Relation between polycyclic aromatic hydrocarbon, Brα and infrared luminosity of local galaxies observed with AKARI
Kazumi Murata, Takao Nakagawa, Hideo Matsuhara, Kenichi Yano
aa r X i v : . [ a s t r o - ph . GA ] J u l PASJ:
Publ. Astron. Soc. Japan , 1– ?? , c (cid:13) Relation between polycyclic aromatic hydrocarbon, Br α and infraredluminosity of local galaxies observed with AKARI Kazumi
Murata , Takao
Nakagawa , Hideo
Matsuhara , and Kenichi
Yano
Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency, Sagamihara, Kanagawa 252-5210, [email protected] (Received 2017 April 3; accepted 2017 June 27)
Abstract
We produce a catalogue of polycyclic aromatic hydrocarbon (PAH) 3.3 µ m, Br α and infrared luminosity( L (IR)) of 412 local galaxies, and investigate a relation between these physical parameters. We measurethe PAH 3.3 µ m and Br α flux using AKARI µ m spectra and the L (IR) using the AKARI -all-sky-survey data. The L (IR) and redshift ranges of our sample are L (IR)=10 . − . L ⊙ and z spec = 0 . − . L (PAH 3.3 µ m) to L (IR) is constant at L (IR) < L ⊙ whereasit decreases with the L (IR) at higher L (IR). Also, the ratio of L (Br α ) to L (IR) decreases with the L (IR).The both L (PAH)/ L (IR) and L (Br α )/ L (IR) ratios are not strongly dependent on galaxy type and dusttemperature. The relative weakness of the two ratios could be attributed to destruction of PAH, a lack ofUV photons exciting PAH molecules or ionising hydrogen gas, extremely high dust attenuation, or activegalactic nucleus contribution to the L (IR). Although we cannot determine the cause of the decreases ofthe luminosity ratios, a clear correlation between them implies that they are related with each other. Thecatalogue presented in our work will be available at the AKARI archive web page.
Key words:
Galaxies: star formation, Galaxies: evolution, Infrared: galaxies
1. Introduction
For understanding galaxy evolution, star formation rate(SFR) is one of the most basic parameters of a galaxy. TheSFR of galaxies has been derived by many researchers withvarious methods. Among various SFR tracers, rest frameUV and optical emission have been widely used for esti-mating SFR, even for high-redshift galaxies. Nonetheless,these tracers are strongly attenuated by interstellar dust,which is more serious for galaxies with more luminous atinfrared (Takeuchi et al. 2010). To correctly estimate SFRof such galaxies, we need SFR tracers that are less affectedby dust attenuation, i.e. those at infrared wavelength.There are three kinds of SFR tracers at infrared: poly-cyclic aromatic hydrocarbon (PAH) emission, hydrogenrecombination lines such as Br α (4.052 µ m), and infraredluminosity ( L (IR)). PAH is a large molecule, exists mainlyat a photo-dissociation region (PDR), absorbs UV pho-tons from young stars and emits the energy at near- tomid-infrared such as 3.29, 6.2, 7.7, and 11.3 µ m (Tielens2008). Because the energy source is UV light from youngstars, PAH emission reflects the star-forming activities ofgalaxies. Br α line is a hydrogen recombination line emit-ted from an HII region. It directly traces the ionisingphotons produced by young stars. In addition, assum-ing the galactic attenuation curve, the dust attenuationis merely A(Br α ) ∼ α ) owing to the long wave-length. These properties make the Br α line an ideal SFRtracer for galaxies with strong dust extinction like ultraluminous infrared galaxies (ULIRGs). L (IR) is luminos-ity integrated over infrared spectrum at 8-1000 µ m. It isemitted by interstellar dust that absorbs UV and optical radiation from young stars. Hence for galaxies with ex-tremely strong dust attenuation like ULIRGs, the L (IR)approximates the bolometric luminosity and becomes anideal SFR tracer (Kennicutt 1998a).Combining these SFR tracers provides us various galaxyinformation because the relation between them is differ-ent in different physical condition of galaxies. For exam-ple, if PAH particles are destroyed by strong interstellarradiation, they are not proportional to the SFR of galax-ies so that L (PAH)/ L (IR) ratio can be used for inferringphysical condition of interstellar medium (Nordon et al.2012; Murata et al. 2014). Hydrogen recombination linesfrom a dusty HII region could be reduced due to a lackof ionising photons (Vald´es et al. 2005; Yano et al. 2016).These phenomena indicate that combining the three SFRtracers enables us to investigate galaxy properties relatedto the star-forming