Spectral decomposition of starbursts and AGNs in 5-8 micron Spitzer IRS spectra of local ULIRGs
E. Nardini G. Risaliti, M. Salvati, E. Sani, M. Imanishi, A. Marconi, R. Maiolino
aa r X i v : . [ a s t r o - ph ] J a n Mon. Not. R. Astron. Soc. , 1–5 (2002) Printed 16 December 2018 (MN L A TEX style file v2.2)
Spectral decomposition of starbursts and AGNs in 5–8 µ m Spitzer
IRS spectra of local ULIRGs
E. Nardini, G. Risaliti, , M. Salvati, E. Sani, M. Imanishi, A. Marconi and R. Maiolino Dipartimento di Astronomia, Universit`a di Firenze, L.go E. Fermi 2, 50125 Firenze, Italy. E-mail: [email protected] INAF - Osservatorio Astrofisico di Arcetri, L.go E. Fermi 5, 50125 Firenze, Italy Harvard-Smithsonian Center for Astrophysics, 60 Garden St. Cambridge, MA 02138 USA National Astronomical Observatory, 2-21-1, Osawa, Mitaka, Tokyo 181-8588, Japan INAF - Osservatorio Astronomico di Roma, via di Frascati 33, 00040 Monte Porzio Catone (RM), Italy
Released Xxxx Xxxxx XX
ABSTRACT
We present an analysis of the 5–8 µ m Spitzer -IRS spectra of a sample of 68 localUltraluminous Infrared Galaxies (ULIRGs). Our diagnostic technique allows a clearseparation of the active galactic nucleus (AGN) and starburst (SB) components inthe observed mid-IR emission, and a simple analytic model provides a quantitative estimate of the AGN/starburst contribution to the bolometric luminosity. We showthat AGNs are ∼
30 times brighter at 6 µ m than starbursts with the same bolomet-ric luminosity, so that even faint AGNs can be detected. Star formation events areconfirmed as the dominant power source for extreme infrared activity, since ∼
85% ofULIRG luminosity arises from the SB component. Nonetheless an AGN is present inthe majority (46/68) of our sources.
Key words: galaxies: active; galaxies: starburst; infrared: galaxies.
Ultraluminous Infrared Galaxies (ULIRGs, L IR > L ⊙ )are the local counterparts of the high-redshift objects dom-inating the cosmic background in the far-infrared and milli-metric bands. Unveiling the nature of their energy source isfundamental in order to understand the star formation his-tory and the obscured AGN activity in the distant Universe.Since their discovery, several infrared indicators havebeen proposed to determine whether the central engine inULIRGs is an AGN or a starburst (SB). The presence ofhigh-ionization lines in the mid-IR spectra of ULIRGs pointsto AGN activity, while intense PAH emission features aretypical of starburst environments (Genzel et al. 1998; Lau-rent et al. 2000). Recently, the absorption feature of amor-phous silicate grains centered at 9.7 µ m has also been usedtogether with the PAH emission to assess the nature of theobscured power source (Spoon et al. 2007). An alternate wayto disentangle AGNs and SBs in ULIRGs has been proposedby Risaliti et al. (2006, hereafter R06), based on the sepa-ration of the two continuum components in 3–4 µ m spectra.This method has been successfully applied to a sample of ∼
50 nearby ULIRGs (Risaliti, Imanishi & Sani 2007, sub-mitted ) and provided an estimate of the average AGN/SBcontribution to ULIRGs. The key reason for using the con-tinuum emission at λ ≃ µ m as a diagnostic is the differ- ence between the 3- µ m to bolometric ratios in AGNs andstarbursts ( ∼ two orders of magnitude larger in the former).This makes the detection of the AGN component possibleeven when the AGN is heavily obscured and/or bolometri-cally weak compared to the starburst. However the originalprescription is limited by the low quality of the available L-band spectra of ULIRGs (R06, Imanishi et al. 2006), whichmakes the results on individual sources highly uncertain,except for the ∼ µ mspectral band, using the observations of the IRS instrument(Houck et al. 2004) onboard Spitzer . We disentangled theAGN and SB contributions to the observed 5–8 µ m emissionof ULIRGs by combining average spectral templates repre-senting the different properties of the two physical processesat work. The high quality of Spitzer -IRS data, in addition tothe relatively low dispersion of the intrinsic continuum prop-erties of both AGNs and starbursts in this spectral range,allows a much more accurate determination of the AGN/SBcomponents than possible at other wavelengths (e.g. X-rays)or with other diagnostic methods based on emission lines.In this paper we present our decomposition method, anddiscuss a simple analytical model providing a quantitative estimate of the AGN/SB contribution to the bolometric lu-minosity of each source. c (cid:13) E. Nardini et al.
