Luminous Infrared Galaxies

FIRSED April 3, 2013 The Great Observatories All-sky LIRG Survey (GOALS)

o • IRAS-detected galaxies with f60μm > 5.24 Jy & |b| > 5 11.0 • 201 galaxies with LIR ≥ 10 L⊙ • HST, Chandra ~ 88 LIRGs; Spitzer, Galex, Herschel ~ 201 LIRGs • http://goals.ipac.caltech.edu

FIRSED April 3, 2013 Luminous & Ultraluminous Infrared Galaxies LIRGs: L [8-1000 μm] ≥ 1011-11.99 L • IR ⊙; 12 ULIRGs: LIR ≥ 10 L⨀ • Spiral galaxies which show an increasing tendency to be involved in interactions or mergers with increasing luminosity • Rich in optically-visible star clusters

NGC 5257/8 Luminous & Ultraluminous Infrared Galaxies CO(1→0)

• Spiral galaxies which show an increasing tendency to be involved in interactions or mergers with increasing luminosity • Rich in optically-visible star clusters • Rich in dust, and both neutral and star-forming molecular gas

Contours = HI CO(1→0) Luminous & Ultraluminous Infrared Galaxies

AGN • Spiral galaxies which show an increasing tendency to be involved in interactions or mergers with increasing luminosity • Rich in optically-visible star clusters • Rich in dust, and both neutral and star-forming molecular gas Contours = HI Evidence of Active Galactic Nuclei Mrk 231 • (AGN) activity

March 14, 1998 9:20 Annual Reviews SANDER2 AR12-18

U/LIRGs and luminosity function 756 SANDERS & MIRABEL

–2 IRAS BGS IRAS 1Jy ULIGs Mrk Seyferts Mrk Starburst ⊗ cDs PG QSOs ) –4 Normal galaxies –1 bol M

–3 ⊗ ⊗

–6 log Φ (Mpc

–8

–10 9 10 11 12 13 log Lbol (L. )

Figure 1 The luminosity function for infrared galaxies compared with other extragalactic objects. • The local spaceReferences: densityIRAS RBGS (Sanders of etLIRG al 1996a), IRAS is1-Jy low Survey of ULIGs (Kim 1995), Palomar- Green QSOs (Schmidt & Green 1983), Markarian starbursts and Seyfert galaxies (Huchra 1977), 11.4 and overtakesand normal that galaxies of (Schechter normal 1976). Determination galaxies of the bolometric beyond luminosity for thea opticallyLbol ~ 10 L⦿ • selected samples was as described in Soifer et al (1986), except for the adoption of a more accurate bolometric correction for QSOs of 11.8 ⌫L⌫ (0.43 µm) (Elvis et al 1994). ⇥

comes from ULIGs at flux levels f60 1–2 Jy corresponding to sources at z > 0.13. No evidence for evolution is found= for the subsample of 2 Jy ULIGs, i.e.⇠ n 3.8 3 (Kim & Sanders 1996). These results appear to be consistent with previous= ± debates in the literature which find n 5.6–7 for redshift surveys with flux limits f 0.5 Jy (Saunders et al 1990, Oliver⇠ et al 1995) but only n 3–4 60 ⇠ ⇠ for surveys with flux limits f60 > 1.5 Jy (Fisher et al 1992), and with analyses of IRAS extragalactic source counts⇠ (Hacking et al 1987, Lonsdale & Hacking 1989, Lonsdale et al 1990, Gregorich et al 1995) that show evidence for strong evolution only at relatively low flux levels ( f60 < 1 Jy). More definitive tests of ⇠ U/LIRGs and the Main Sequence

Starbursts

Main Sequence

11.4 ♁ L IR > 10 L⊙ 11.4 + LIR ≤ 10 L⊙

• The high end luminosity of the LIRG population lie above the main sequence with SFR ~ 100s... • ... if the IR is tracing only star formation Ultraluminous IR Galaxies & the evolution of QSOs?

Gas-rich Disk Progenitors with seed SMBH (Sanders et al. 1988) Ultraluminous IR Galaxies & the evolution of QSOs?

Enhanced Star Formation; SMBH building (Sanders et al. 1988) Ultraluminous IR Galaxies & the evolution of QSOs?