activity.The Japanese infrared astronomical satellite AKARI (Murakami et al. 2007) can provide these SFR tracers. Aspectroscopic observation at 2-5 µ m wavelength with theInfrared Camera (IRC; Onaka et al. 2007) is capable ofmeasuring PAH 3.3 µ m and Br α emission. Furthermore,using the all-sky-survey data with the far-infrared sur-veyer (FIS; Kawada et al. 2007) we can estimate the L (IR)of galaxies. Some researchers have provided PAH and Br α luminosity of local galaxies using these unique capabili-ties (Imanishi et al. 2008; Imanishi et al. 2010; Kim etal. 2012; Lee et al. 2012; Yamada et al. 2013; Ichikawaet al. 2014; Castro et al. 2014). For example Imanishi etal.(2008,2010) provided PAH flux of 154 (U)LIRGs, for 67of which they also provided Br α flux. However, a signifi-cant amount of local galaxy data is still unpublished. K. Murata et al. [Vol. ,In this work, we analyse all AKARI /IRC spectroscopicdata and produce a catalogue of 412 local galaxies withPAH 3.3 µ m and Br α flux. We also estimate and provideinfrared luminosity of these galaxies using AKARI -all-sky-survey data. Using these data we investigate the relationbetween the three SFR tracers. This paper is organisedas the following. In section 2 and 3, we describe dataused in our study and how we estimated the PAH 3.3 µ m,Br α , and infrared luminosity. In section 4, we show ourresults and present the catalogue. In section 5, we discussthe relation between the three SFR tracers. Finally wesummarise our study in section 6. Throughout our workwe adopt a cosmology with Ω M =0.3, Ω Λ =0.7, and H =70km s − Mpc − .
2. Data
We used spectra observed with the
AKARI /IRC grismspectroscopy. The IRC has a 1’ ×
1’ aperture mask, inwhich the light from galaxies is incident and dispersedwith wavelength resolution of R ∼ AKARI oper-ation is divided into two phases: Phase 1-2, and phase 3.In the phase 1-2, liquid helium is used as a cryogen andobservation was conducted at 2006 April to 2007 August.In the phase 3, however, liquid helium is exhausted andthe sensitivity is degraded. The observed period of thisphase is from 2008 June to 2010 February. The spec-tra were recently reduced by the
AKARI data processingand analysis team in a standard manner with the latestpipeline (Usui et al. in preparation). They extracted thespectra within 10 arcsec width in spatial direction fromthe reduced images. We used sixteen observation programmes for our sam-ple as summarised in table 1. Here we briefly describe howtargets are selected in each programme. We note that weincluded only galaxies with noticeable PAH 3.3 µ m emis-sion in the spectra into our sample, and that some galaxieswere duplicately observed in different programmes. AGNUL:
This programme is planned for observingLIRGs and ULIRGs selected with the Bright GalaxySample (BGS; Soifer et al. 1987; Sanders et al. 1995),the revised BGS (Sanders et al. 2003), and the IRAS 1Jy sample (Kim and Sanders 1998). Although the tar-get selection is not complete, the completeness is limitedonly by the sky position of galaxies. The programme isconducted both in the phase 1-2 and phase 3.
AMUSE:
This programme is for observing galaxiesfrom the 5 mJy unbiased
Spitzer extra galactic survey(5MUSES; Wu et al. 2010). The galaxies are brighterthan 5 mJy at 24 µ m and located in the Spitzer -first-look-survey and SWIRE fields.
BRSFR:
This programme is for observing galaxieswhose radial velocity is ∼ L ⊙ . The original version of the pipeline was developed in Ohyama etal. (2007).
Table 1.
Observation programmes used in our catalogue.The second column indicates the number of galaxies in ourcatalogue. We have 10 galaxies taken with multiple pro-grammes, leading the sum of the second column is over thenumber of the catalogued galaxies.
Programme F µm > L (IR) < L ⊙ CLNSL 6 composite in BPTCOABS 1 bright Seyfert-2DTIRC 11 data checkEGANS 8 SDSS at z =0.1-0.5GOALS 48 LIRGH2IRC 1 H detectedISBEG 19 blue early typeMSAGN 82 AKARI /mid-IR AGNMSFGO 11
AKARI /mid-IRNISIG 10 star-formingNULIZ 18 ULIRGQSONG 3 QSOSYDUS 3 Seyfert 1&2
CLNSL:
This programme is for observing compos-ite galaxies and low ionisation nuclear emission regions(LINERs) selected from infrared galaxies in Hwang et al.(2007). COABS:
This programme is for observing nuclei ofbright Seyfert-2 galaxies to investigate the physical con-dition of molecular tori.
DTIRC:
This is a programme for data checks by theIRC data team. The targets include standard stars andsome galaxies.