In order to perform a detailed study of a representative sam-ple of ULIRGs in the local Universe, we selected 68 sourceswith z < .
15 and a 60- µ m flux density f > µ m ensures an unbiased selection with respect to therelative AGN/SB contributions.IRS observations were obtained within three differ-ent programs: PID 105 (PI J.R.Houck), PID 2306 (PIM.Imanishi), PID 3187 (PI S.Veilleux). The coadded im-ages provided by the Spitzer Science Center (after the treat-ment with pipeline version S13.0) have been background-subtracted by differencing the two observations in the nod-ding cycle. The spectra have been extracted and calibratedfollowing the standard procedure for point-like sources withthe package
SPICE . The flux uncertainties have been esti-mated from source and background counts (in e − /s ). Fi-nally, we performed a smooth connection between the Short-Low spectral orders, with no necessity of relative scaling.Out of the 68 spectra, we already published 48 in Iman-ishi et al. 2007. Six more spectra are shown in Armus etal. 2006, 2007. The remaining 14 spectra are analyzed herefor the first time and will be fully presented in a forthcomingpaper (Nardini et al. 2008, in prep. ). µ m AGN/SB SEPARATION Despite the diversity of the global IRS spectra of pure AGNsand pure SBs, and the complexity of the physics involved, lit-tle dispersion is seen at wavelengths shortward of the 9.7 µ msilicate feature. This makes possible the use of universalAGN/SB templates to reproduce the spectral properties ofULIRGs in the 5–8 µ m interval. In the following we describethe templates adopted in our model. Starburst.
The mid-IR spectral features of local SBgalaxies show very little variations from one object to an-other in the 5–8 µ m wavelength range, while larger differ-ences are present in the ∼ µ m band (Brandl et al. 2006,hereafter B06). In order to check if this is the case in theULIRG luminosity range as well, we analyzed the sourcesin our sample estabilished to be starburst-dominated bymulti-band studies. We did not find significant variationsamong the spectra, concluding that a fixed template canbe used to represent the 5–8 µ m SB component in localULIRGs. We built such template using the five brighest ob-jects among the pure SBs in our sample (IRAS 10190+1322,IRAS 12112+0305, IRAS 17208 − − − µ mbefore adding. Our SB template is shown in Fig.1, togetherwith its dispersion in the whole IRS spectral band and theB06 template. The little spectral dispersion below 8 µ mamong SBs of different luminosity (to be compared withtheir large differences at longer wavelengths) is in itself aninteresting result, which should be fully investigated throughdetailed emission and radiative transfer models. Concerningthis we only notice that such a remarkable similarity canresult from the spatial integration over a large number ofindividual star-forming regions. Figure 1.
Comparison between the SB template of B06 ( dashedred line ) and our template ( solid blue line ), constructed from theemission of the 5 brightest SB-dominated ULIRGs. The shadedarea shows the σ rms dispersion in the 5 ULIRG spectra. Thevertical long-dashed green lines enclose the fitting region. AGN.