Optically luminous QSO phase (Sanders et al. 1988) Evolutionary byproduct

Elliptical/SO Galaxy (e.g., Barnes & Hernquist 1992) Starburst/AGN; Dust-clearing; Gas-rich Dust-enshrouded AGN-dominant Elliptical Progenitors phase; phase; Byproduct cool IR colors warm IR colors SED evolution: Cool → Warm

(f25μm/f60μm < 0.20) (f25μm/f60μm ≥ 0.20)

Dust-enshrouded Dust-clearing phase phase The importance of IR for LIRG studies

• νIRLIR / νoptLopt ↑ as Lbol ↑ (Sanders & Mirabel 1996): see Vivian U’s talk for an update –29–

The importance of IR for LIRG studies

• The SFR(UV) vs. SFR(IR) discrepancy grows with increasing LIR (Howell et al. 2010) Fig. 2.— Left: Histogram showing the ratio of SFR(UV) to SFR(UV+IR). The solid line is the full GOALS GALEX sample. Colored lines show the GOALS GALEX sample divided into luminosity bins as in Fig. 1: 11 11.4 11.4 11.8 11.8 10 < LIR < 10 L! (red dotted line), 10 < LIR < 10 L! (green dashed line), and LIR > 10 L! (blue dashed line). The FUV contribution to SFR is small for (U)LIRGs and decreases with increasing LIR. Right: LIR plotted against the ratio of SFR(UV) to SFR(UV+IR). Median ratios of the star formation 11 11.4 11.4 11.8 rates are shown for each luminosity bin (red: 10 < LIR < 10 L!,green:10 < LIR < 10 L!,blue: 11.8 LIR > 10 L!)alongwith1σ standard deviations of the mean. Although anticorrelated (Spearman rank correlation coeﬃcient of -0.47) the correlation is not linear. An imaging census of light from a LIRG HST Far-UV • Unobscured massive star formation

5 kpc (Inami et al. 2010) HST Optical • Unobscured star formation

5 kpc (Kim et al. 2013) 1.6μm HST NIR • embedded NRG sources + old stellar population

5 kpc (Inami et al. 2010) 70 INAMI ET AL. Vol. 140

10.00 IIZw096 source D - IRS SL + LL

1.00

0.10 Flux density (Jy)

0.01

5 10 15 20 25 30 Rest wavelength (microns) Spitzer mid-IR [NeII] 0.4 IIZw096 source D - IRS SL only 0.3 PAH Embedded Star 0.2 • Flux density (Jy) [SIV] 0.1 Formation &

0.0 6 8 10 12 14 AGN Rest wavelength (microns) Figure 6. Spitzer IRS spectra of source D (see Figure 3). The full low-resolution spectrum is at the top left, while a close-up of the SL spectrum is shown at the bottom left. The SH and LH spectra are shown at the top right and the bottom right, respectively. SomeDust residual order tiltingemission is evident in the two reddest ordersofLH.All spectra are shown in the rest frame. • originates from source C+D. When this is done, we derive a

ratio of FPAHtot /FIR 0.02. Since the continuum emission at the overlap wavelengths∼ of SL and LL has a difference less than 10%, dust emission in our low-res spectrum is dominated by source C+D.