EGANS:
This is a programme for observing SDSSgalaxies at z =0.1-0.5. Our sample includes eight galax-ies from this programme. GOALS:
In this programme the sample is based on theGreat Observatory All-Sky LIRG Survey (GOALS; Armuset al. 2009). This is a non-bias survey for investigatinglocal LIRGs selected from the revised BGS.
H2IRC:
This programme is conducted for observinggalaxies with strong emission of molecular hydrogen.
ISBEG:
This programme is for observing early typegalaxies with unusually blue colour selected from theSDSS (Lee et al. 2010). The redshift range of the tar-gets is z =0.02-0.1. MSAGN:
It is a follow-up programme of sourcesdetected with the
AKARI -mid-infrared-all-sky survey.Galaxies with F (9 µm ) /F ( Ks ) > | b | >
30 deg were selected as a target (Oyabu et al. 2011).
MSFGO:
It is also a follow-up programme of sourcesdetected with the
AKARI -mid-infrared-all-sky survey.The sources located at galactic latitude of | b | <
30 degwere selected as a target.
NISIG:
This is a programme for observing star-forming In a BPT diagram (Baldwin, Phillips & Terlevich 1981), theylocated in an intermediate region between Seyfert and HII galaxyregions. o. ] PAH, Br α , and LIR of local galaxies observed with AKARI
NULIZ:
The targets were selected from 320 ULIRGsin Hwang et al. (2007). Their ULIRGs were se-lected by cross-matching the IRAS faint source cataloguewith galaxy redshift surveys: the SDSS DR4 (Adelman-McCarthy et al. 2006), 2dF Galaxy Redshift survey(Colless et al. 2001), and 6dF Galaxy survey (Jones etal. 2004).
QSONG:
This is a programme for observing both lowand high redshift quasi stellar objects (QSOs; Kim et al.2015; Jun et al. 2015). We used only low redshift QSOsshowing PAH 3.3 µ m emission. SYDUS:
This is a programme for observing a 12 µ msample of nearby Seyfert 1 and 2 galaxies. We used the data from
AKARI -far-infrared-all-sky sur-vey for estimating the L (IR) of our sample. We used boththe bright source catalogue version 2 (BSC2; Yamamuraet al. in prep.) and the intensity map (Doi et al.2015; Takita et al. 2015). The detection limit of the BSC2is 0.44 Jy at the Wide-S band (90 µ m) and 3.4 Jy at theWide-L band (140 µ m). The spatial resolution is 78 arc-sec at the Wide-S and 88 arcsec at the Wide-L bands,respectively (Takita et al. 2015).
3. Method
While we basically used spectra provided by the
AKARI data processing and analysis team, we extracted spectrafrom the provided images for some galaxies because of thefollowing two reasons. First, some spectra are failed to beextracted due to a wrong position setting. Second, evenif multiple galaxies were observed in one image, only onespectrum is extracted.While the
AKARI team extracted the spectra with afixed box size of 10 arcsec, we applied a Gaussian fittingto optimise the aperture size for each galaxy. First wedetermined the spatial width and position of the spectrawith a Gaussian fitting. The fitting was repeated along thewavelength direction. While this leads different width andposition in different wavelength, we took median valuesand fixed them.Once width and position were fixed, we extracted thespectra using a Gaussian fitting with only one free pa-rameter of flux. We visually confirmed that the fittingwas correctly conducted. We combined the 1D spectra ofgalaxies that were observed multiple times . The spectra were scaled to the 3.4 µ m flux from theWide field Infrared Survey Explorer ( WISE ; Wright et al.2010) all-sky catalogue owing to the two reasons.First, the sensitivity of the IRC is known to be changedwith the detector temperature, and the current version of Both data products are available in the
AKARI archive page. Typically, they were observed three times. F ( . ) W I S E / F ( . ) AKA R I w1rsemi[arcsec] Fig. 1.
Flux ratio between
WISE and
AKARI against galaxysize, w rsemi . The F (3 . AKARI was estimated from thespectra convolved with the response curve of the
WISE
W1band. The sample used in this histogram is from our finalcatalogue and galaxies rejected from the catalogue are notincluded (see section 3.3). the pipeline does not provide correction for the sensitivityvariation while the function is expected to be implementedin the next release.Second, in case when the size of a galaxy is larger com-pared to the aperture used for extracting spectra, theaperture correction is needed. Among the
WISE magni-tudes, we used the w1mag 8 for smaller galaxies than the w1rsemi of 30 arcsec and the w1gmag for larger galaxies.The w1mag 8 is a W1 24.75 arcsec radius aperture mag-nitude while the w1gmag is a magnitude in an ellipticalaperture whose size is determined by the 2MASS extendedsource catalogue. To convert a magnitude from Vega toAB, we added 2.699 mag to both w1mag 8 and w1gmag .Fig.1 shows the scaling factors applied in our studyagainst galaxy size. Despite the scatter, it shows thatlarger galaxies have a larger scaling factor. It also showsthat most of our sample have a scale factor of less than 2,and there is only one object with F W ISE /F AKARI largerthan 10.