Our recent L-band analysis of bright ULIRGsshows that the intrinsic AGN emission is due to hot dustgrains and the flux density is well described by a featurelesspower law with a fixed spectral index: f ν ∝ λ . . Here weadopt the same spectral shape up to 8 µ m, in agreement withnew Spitzer observations of a large sample of local type 1quasars (Netzer et al. 2007).An active nucleus is much more compact than a cir-cumnuclear starburst region. As a consequence, the near-IRradiation due to thin dust reprocessing can be itself stronglyreddened because of a compact absorber along the line ofsight. We therefore introduce an exponential attenuationfactor e − τ ( λ ) , where the optical depth follows the conven-tional law τ ( λ ) ∝ λ − . (Draine 1989). A similar correc-tion is not needed in the SB template. We stress that thisdoes not imply that the starburst spectrum is not affectedby inner obscuration; the possible effects of this obscuration(which are clear at longer wavelengths, e.g. in the silicate ab-sorption features at 9.7 and ∼ µ m) are however alreadyaccounted for in the adopted observational template.Summarizing, the different contributions to the ob-served energy output of a ULIRG can be parametrized asfollows: f obs ν ( λ ) = f int (cid:2) (1 − α ) u sb ν + α u agn ν e − τ ( λ ) (cid:3) (1)where α is the AGN contribution to the 5–8 µ m intrinsicflux density f int , while u sb ν and u agn ν are the SB and AGNtemplates normalized at 6 µ m. Apart from the flux normal-ization, our model contains only two free parameters, i.e. α and the optical depth to the AGN τ (6 µ m). They are bothshown in Tab.1.Additional high-ionization emission lines and molecu-lar absorption features (due to ices and aliphatic hydrocar-bons), whenever present, were fitted by means of gaussianprofiles except for the water ice absorption at ∼ µ m, repro-duced with the laboratory profile from the Leiden databasecorresponding to pure H O ice at 30 K. c (cid:13) , 1–5 GN and starburst in ULIRGs Figure 2.
Three representative examples of the typical 5–8 µ mspectral shapes of ULIRGs. The differences among the spectraare entirely due to the different AGN contribution and its ob-scuration. Whenever the AGN is the dominant power source, asin Mrk 231, a strong continuum almost obliterates the PAH fea-tures. On the contrary, in Mrk 273 the AGN is fainter and thespectral outline of a starburst is clearly identified. A similar spec-trum is exhibited by IRAS 20551 − green filledcircles ) and their best fits ( black thin line ), we have included thereddened AGN ( red dot-dashed line ) and starburst ( blue dottedline ) components. Fig.2 shows the spectral decomposition of three rep-resentative ULIRGs. In spite of the great diversity of theobserved spectra, our simple model provides a good fit ofeach spectrum in the sample: the residuals from the bestfits are in all sources smaller than 10% at all wavelengths(though the fits are not formally acceptable in a statisticalsense, with a reduced χ &
2, due to the small error barsand the remaining unfitted minor features). In particular,both the PAH emission and the continuum are always wellreproduced: this implies that the large variations in the 5–8 µ m spectral shape of ULIRGs are entirely due to the AGNcontribution and its obscuration. A detailed analysis of theresults for each source and a physical interpretation of pe-culiar cases are the subject of a forthcoming paper (Nardiniet al. 2008, in prep. ). The large difference between the 5–8 µ m to bolometric ratiosin AGNs and SBs implies that this ratio is itself an indicator of AGN activity, and can be used (a) to test the consistencyof our decomposition method and (b) to estimate the relativeAGN and SB contributions to the bolometric luminosity ofour sample. We define the 6- µ m to bolometric ratio as: R = (cid:18) ν f int F IR (cid:19) . (2)where F IR is the total infrared flux, estimated as in Sanders& Mirabel (1996). Since the integrated infrared luminosity ofULIRGs coincides almost exactly with their bolometric lu-minosity, R is a fair approximation to the fraction of the to-tal energy output that is intrinsically emitted in the 5–8 µ mrange. Reminding that the intrinsic AGN/SB contributionsare α f int and (1 − α ) f int respectively, and decomposing F IR as F agnIR + F sbIR , a simple connection between R and α is brought out: R = R agn R sb α R sb + (1 − α ) R agn , (3)provided that R agn and R sb , the equivalents of R for pure(unobscured) AGNs and pure SBs, are defined as in Eq.2.We fitted the theoretical R ( α ) relation (Eq.3) to our dataconsidering R agn and R sb as free parameters, and found:log R agn = − . +0 . − . and log R sb = − . +0 . − . . (4)We note that R agn turns out to be somewhat higher than tra-ditional estimates based on AGN spectral energy distribu-tions: for example, we derive log R agn ∼ − . ∼
30 times more luminousat 6 µ m than starbursts with the same bolometric lumi-nosity. We are now able to quantify the AGN contribution( α bol = F agnIR /F IR ) to the total infrared luminosity of eachsource: α bol = α α + ( R agn /R sb )(1 − α ) , (5)where R agn /R sb ≃
28. The values of α bol are listed inTab.1. Our estimates for the ∼
15 brightest sources are ingood agreement with those of R06 and with the Genzel etal. (1998) and Laurent et al. (2000) mid-IR diagnostic di-agrams. Considering the whole sample, our results can becompared with the optical classification and with L-bandand hard X-ray studies, when available. A substantial agree-ment is obtained in all cases. It is worth noticing that theoptical classification alone gives incomplete information: allthe sources classified as Seyferts show clear traces of AGNactivity, but 7 out of 8 among the ULIRGs with α bol > . τ > ii regions.LINERs are again confirmed to be rather heterogeneous withrespect to the nature of their energy source. Such ambigui-ties can be solved by applying our diagnostic.By inverting Eq.5 the relation between R and α bol takesthe neat form R = α bol R agn + (1 − α bol ) R sb , and is plottedin Fig.3a. As a final check we have computed for each sourcethe following quantities: b R agn = (cid:18) ν α f int α bol F IR (cid:19) and b R sb = (cid:20) ν (1 − α ) f int (1 − α bol ) F IR (cid:21) . (6) c (cid:13) , 1–5 E. Nardini et al.
