughly consistent with the local starburst re-lation, the fitted line is much flatter than that, because of the exis-tence of galaxies well above the local starburst line with blue UVcolor. It is di ffi cult to conclude in the case of ULIRGs becauseof poor statistics, but the fitted linear relation seems to locateabove the local starburst relation. We have seen that galax- ies with relatively quiescent galaxies ( L < L ⊙ ) in oursample locate well below the relation of Meurer et al. (1999).Similar trend was reported for optically selected galaxies (e.g.,Boissier et al., 2007; Cortese et al., 2006) Boissier et al. (2007)investigated this relation for normal galaxies selected at opti-cal wavelength with the same functional form with this study[eq. (20)] They found a much flatter relation than Meurer’s orig-inal relation, which is presented in Right panel of Figure 15 withdot-dot-dot-dashed line. Cortese et al. (2006) reported a similarresult. What is newly clarified in this study is that the same trendis seen for purely FIR-selected sample of SF galaxies. In addi-tion, Boissier et al.’s relation does not represent the relation forour quiescent galaxy sample. This is naturally understood thatour sample includes strongly reddened galaxies by dust whichcan never be picked up by UV-selection. By using optical–NIR flux densities, we can estimate the stel-lar mass of the sample galaxies. Here we discuss dust attenua-tion properties with respect to the stellar mass. There are vari-ous methods to estimate stellar mass. In this work, we used anSDSS-based method proposed by Yang et al. (2007); these au-thors assume the Kroupa IMF (Kroupa, 2001), while we usedthe Salpeter IMF (Salpeter, 1955) to estimate the SFR of galax-ies. For consistency, we convert their stellar mass estimates usingthe conversion factor given by Bell et al. (2003). As for the accu-racy of the stellar mass estimates, the relation has been calibratedfrom SDSS + K s -band using the rela-tions derived by Bell et al. (2003), and the trends did not changesignificantly. Further exploration for the stellar mass estimationwill be discussed elsewhere.Figure 16 shows the relation between stellar mass M ∗ anddust attenuation in terms of L TIR / L FUV . At a first glance, thedependence of dust attenuation on stellar mass is very strong.Though we should be cautious on the selection e ff ect on thisplot again, the strong dependence might be partially physical.This implies that larger galaxies are more extinguished. Hence,dust attenuation is closely related to the physical size of galaxies.The presence of low mass and high L TIR / L FUV galaxies ∼ M ∗ ≃ –10 M ⊙ is worth mentioning here.Iglesias-P´aramo et al. (2006) made essentially the same analy-sis by GALEX and IRAS. In their studies, such galaxies did notexist in their IR-selected samples (their Figure 12). If the fluxmeasurement is secure, these galaxies deserve a close inspectionto examine their nature.Next, we show the relation between SFR per unit stellarmass, SFR / M ∗ , which is often referred to as the specific SFR(SSFR), and dust attenuation L TIR / L FUV in Figure 17. It is in-deed striking that obviously there is no correlation between thesequantities.First conclusion would be that there is no correlation be-tween global attenuation and SSFR. Generally, many authorsshowed that dust attenuation increases with SFR (see, e.g.,Figure 7 of Buat et al., 2007a). Then, if there is no link withthe SSFR it implies that dust attenuation is linked to the totalamount of SF scaled with galaxy size, because SFR was posi-tively correlated to M ∗ seen above. It may be interpreted as fol-lows: the attenuation is not related to the current-to-past SFR ra-tio, or roughly, the SF history. Probably this could be explainedby the short lifetime of dust grains in the ISM.We should mention that, however, Iglesias-P´aramo et al.(2006) has shown that even for the FIR-selected galaxies there is some correlation between these two quantities (see theirFigure 10b). This issue seems to require further examination.