For measuring PAH 3.3 µ m and Br α flux, we performeda Gaussian profile fitting to the spectra. At first, we deter-mined the baseline with a linear fitting around each lineposition. The free parameters are slope and intercept. Wetypically used rest frame 3.05-3.2 µ m and 3.4-3.72 µ m fora PAH 3.3 µ m line and 3.9-4.0 µ m and 4.1-4.2 µ m fora Br α line. In case S/N ratio at these wavelength rangewas low, the range was slightly modified, depending on http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/sec4 4h.html K. Murata et al. [Vol. ,galaxies.Second, we fitted each emission line using a Gaussianwith the determined baseline. The free parameters wereflux, central wavelength, and line width. The fitted restframe wavelength range is typically 3.2-3.4 µ m for a PAH3.3 µ m line and 4.0-4.1 µ m for a Br α line. Redshifts wereestimated from the central wavelength. Equivalent widthswere measured from the ratio between line and continuumflux. The properties of these physical values are shown insection 4. In case the spectrum show a sub-peak of PAHemission at 3.4 µ m, we avoided it in the fitting processes,following Imanishi et al. (2010).We included only galaxies whose PAH emission is de-tected with over 3 σ into our sample, where we estimatedthe flux error only from the fitting residual. As a result,our sample includes 412 galaxies, among which we alsodetected the Br α emission from 264 galaxies. The luminosity distances of galaxies were estimatedfrom their redshifts. The redshifts were obtainedfrom the SIMBAD astronomical database or from theNASA/IPAC Extragalactic Database (NED) . For galax-ies with z spec < .
01 or unknown redshift we used the“Redshift-Independent Distances” from the NED insteadof distances estimated from the redshift. We did notfind the redshift-independent distances for nine galaxiesat z spec < .
01. Among which we estimated the distanceof IC0836 ( z spec = 0 . z spec = 0 . z spec = 0 . We estimated the infrared luminosity of our sample us-ing the bright source catalogue version 2 (BSC2) of the
AKARI -far-infrared-all-sky survey. We cross-matched thecatalogue with our sample using a search radius of 30arcsec. Among our sample, 340 galaxies were matchedwith the BSC2. For the remaining galaxies, we performedaperture photometry on the FIS all-sky map (Doi et al.2015; Takita et al. 2015) and 43 galaxies were detectedwith over 3 σ at the 90 µ m band. Most of them are fainterthan F ν (90) < µ m map, for which we couldmisidentify or overestimate their flux, and we flagged themin the final catalogue (FISflag in the catalogue; see table2). We have FIS photometry for 383 galaxies in total.Among these, the luminosity distance for three galaxiescould not be constrained thus we did not derive infraredluminosity for these objects.To estimate infrared luminosity, we applied an equa-tion suggested in Solarz et al. (2016). They derivedthe equation by comparing the BSC2 with the InfraredAstronomical Satellite (IRAS) all-sky survey. The derived http://simbad.u-strasbg.fr/simbad/ https://ned.ipac.caltech.edu/ equation is as follows.log L (IR) = 1 .
016 log L AKARI /L ⊙ + 0 .
349 ( ± . , (1)where, L AKARI = ∆ ν (Wide − S) L ν (90 µ m)+∆ ν (Wide − L) L ν (140 µ m) , (2)∆ ν (Wide − S) = 1 . × [Hz] , ∆ ν (Wide − L) = 0 . × [Hz] . We did not use the 65 µ m band flux because we found thatit produces a significant deviation in L (IR) for some galax-ies. For galaxies without 140 µ m band flux, we used only90 µ m flux to estimate L (IR) using the following equation,which was determined to match the L (IR) derived above.log L (IR) = 0 .
976 log [ ∆ ν (Wide − S) L ν (90 µ m) / L ⊙ ]+0 .
995 ( ± .
11) (3)We added 0.11 dex to the error of the L (IR) due to theuse of the above equations.