Table 1.
Spectral parameters for the 68 sources in our sample. α : AGN contribution to the intrinsic continuum emission at 6 µ m(in percent). τ : Optical depth of the AGN component at 6 µ m (we assume τ = 0 for the sources with no detected AGN). α bol : AGNcontribution to the bolometric luminosity (in percent). The errors in α bol are due to the statistical uncertainty both in the flux amplitudeof the AGN/SB components and in the ratios R agn and R sb . The systematic effects are discussed in the text. O/X/L : SB/AGN/LINERclassification based on optical, X-ray and L-band spectroscopy. A: AGN, L: LINER, A*: AGN, tentative detection. References: : Veilleuxet al. 1999, : Veilleux et al. 1995, : Duc et al. 1997, : Iwasawa et al. 2005, : Severgnini et al. 2001, : Franceschini et al. 2003, : Balestraet al. 2005, : Vignati et al. 1999, : Imanishi et al. 2003, : Imanishi et al. 2006, : Risaliti et al. 2006, : Sani et al. 2007. † : Sourceswith simultaneous α bol > . τ > Source α τ α bol O/X/L
Source α τ α bol O/X/L
ARP 220 75 ± ± . +3 . − . L /SB /SB IRAS 14197+0813 75 ± ± +5 − –/–/–IRAS 00091 − † ± ± +8 − SB /–/– IRAS 14252 − < < . < . /–/SB IRAS 00188 − ± ± ± /–/A* IRAS 14348 − +3 − < .
09 3 . +2 . − . L /–/SB IRAS 00456 − < . < .
04 SB /–/– IRAS 15130 − +1 − < .
01 28 +9 − A /–/A IRAS 00482 − < < . < . /–/– IRAS 15206+3342 52 ± ± . +1 . − . SB /–/SB IRAS 01003 − † ± ± ± /–/– IRAS 15225+2350 89 ± ± +7 − SB /–/SB IRAS 01166 − ± ± +7 − SB /–/– IRAS 15250+3609 94 ± ± +9 − L /SB /–IRAS 01298 − † ± ± +8 − SB /–/– IRAS 15462 − +1 − < .
01 25 +10 − A /–/–IRAS 01569 − ± ± +6 − SB /–/– IRAS 16090 − ± ± +7 − L /–/A* IRAS 02411+0353 < < . < . /–/– IRAS 16156+0146 90 ± ± +8 − A /–/–IRAS 03250+1606 < . < . < . /–/A* IRAS 16468+5200 85 ± ± +6 − L /–/SB IRAS 04103 − ± ± . +2 . − . L /–/– IRAS 16474+3430 < . < . < . /–/A* IRAS 05189 − +1 − < .
01 28 +8 − A /A /A IRAS 16487+5447 21 ± < .
04 1 . +0 . − . L /–/A* IRAS 08572+3915 99 ± ± +5 − L /–/A IRAS 17028+5817 < . < .
05 L /–/A* IRAS 09039+0503 61 ± ± . +2 . − . L /–/A* IRAS 17044+6720 91 ± ± +8 − L /–/A IRAS 09116+0334 < . < . < .
07 L /–/A* IRAS 17179+5444 84 ± ± +6 − A /–/A IRAS 09539+0857 † ± ± +8 − L /–/SB IRAS 17208 − < . < . < . /SB /SB IRAS 10190+1322 < . < .