5. Conclusion
In order to explore SEDs, star formation, and dust extinctionproperties of galaxies in the Local Universe, we have constructeda multiband galaxy sample based on the AKARI FIS All-SkySurvey and GALEX All-Sky Imaging Survey (AIS). We startfrom AKARI All-Sky Survey Bright Source Catalogue β -1.1,and selected galaxies by matching the AKARI sources withthose of the IRAS PSC z . Next, we have measured total GALEXFUV and NUV flux densities by a photometry software whichwe have developed specifically for this purpose. Then, we havematched this sample with SDSS and 2MASS galaxies to ob-tain the basic sample. The basic sample consists of 776 galax-ies. After removing objects with photometry contaminated byforegound sources (mainly in SDSS), we have defined the “se-cure sample” which contains 607 galaxies. Based on this galaxysample, we have explored various properties of galaxies relatedto star formation and dust extinction.Summary and conclusions of this study are as follows:1. The sample galaxies have redshifts < ∼ .
15, and their 90- µ m luminosities range from 10 to 10 L ⊙ , with a peak at10 L ⊙ .2. The SEDs display a very large variety, especially more thanfour orders of magnitude at M-FIR, but if we sort the sampleby 90 µ m, their average SED has a coherent trend: the moreluminous at 90 µ m, the redder the global SED becomes.3. The M r -NUV − r color-magnitude relation of our sampledoes not show a bimodality which is almost always expectedin optically selected galaxy samples. The distribution is uni-modal, centered on the green valley between blue cloud andred sequence seen in optical surveys.4. We have established formulae to convert FIR luminosityfrom AKARI bands to the total infrared (IR) luminosity L TIR .5. The luminosity related to star formation activity ( L SF ) isdominated by the contribution of L TIR even if we take intoaccount the FIR emission from dust heated by old stars.6. With these formulae, we calculated the star formation di-rectly visible with FUV and hidden by dust. At high star for-mation rate (SFR) ( >
20 M ⊙ yr − ), the fraction of directlyvisible SFR, SFR FUV , decreases.7. We estimated the ratio, L TIR / L FUV , which is a direct measureof the FUV attenuation A FUV . The distribution of L TIR / L FUV is consistent with a previous result based on GALEX andIRAS (Buat et al., 2007a).8. We also examined the L TIR / L FUV -UV slope (FUV − NUV)relation. The majority of the sample has L TIR / L FUV ratios5 to 10 times lower than expected from the local starburstrelation (Kong et al., 2004), while some LIRGs and all theULIRGs of this sample have higher L TIR / L FUV ratios. Thistrend was already reported from a previous GALEX-IRASstudy (Buat et al., 2005) obtained by a much smaller sample,and we have confirmed their conclusion.9. By making use of stellar mass information derived fromSDSS flux densities in this work), we have examined the dustattenuation properties in terms of stellar mass. We found thatthe attenuation indicator L TIR / L FUV is correlated to stellarmass of galaxies, M ∗ , but there is no correlation with spe-cific SFR (SSFR), SFR / M ∗ . This may mean that L TIR / L FUV is not linked to the SF history, but simply scales with the size . T. Takeuchi et al.: Star formation and dust extinction of galaxies from AKARI-GALEX 11
Fig. 15.
Distribution of L TIR - L FUV luminosity ratio as a function of UV color FUV − NUV. Dots: galaxies with 60 µ m luminosity L < L ⊙ , filled squares: galaxies with 10 ≤ L < L ⊙ (IR luminous galaxies: LIRGs), and filled circles: galaxies with10 ≤ L < L ⊙ (Ultraluminous IR luminous galaxies: ULIRGs). Solid lines in each panel represent the relation for Localstarbursts proposed by Meurer et al. (1999), which was converted into the relation between the FUV-NUV color [eq. (19)]. Leftpanel shows the raw distribution of L TIR - L FUV luminosity ratio, while in Right panel functional fits are overplotted on the data. Forgalaxies with L < L ⊙ , eq. (19) is used, while for more luminous galaxies, simple linear relations are adopted because of poorfit of the function. Dotted line: a fit to galaxies with L < L ⊙ by eq. (20); Dot-dot-dot-dashed line: a fit to a sample of normalgalaxies by eq. (20) presented by Boissier et al. (2007); Dashed line: a linear fit to LIRGs; Dot-dashed line: a linear fit to ULIRGs.of galaxies. However, this is at odds with previous result ofIglesias-P´aramo et al. (2006).This sample will serve as an important reference sample at z = / future observationalprojects, like Herschel: for instance, this can be used to constructa set of SEDs for discussing higher- z observational strategy, oras a baseline test sample to investigate a method of extractinggalaxies only from FIR flux information (e.g., Pollo et al., 2010).However, since our first sample is not complete in manysenses, further analysis will be desired. We plan to construct alarger multiwavelength sample from the next release of AKARIFIS All-Sky Survey in the near future. Acknowledgements.