4. The PAH 3.3 µ m and Br α flux catalogue In this section we provide a catalogue of flux and equiv-alent width of PAH 3.3 µ m and Br α emission and infraredluminosity. We catalogued 412 local galaxies whose PAH3.3 µ m emission was detected over 3 σ . Among our sam-ple, 264 galaxies have Br α flux and equivalent width, and380 have L (IR). In our knowledge, it is the largest cata-logue of PAH 3.3 µ m and Br α emission. The cataloguecan be obtained from the AKARI archive page . We alsopresent a part of the catalogue in table 2. We show theproperties of our catalogue in the following subsections. Fig.2 shows histograms of flux (a;left) and equivalentwidth (b;right) of PAH and Br α emission. From Fig.2awe can see that most of our sample is distributed at F lux > × − ergs − cm − for both PAH and Br α emis-sion.On the other hand, the distribution edge of the equiv-alent width is different between the two lines; the edge ofthe EW(Br α ) is ∼ × lower than that of the EW(PAH).This implies a less bias against low equivalent width,which is clearly different from a narrow-band-imaging-emission-line survey. We investigate the difference between redshifts derivedfrom PAH and Br α , and those from literature. Fig.3ashows that our redshifts are underestimated by ∆ z ∼ σ ( z ) ∼ , uncertainty in the WCS directly affects the Because the slit mask is as large as 1’ × o. ] PAH, Br α , and LIR of local galaxies observed with AKARI N u m b e r Flux[10 -14 erg s -1 cm -2 ](a) PAH 3.3 µ mBr α N u m b e r Equivalent width[nm](b)PAH 3.3 µ mBr α Fig. 2.
Histogram of flux (a) and equivalent width (b) of PAH (solid red line) and Br α (green broken line) emission in our catalogue.The bin sizes are 0.2 dex for flux histogram and 0.1 dex for those of equivalent width. wavelength calibration. This underestimation cannot benegligible, especially at z < .
01. Hence, we did not useredshifts measured in our work. This leads five galaxieswith no redshift, and hence no distance information.The figure also shows that the difference between z (PAH) and z (Br α ). The mean and standard devia-tion is ∆ z = − . σ (∆ z ) = 0 . z =0.002-0.003, this difference is small enough.In other words, the non-linear term of wavelength-calibration issue is negligible.We show the redshift distribution of our sample inFig.3b. We can see that the redshift is distributed mostlyat z spec ∼ z spec ∼ In Fig.4 we show PAH 3.3 µ m, Br α and infrared lu-minosity against the redshift. Among our whole sample,70% have L (IR) > L ⊙ ; 179 (43%) and 111 (27%) outof 412 galaxies are LIRGs and ULIRGs. Specifically, 90%of the galaxies at z spec > . α luminosity are roughly three and fourorders of magnitudes lower than the L (IR), respectively. We classified our sample into five categories: SF,AGN, LINER, composite and unknown. We referred theSIMBAD and the Sloan digital sky survey (SDSS) datarelease 12. By using the SIMBAD, Seyfert (Sy), radiogalaxies (rG), AGN, QSO and possible quasar were re-garded as AGN, LINER was regarded as LINER, and HIIgalaxy and starburst galaxy were regarded as SF. Fromthe SDSS catalogue, we used the “BPT” parameter, whichclassified sample into the above five categories. Galaxieswith a BPT parameter of “Seyfert/LINER” were classifiedas Seyfert. In case the SIMBAD and the SDSS disagree,we prioritised the former. In total our catalogue has 59SF, 101 AGN, 50 LINER, 20 composite, 182 unknown (no information) galaxies. K. Murata et al. [Vol. , N u m b e r ∆ z(a) z(Br α ) - z spec z(PAH) - z spec z(PAH) - z(Br α ) N u m b e r Redshift (b)
Fig. 3. (a):Difference between redshifts from PAH emission, Br α emission, and literature ( z spec ). Our red-shifts, z (PAH) and z (Br α ), are underestimated by ∆ z ∼ L u m i no s it y [ L S un ] RedshiftL(IR)L(PAH)L(Br α ) Fig. 4.
Distribution of PAH 3.3 µ m (green), Br α (blue), andinfrared luminosity (red) against the redshift. o .] P AH , B r α , a nd L I R o f l o c a l ga l a x i e s o b s e r v e d w i t h A K A R I Table 2.
PAH 3.3 µ m, Br α , and L (IR) catalogue of local galaxies observed with AKARI . This is a part of the catalogue. The full catalogue will be available from
AKARI archiveweb page.