02 SB /–/SB IRAS 19254 − ± ± +9 − A /A /A IRAS 10378+1109 73 +1 − < .
01 9 . +3 . − . L /–/A* IRAS 20100 − ± ± +6 − SB /A /SB IRAS 10485 − ± ± . +2 . − . L /–/A* IRAS 20414 − < . < . < .
09 SB /–/SB IRAS 10494+4424 < . < .
02 L /–/A* IRAS 20551 − † ± ± +8 − L /A /A IRAS 11095 − † ± ± +9 − L /–/SB IRAS 21208 − < . < .
04 SB /–/SB IRAS 11130 − ± ± +6 − L /–/– IRAS 21219 − +1 − < .
01 83 +13 − A /–/A IRAS 11387+4116 < . < .
03 SB /–/SB IRAS 21329 − ± < .
07 2 . +1 . − . L /–/A* IRAS 11506+1331 54 +1 − < .
01 4 . +1 . − . SB /–/A* IRAS 22206 − < . < . < . /–/–IRAS 12072 − † ± ± +9 − A /–/A IRAS 22491 − < . < .
06 SB /SB /–IRAS 12112+0305 < < . < . /SB /SB IRAS 23128 − ± ± . +1 . − . L /A /A IRAS 12127 − ± ± +15 − L /–/A IRAS 23234+0946 30 +2 − < .
05 1 . +0 . − . L /–/SB IRAS 12359 − +2 − < .
01 4 . +2 . − . L /–/A* IRAS 23327+2913 73 ± ± . +3 . − . L /–/SB IRAS 13335 − < . < .
02 L /–/– MRK 231 93 ± < .
12 32 +8 − A /A /A IRAS 13454 − +1 − < .
01 5 . +2 . − . A /–/– MRK 273 67 +1 − < .
01 7 . +2 . − . A /A /A IRAS 13509+0442 < . < .
03 SB /–/SB NGC 6240 65 +6 − ± . +4 . − . L /A /A IRAS 13539+2920 < . < .
02 SB /–/SB
4C +12.50 97 ± ± +10 − A /–/–IRAS 14060+2919 < . < .
02 SB /–/SB UGC 5101 † ± ± +9 − L /A /A The results are shown in Fig.3b and prove that our de-composition method is reliable in estimating the AGN/SBcontributions both to the 6- µ m and to the bolometric lu-minosity of local ULIRGs. In fact, the estimated 6- µ m tobolometric ratios of the SB component in the compositesources (which are located at the bottom right of the plotand are heavily dependent on our modeling) are fully consis-tent, within the errors, with the ratios of the pure starbursts(located at the bottom left and directly computed from themeasured 6- µ m and IRAS fluxes). This success is promis-ing in anticipation of a forthcoming study about the roleof black hole accretion and star formation in the intenseinfrared activity characterising the distant galaxies.However, it is important to keep in mind the main lim-itations of our approach:1) The narrow wavelength range used in this work simplifiesthe decomposition analysis, but prevents us from a completestudy of the dust composition, density and geometrical dis- tribution. These elements strongly affect the overall mid-IRemission longward of the silicate absorption feature, and canbe investigated only through an analysis of the whole IRSspectrum.2) While the 5–8 µ m templates seem to have little dispersion(as discussed in detail in Section 3), the spread in the 6- µ mto bolometric ratios R agn and R sb can be much higher, mak-ing our estimates of the AGN/SB bolometric fractions moreuncertain than those in the 5–8 µ m band. The uncertaintiesin α bol reported in Tab.1 are obtained assuming the meanratios, with the errors on the mean given in Eq.4. However,the dispersion around the best fit in Fig.3a is significantlylarger. We therefore consider this dispersion (0.3 dex, con-stant at all values of α bol ) as the actual uncertainty in thebolometric ratios of the individual sources. The numericalresults on the individuale sources are anyway precise enoughto estabilish which is the dominant source of the observedluminosity. c (cid:13) , 1–5 GN and starburst in ULIRGs Figure 3. (a)
Ratio R between absorption-corrected 6- µ m lu-minosity and bolometric luminosity, versus the AGN bolometriccontribution α bol . The error bars of R are due to the uncertaintiesin the total infrared flux F IR and in the intrinsic AGN fraction, α . The solid line is the best fit of the R – α relation from Eq.3(plotted as a function of α bol using Eq.5). (b) Same as above,with the AGN and SB components plotted separately.