We deeply thank the anonymous referee for her / his care-ful reading of the original manuscript, useful suggestions and comments whichimproved the clarity of the paper. This work is based on observations withAKARI, a JAXA project with the participation of ESA. TTT has been supportedby Program for Improvement of Research Environment for Young Researchersfrom Special Coordination Funds for Promoting Science and Technology, andthe Grant-in-Aid for the Scientific Research Fund (20740105) commissioned bythe Ministry of Education, Culture, Sports, Science and Technology (MEXT)of Japan. VB and DB have been supported by the Centre National des EtudesSpatiales (CNES) and the Programme National Galaxies (PNG). We thankAgnieszka Pollo, Mai Fujiwara, Akira Ikeyama, Ryosuke Asano, Akio K. Inoue,Hiroshi Shibai, Yasuo Doi, Hideaki Fujiwara, Mitsunobu Kawada, HidehiroKaneda, Hiroyuki Hirashita, and TakakoT. Ishii for fruitful discussions and com-ments. TTT, FTY, and KLM are partially supported from the Grand-in-Aid forthe Global COE Program “Quest for Fundamental Principles in the Universe:from Particles to the Solar System and the Cosmos” from the MEXT. References
Bell, E. F., McIntosh, D. H., Katz, N., & Weinberg, M. D. 2003, ApJS, 149, 289Bell, E. F., et al. 2005, ApJ, 625, 23Boissier, S., et al. 2007, ApJS, 173, 524Buat, V., Iglesias-P´aramo, J, Seibert, M. et al. 2005, ApJ, 619, L51Buat, V., Takeuchi, T. T., Iglesias-P´aramo, J, et al. 2007a, ApJS, 173, 404Buat, V., Marcillac, D., Burgarella, D. et al. 2007b, A&A, 469, 19 * [ M O • ])0.11.010.0100.01000.010000.0 L T I R / L F U V Fig. 16.
Relation between stellar mass M ∗ and dust attenuationindicator L TIR / L FUV . Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245Chary, R., Elbaz, D. 2001, ApJ, 556, 562Cortese, L., et al. 2006, ApJ, 637, 242Dale, D. A., Helou, G., Contursi, A. et al. 2001, ApJ, 549, 215Dale, D. A. , Helou, G. 2002, ApJ, 576, 159Dale, D. A. , Bendo, G. J., Engelbracht, C. W. et al. 2005, ApJ, 633, 857Daddi, E., et al. 2007, ApJ, 670, 156Dwek, E., & Scalo, J. M. 1980, ApJ, 239, 193Dwek, E. 1998, ApJ, 501, 643Genzel, R., & Cesarsky, C. J. 2000, ARA&A, 38, 761Gil de Paz, A., et al. 2007, ApJS, 173, 185Goldader, J. D., Meurer, G., Heckman, T. M., Seibert, M., Sanders, D. B.,Calzetti, D., & Steidel, C. C. 2002, ApJ, 568, 651 -12 -11 -10 -9 -8 SFR/M * [yr -1 ]0.11.010.0100.01000.010000.0 L T I R / L F U V Fig. 17.