Name RA(J2000) Dec(J2000) z spec D L w1rsemi a Scale b log L (IR) log L err(IR) FISflag c type Programme(Mpc) (arcsec) (L ⊙ )2MFGC12572 234.04889 -5.39759 0.0270 118.00 22.35 1.12 11.46 0.12 0 LINER AGNUL P35MUSES171 241.16930 55.56920 0.0780 353.66 8.98 1.11 11.33 0.14 0 Composite AMUSEAGC221050 192.55768 7.57904 0.0383 169.00 13.97 1.13 11.48 0.12 0 SF MSAGNArp148 165.97421 40.85007 0.0345 151.75 13.45 1.09 11.63 0.11 0 Unknown AGNUL P3CGCG049-057 228.30526 7.22500 0.0130 56.22 16.05 1.06 11.27 0.11 0 Unknown AGNUL P3ESO0255-IG007 96.84040 -47.17670 0.0401 177.06 12.04 1.10 11.85 0.11 0 Unknown GOALS,DTIRCIC0836 193.97508 63.61233 0.0092 39.66 38.48 1.25 10.16 0.12 0 Unknown MSAGNIIIZw035 26.12708 17.10139 0.0274 119.80 11.75 0.94 11.59 0.11 0 Unknown GOALSIRAS01173+1405 20.01130 14.36190 0.0312 136.93 14.21 1.04 11.64 0.11 0 LINER AGNUL P3PAH Br αz zerr Flux Fluxerr EW EWerr z zerr
Flux Fluxerr EW EWerr(erg / s / cm ) (erg / s / cm ) ( µ m) ( µ m) (erg / s / cm ) (erg / s / cm ) ( µ m) ( µ m)0.0248 0.0006 25.47 1.35 0.0553 0.0029 0.0234 0.0007 1.07 0.24 0.0035 0.00080.0663 0.0011 3.20 0.31 0.0579 0.0057 0.0610 0.0004 0.26 0.04 0.0086 0.00120.0347 0.0002 37.67 1.17 0.1188 0.0021 0.0350 0.0001 3.78 0.13 0.0185 0.00040.0309 0.0003 27.86 0.84 0.0980 0.0030 0.0322 0.0002 2.35 0.12 0.0122 0.00060.0096 0.0004 15.23 0.54 0.0439 0.0016 0.0105 0.0005 1.67 0.16 0.0080 0.00080.0337 0.0002 66.34 1.28 0.1605 0.0031 0.0349 0.0001 7.05 0.21 0.0225 0.00070.0109 0.0006 27.06 1.16 0.0387 0.0013 0.0077 0.0009 4.76 0.47 0.0116 0.00110.0204 0.0004 9.52 0.49 0.0469 0.0019 0.0183 0.0010 1.79 0.21 0.0129 0.00140.0227 0.0004 40.89 1.43 0.1182 0.0041 0.0243 0.0001 5.22 0.14 0.0142 0.0004 a Semi-major axis of the elliptical aperture used to measure source in the
WISE µ m band. b F (3 . WISE /F (3 . AKARI ratio use to scale line flux. c K. Murata et al. [Vol. ,
5. Discussion
In the previous section, we produced a catalogue of PAH3.3 µ m and Br α flux, together with the infrared luminos-ity. In this section we investigate relationships betweenthese values. L (IR) In Fig.5a we show a ratio between the L (PAH) andinfrared luminosity against the infrared luminosity. Atlog L(IR)/L ⊙ <
11, we see that the L (PAH)/ L (IR) ratiois constant, ∼ − . At the higher L (IR), in contrast, thisratio decreases with the L (IR) and it is down to ∼ − at log L (IR)/L ⊙ ∼ .
5, consistent with e.g. Yamada etal. (2013).Four optical types are colour-coded in the figure: SFincluding composite ones (blue), LINER (green), AGN(red), and unknown (grey). We show median values with68% range in each 0.7 dex log LIR bin in Fig.5c. We foundno strong dependence of L (PAH)/ L (IR) on the galaxytype; any types of galaxies at higher L (IR) have lower L (PAH)/ L (IR).In Fig.5e we show the same figure for galaxies with F ν (90) /F ν (140) > F ν (90) /F ν (140) < z spec < . F ν (90) /F ν (140) ratio reflects the dust temperature andthe ratio of unity corresponds to T dust ∼
45 K. Wecan see that galaxies with higher F ν (90) /F ν (140) ratiotend to distribute at higher L (IR). Nonetheless, at fixed L (IR), we found no dependence of L (PAH)/ L (IR) on the F ν (90) /F ν (140) ratio. α ) vs L (IR) In Fig.5b we compare SFR derived from the Br α emission and that from the L (IR), as a function of L (IR). For estimating SFR we used the following equa-tions (Kennicutt, Tamblyn, & Congdon 1994; Kennicutt1998b).SFR = 7 . × − L (H α ) [erg / s]= 2 . × − L (Br α ) [erg / s]= 1 . × − L (Br α ) [L ⊙ ] , (4)SFR = 1 . × − L(IR) [L ⊙ ] , (5)where we assume H α /Br α ratio of 36, according to thecase B with T=10000 K and n e =10 − [cm − ] (Hummerand Storey 1987), and the Salpeter IMF. In Fig.5b wecan see that the SFR(Br α ) is consistent with SFR(IR) atthe log L (IR)/L ⊙ ∼ L (IR),the SFR(Br α )/SFR(IR) ratio gradually decreases with the L (IR) and it is down to 0.1 at log L (IR)/ L ⊙ ∼ α )/SFR(IR) ratio also has no strong depen-dence on the galaxy optical type. The relation is similar tothat of the L (PAH)/ L (IR) ratio. Both ratios drop ∼ .