Overall, if we consider as a confidence limit the value α bol = 0 .
01 (i.e. the dispersion around our starburst tem-plate), an AGN is present in 46 of the 68 ULIRGs in oursample (including several of those optically classified as H ii regions), but it is significant ( α bol & .
25) only in ∼
30% ofthe cases. The SB process is responsible for almost 90% ofthe observed infrared luminosity of ULIRGs, with no signif-icant (i.e. > µ m Spitzer
MIPS surveys. IRS spec-troscopy shows that they are mostly z ∼ ∼
30 times higherAGN relative emission in the 5–8 µ m rest-frame wavelengthrange we have pointed out. The use of average templates for AGN and SB emission hasallowed us to disentangle the two components in the 5–8 µ m spectra of 68 local ULIRGs, observed with the Spitzer SpaceTelescope . We have been able to detect an AGN in morethan 60% of our sources, and estimate its contribution tothe bolometric luminosity. In a statistical sense, we con-firm that local Ultraluminous Infrared Galaxies are poweredfor ∼
85% by intense star formation and for the remaining ∼
15% by AGN activity. Our method proves to be successfulin unveiling an intrinsically faint or obscured AGN inside aULIRG. In this context we also put on a sound basis ourinitial assumption that the wavelength interval 5–8 µ m isan appropriate spectral range in order to search for AGNs:an AGN turns out to be approximately 30 times more lu-minous at 6 µ m than a starburst with the same bolometricluminosity. ACKNOWLEDGMENTS
We are grateful to the anonymous referee for his/her help-ful and constructive comments. We acknowledge financialsupport from the prin-miur 2006025203 grant and the ASI-INAF grant I/023/05/0.
REFERENCES
Armus L., et al., 2006, ApJ, 640, 204Armus L., et al., 2007, ApJ, 656, 148Balestra I., Boller T., Gallo L., Lutz D., Hess S., 2005, A&A, 442,469Brandl B. R., et al., 2006, ApJ, 653, 1129 (B06)Draine B. T., 1989, ESASP, 290, 93Duc P.-A., Mirabel I. F., Maza J., 1997, A&AS, 124, 533Elvis M., et al., 1994, ApJS, 95, 1Franceschini A., et al., 2003, MNRAS, 343, 1181Genzel R., et al., 1998, ApJ, 498, 579Houck J. R., et al., 2004, ApJS, 154, 18Houck J. R., et al., 2005, ApJ, 622, L105Imanishi M., Terashima Y., Anabuki N., Nakagawa T., 2003, ApJ,596, L167Imanishi M., Dudley C. C., Maloney P. R., 2006, ApJ, 637, 114Imanishi M., Dudley C. C., Maiolino R., Maloney P. R., NakagawaT., Risaliti G., 2007, ApJS, 171, 72Iwasawa K., Sanders D. B., Evans A. S., Trentham N., MiniuttiG., Spoon H. W. W., 2005, MNRAS, 357, 565Kim D.-C., Sanders D. B., 1998, ApJS, 119, 41Laurent O., Mirabel I. F., Charmandaris V., Gallais P., MaddenS. C., Sauvage M., Vigroux L., Cesarsky C., 2000, A&A, 359,887Netzer H., et al., 2007, ApJ, 666, 806Risaliti G., et al., 2006, MNRAS, 365, 303 (R06)Sanders D. B., Mirabel I. F., 1996, ARA&A, 34, 749Sani E., et al., 2007, arXiv, 709, arXiv:0709.1344Severgnini P., Risaliti G., Marconi A., Maiolino R., Salvati M.,2001, A&A, 368, 44Spoon H. W. W., Marshall J. A., Houck J. R., Elitzur M., Hao L.,Armus L., Brandl B. R., Charmandaris V., 2007, ApJ, 654,L49Veilleux S., Kim D.-C., Sanders D. B., Mazzarella J. M., SoiferB. T., 1995, ApJS, 98, 171Veilleux S., Kim D.-C., Sanders D. B., 1999, ApJ, 522, 113Vignati P., et al., 1999, A&A, 349, L57c (cid:13)000