Relation between SFR per unit stellar mass, SFR / M ∗ (specific SFR: SSFR), and dust attenuation L TIR / L FUV . Helou, G., Khan, I. R., Malek, L., & Boehmer, L. 1988, ApJS, 68, 151Hirashita, H., Buat, V., & Inoue, A. K. 2003, A&A, 410, 83Hirashita, H., Kaneda, H., Onaka, T., & Suzuki, T. 2008, PASJ, 60, 477Hopkins, A. M., & Beacom, J. F. 2006, ApJ, 651, 142Iglesias-P´aramo, J., Buat, V., Takeuchi, T. T., et al. 2006, ApJS, 164, 38Jarrett, T. H., Chester, T., Cutri, R., Schneider, S., Skrutskie, M., & Huchra, J. P.2000, AJ, 119, 2498Jeong, W.-S., et al. 2007, PASJ, 59, 429Johnson, B. D., et al. 2007, ApJS, 173, 377Kartaltepe, J. S., et al. 2009, Astronomical Society of the Pacific ConferenceSeries, 408, 272Kawada, M., et al. 2007, PASJ, 59, 389Kong, X., Charlot, S., Brinchmann, J., & Fall, S. M. 2004, MNRAS, 349, 769Kroupa, P. 2001, MNRAS, 322, 231Martin, D. C., et al. 2005, ApJ, 619, L1Martin, D. C., et al. 2007, ApJS, 173, 342Meurer, G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64Meurer, G. R., Heckman, T. M., Leitherer, C., Kinney, A., Robert, C., & Garnett,D. R. 1995, AJ, 110, 2665Morrissey, P., et al. 2007, ApJS, 173, 682Murakami, H., et al. 2007, PASJ, 59, 369Nozawa, T., Kozasa, T., Umeda, H., Maeda, K., & Nomoto, K. 2003, ApJ, 598,785Onaka, T., et al. 2007, PASJ, 59, S401Pollo, A., Rybka, P., & Takeuchi, T. T. 2010, A&A, submitted to this volumeNeugebauer, G., et al. 1984, ApJ, 278, L1Reddy, N. A., Steidel, C. C., Pettini, M. et al. 2008, ApJS, 175, 48Reddy, N. A., & Steidel, C. C. 2009, ApJ, 692, 778Salim, S., et al. 2007, ApJS, 173, 267Salpeter, E. E. 1955, ApJ, 121, 161Sanders, D. B., & Mirabel, I. F. 1996, ARA&A, 34, 749Saunders, W., et al. 2000, MNRAS, 317, 55Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Seibert, M., et al. 2005, ApJ, 619, L55Skrutskie, M. F., et al. 2006, AJ, 131, 1163Soifer, B. T., Houck, J. R., & Neugebauer, G. 1987, ARA&A, 25, 187Soifer, B. T., Helou, G., & Werner, M. 2008, ARA&A, 46, 201Takeuchi, T. T., Hirashita, H., Ohta, K., Hattori, T. G., Ishii, T. T., & Shibai, H.1999, PASP, 111, 288Takeuchi, T. T., Ishii, T. T., Hirashita, H., Yoshikawa, K., Matsuhara, H., Kawara,K., & Okuda, H. 2001, PASJ, 53, 37Takeuchi, T. T., Kawabe, R., Kohno, K., Nakanishi, K., Ishii, T. T., Hirashita, H.,& Yoshikawa, K. 2001, PASP, 113, 586Takeuchi, T. T., Yoshikawa, K., & Ishii, T. T. 2003, ApJ, 587, L89Takeuchi, T. T., Hirashita, H., Ishii, T. T., Hunt, L. K., & Ferrara, A. 2003,MNRAS, 343, 839Takeuchi, T. T., & Ishii, T. T. 2004, ApJ, 604, 40Takeuchi, T. T., Buat, V., Burgarella, D. 2005a, A&A, 440, L17 Takeuchi, T. T., Buat, V.,Iglesias-P´aramo, J, Boselli, A., Burgarella, D. 2005b,A&A, 432, 423Takeuchi, T. T., Ishii, T. T., Nozawa, T., Kozasa, T., & Hirashita, H. 2005c,MNRAS, 362, 592Vergani, D., et al. 2009, arXiv:0909.1968Verma, A., Charmandaris, V., Klaas, U., Lutz, D., & Haas, M. 2005, SpaceScience Reviews, 119, 355Yamamura, I., et al. 2009 in AKARI, a light to illuminate the misty Universe, eds.T. Onaka, G. White, T. Nakagawa, I. Yamamura, ASP Conf. Ser., in pressYamamura, I., Fukuda, Y., & Makiuti, S. 2008, AKARI / FIS All-Sky SurveyBright Source Catalogue Version β -1 Release Note (Rev. 1)Yang, X., Mo, H. J., van den Bosch, F. C., Pasquali, A., Li, C., & Barden, M.2007, ApJ, 671, 153Zheng, X. Z. et al. 2006, ApJ, 640, 784 Appendix A: Luminosity vs. redshift diagrams ofthe sample