5, and their varianceis σ ∼ z spec < . F ν (90) /F ν (140) ratio. Same as the L (PAH)/ L (IR) ratio, we found no strong dependence ofSFR(Br α )/SFR(IR) on the F ν (90) /F ν (140) ratio. α To understand the behaviour of PAH and Br α emissionlines in detail, we show L (Br α ) to L (PAH) ratio againstthe L (IR) in Fig.6a. The L (Br α )/ L (PAH) ratio is consis-tent with that reported in Yano et al.(2016; grey brokenline) for ULIRGs whereas at around L (IR) ∼ L ⊙ theratio is > L (IR) increase is faster than the de-crease of Br α along the L (IR). We see only a few galaxiesin the bottom side of the figure, which could be a se-lection bias; as Br α emission is significantly fainter thanthe PAH 3.3 µ m emission, we may not detect the Br α emission from galaxies in this region. Galaxies with nodetection of Br α emission have log L(IR) of 10 − . L ⊙ .Assuming their Br α flux is < × − erg/s/cm (seeFig.2a), half of them would distribute at log L (IR)/L ⊙ < L (Br α )/ L (PAH) < -1.3, and the other half woulddistribute at log L (IR)/L ⊙ >
12 and log L (Br α )/ L (PAH) <
0. In Fig.6b, we compare L (Br α )/ L (IR) and L (PAH)/ L (IR) of LIRGs, ULIRGs, and the othersample. We can see again from this figure that galax-ies with higher L (IR) have lower L (Br α )/ L (IR) and L (PAH)/ L (IR) ratios. The figure shows a clear correla-tion, where the correlation coefficient is 0.703. The clearcorrelation between these ratios implies that the originsof low L (PAH)/ L (IR) and L (Br α )/ L (IR) are the same,or at least, related to each other. The figure also showsthat our LIRG sample has lower L (Br α )/ L (IR) ratio thanthat reported from Yano et al.(2016) for ULIRG sample.It implies destruction of PAH molecules in ULIRGs (seealso section 5.4). In the previous sections, we found that both L (PAH)/ L (IR) and L (Br α )/ L (IR) are lower at higher L (IR), and they correlate with each other. The relativeweakness of the two ratios and their correlation can beinterpreted in the following four ways.(a) Large extinction of PAH and/or Br α emissions:Despite the long wavelength of these emissions, dust ex-tinction could not be negligible. In the galactic extinc-tion curve the k λ is 0.286 at 3.29 µ m and 0.253 at 4.052 µ m, respectively. It is 0.08 ∼ ∼
30 mag, both L (PAH)/ L (IR) and L (Br α )/ L (IR)could be lower by one order of magnitude.If this is the case we have to note two things; First,a significant amount of dust should be in or around thePDR because PAH molecules are in the PDR. Second,the most energy is likely to come from the very centre ofgalaxies because extremely high dust attenuation in theentire region is not likely.(b) Destruction of PAH molecules: Harsh radiationfrom an active galactic nucleus (AGN) or strong star-o. ] PAH, Br α , and LIR of local galaxies observed with AKARI -5-4.5-4-3.5-3-2.5-2 9.5 10 10.5 11 11.5 12 12.5 13 l og L ( P AH ) / L (I R ) log L(IR)/L Sun (a)AGNLINERSF -1.5-1-0.5 0 0.5 1 9.5 10 10.5 11 11.5 12 12.5 13 l og [ SF R ( B r α ) / SF R (I R )] log L(IR)/L Sun (b)AGNLINERSF-4.4-4.2-4-3.8-3.6-3.4-3.2-3-2.8-2.6 10 10.5 11 11.5 12 12.5 l og L ( P AH ) / L (I R ) log L(IR) / L Sun (c; median)UnknownAGNLINERSF -1.2-1-0.8-0.6-0.4-0.2 0 0.2 0.4 10 10.5 11 11.5 12 12.5 l og SF R ( B r α ) / SF R (I R ) log L(IR)/L Sun (d;median)UnknownAGNLINERSF-5-4.5-4-3.5-3-2.5-2 9.5 10 10.5 11 11.5 12 12.5 l og L ( P AH ) / L (I R ) log L(IR)/L Sun (e)F90/F140 > 1F90/F140 < 1 -1.5-1-0.5 0 0.5 1 9.5 10 10.5 11 11.5 12 12.5 l og [ SF R ( B r α ) / SF R (I R )] log L(IR)/L Sun (f)F90/F140 > 1F90/F140 < 1
Fig. 5. (a): Ratio between PAH 3.3 µ m luminosity to L (IR) against the L (IR). The sample is divided into four type: SF (blue),LINER (green), AGN (red), and unknown (grey). (b): Ratio between SFR(Br α ) to SFR(IR) against the L (IR). (c): Medians of the L (PAH)/ L (IR) ratio for each galaxy type. (d): Same as the panel (c) but for SFR(Br α )/SFR(IR) ratio. (e): Same as the panel(a) but the sample is divided with F ν (90) /F ν (140). (f): Same as the panel (b) but the sample is divided with F ν (90) /F ν (140). burst may destroy them (Nordon et al. 2012; Murata etal. 2014). This can explain only lower L (PAH)/ L (IR) ra-tios. Since our ULIRG sample shows lower L (PAH)/ L (IR)ratios than those of LIRG sample, destruction ofPAH molecules may be more effective in ULIRGs.Nonetheless, the correlation between L (Br α )/ L (IR) and L (PAH)/ L (IR) ratios indicates other phenomenon tomake L (Br α )/ L (IR) and L (PAH)/ L (IR) ratios lower.(c) Lack of UV photon ionising hydrogen gas and excit-ing PAH molecules: If UV photons produced from youngstars are absorbed by dust before it ionises gas and excitesPAH molecules, both the hydrogen recombination lines and PAH emission is weakened (Vald´es et al. 2005; Murataet al. 2014; Yano et al. 2016), which results in the drop of L (PAH)/ L (IR) and the L ( Brα )/ L (IR).To explain the observed (minimum)SFR(Br α )/SFR(IR) < γ and Pa α lines. They showed that the fraction shouldbe ∼
80% for explaining observed ratios between L (Br γ )or L (Pa α ) to L (IR). Yano et al. (2016) also suggestedthat the fraction is, on average, ∼
45% for 13 local HII0 K. Murata et al. [Vol. , -2.5-2-1.5-1-0.5 0 0.5 9.5 10 10.5 11 11.5 12 12.5 13 l og L ( B r α ) / L ( P AH ) log L(IR)/L Sun (a) Yano et al.2016AGNLINERSF-5.5-5-4.5-4-3.5-3-4.5 -4 -3.5 -3 -2.5 -2 l og [ L ( B r α ) / L (I R )] log [L(PAH 3.3 µ m)/L(IR)] (b)LIRGULIRGOther Fig. 6.
Comparison between Br α and PAH luminosity. (a): Ratio between them against the L (IR). A ratio estimatedfor local ULIRGs in Yano et al. (2016) is also shown in the grey broken line. The colour code is the same as Fig.5a.(b): Both luminosity are normalised with their L (IR). They are classified with their L (IR): LIRGs (magenta), ULIRGs(cyan), and fainter than LIRGs (black). The grey broken line indicates, again, the ratio estimated in Yano et al. (2016). ULIRGs using a ratio between Br α and L (IR). This valueis lower than ours, which could be due to their smallsample size.(d) AGN contribution to L (IR): If AGNs significantlycontribute their energy to the L (IR) and do not to the L (Br α ) and L (PAH), we expect a lower L (Br α )/ L (IR)and L (PAH)/ L (IR) ratios than those of normal star-forming galaxies.We found no strong dependence of L (PAH)/ L (IR) and L (Br α )/ L (IR) on galaxy type. It implies the AGN contri-bution is not a dominant cause. This is inconsistent withYano et al. (2016) who showed that the L (Br α )/ L (IR)ratio depends on the galaxy type; the ratio of HII galax-ies is significantly higher than that of LINER and Seyfertgalaxies. This inconsistency could be due to the samplesize; we have as many as 230 classified galaxies while Yanoet al. (2016) have only ∼
30 galaxies. Hence we concludethat our result is more robust.Nonetheless, we have to consider obscured AGNs.Imanishi et al. (2010) showed that even HII galaxies couldhave an AGN. If it significantly contributes the energy into L (IR), both L (PAH)/ L (IR) and L (Br α )/ L (IR) could belower. Symeonidis et al. (2016) showed that the intrin- sic far-infrared emission from QSOs is higher than previ-ously known. They implied that contribution from a pow-erful AGN to any broad band flux cannot be neglectedfor estimating the SFR of the host galaxy. Hence, al-though we found no dependence of L (PAH)/ L (IR) and L (Br α )/ L (IR) on galaxy type, we cannot reject the sce-nario (d).
6. Summary