3.5. AKARI Near-infrared Spectroscopy The spectrum of sources C and D taken with AKARI is well separated from that of the other sources (A and B) in II Zw 096, although they are all in the Np aperture, because the direction of spectral dispersion is nearly perpendicular to the vector separating C, D, and A. The AKARI near-infrared spectrum of source C+D is shown in Figure 7.Strong3.3 µm PAH emission, 3.4 µmemissionfromaliphatichydrocarbon grains, and Brα emission lines are detected. The line fluxes are (2.37 0.14) 10 16 Wm 2 and (0.50 0.05) 10 16 Wm 2 ± × − − ± × − − for the PAH and the Brα,respectively,afterfittingbothwith Figure 7. AKARI near-infrared spectrum of sources C and D. The wavelength aGaussianfunction(seeTables3 and 4). The EQW of the scale is given in the rest frame. 3.3 µmPAHis0.14 µm. Because the AKARI spectral resolution between 2.5and5µmisonlyR 120, the Brα recombination line is not resolved. ∼ For all clusters, we use a 0$$.1radiuscircularapertureforthe The spectrum of source A (not shown) also exhibits 3.3 µm HST photometric measurements. We estimate that the clusters contribute 14% of the total FUV luminosity ( 8 1036 W) PAH and Brα emission. The 3.3 µmPAHhasafluxof % ≈ × 16 2 in II Zw 096 (uncorrected for reddening). (1.77 0.12) 10− Wm− and an EQW of 0.08 µm. The Brα line± flux is× (0.33 0.06) 10 16 Wm 2. In Figure 8,wepresentanFUV–optical(F140LP F435W ± × − − versus F435W F814W) color–color diagram of the− clusters. − 3.6. Bright Star Clusters in the Remnant We additionally show the colors of a number of diffuse emission regions throughout the merger system. For the diffuse emission The high-resolution HST imaging data show that a large regions, we use 0$$.5–0$$.8radiuscircularapertures,dependingon number of bright star clusters surround the II Zw 096 nuclei the location in the FUV image to avoid crowded regions. and are present throughout the merging disks (see Figure 2). Two potential explanations for the observed color distribution 5 kpc From the B- and I-band images, we identify 128 clusters with are considered. First, if all the clusters are coeval, then their apparent magnitudes of 18.3mag ! mB ! 26.7mag(Vega) different colors result from variable extinction. In other words, and 17.9mag! mI ! 25.7mag(Vega)with80%completeness those clusters or diffuse emission regions which have colors of limits of 26.0 mag and 25.5 mag, respectively. Of these, 97 and F435W F814W 1.0magorF140LP F435W 0.0mag − " − " 88 are detected in the FUV and H-band images, respectively. are heavily reddened(Inami (AV " 3 mag).et Whenal. we2010) de-redden an 70 INAMI ET AL. Vol. 140

10.00 SpitzerIIZw096 source D - IRSmid-IR SL + LL •1.00Embedded Star 0.10

Flux density (Jy) Formation &

0.01AGN Dust5 emission10 15 20 25 30 • Rest wavelength (microns)

[NeII] 0.4 IIZw096 source D - IRS SL only

0.3 PAH

0.2 Flux density (Jy) [SIV] 0.1

0.0 70 INAMI ET AL. 6 8 10 12 14 Vol. 140 Rest wavelength (microns) 10.00 Figure 6. Spitzer IRS spectra of source D (see Figure 3). The full low-resolution spectrum is at the top left, while a close-up of the SL spectrum is shown at the bottom IIZw096 source D - IRS SL +left. LL The SH and LH spectra are shown at the top right and the bottom right, respectively. Some residual order tilting is evident in the two reddest ordersofLH.All spectra are shown in the rest frame. 1.00 originates from source C+D. When this is done, we derive a ratio of FPAH /FIR 0.02. Since the continuum emission at tot ∼ 0.10 the overlap wavelengths of SL and LL has a difference less than 10%, dust emission in our low-res spectrum is dominated by Flux density (Jy) source C+D. 0.01 3.5. AKARI Near-infrared Spectroscopy 5 kpc 5 10 15 20 25 30 Rest wavelength (microns)The spectrum of sources C and D taken with AKARI is well separated from that of the other sources (A and B) in

II Zw 096,[NeII] although they are all in(Inami the Np aperture, et al. because2010) 0.4 IIZw096 source D - IRS SL theonly direction of spectral dispersion is nearly perpendicular to the vector separating C, D, and A. The AKARI near-infrared 0.3 PAH spectrum of source C+D is shown in Figure 7.Strong3.3 µm PAH emission, 3.4 µmemissionfromaliphatichydrocarbon 0.2 grains, and Brα emission lines are detected. The line fluxes are 16 2 16 2 (2.37 0.14) 10− Wm− and (0.50 0.05) 10− Wm− Flux density (Jy) [SIV] ± × ± × 0.1 for the PAH and the Brα,respectively,afterfittingbothwith Figure 7. AKARI near-infrared spectrum of sources C and D. The wavelength aGaussianfunction(seeTables3 and 4). The EQW of the scale is given in the rest frame. 0.0 3.3 µmPAHis0.14 µm. Because the AKARI spectral resolution between 2.5and5µmisonlyR 120, the Brα recombination 6 8 10 12 14 For all clusters, we use a 0.1radiuscircularapertureforthe Rest wavelength (microns)line is not resolved. ∼ $$ The spectrum of source A (not shown) also exhibits 3.3 µm HST photometric measurements. We estimate that the clusters Figure 6. Spitzer IRS spectra of source D (see Figure 3). The full low-resolution spectrum is at the top left, while a close-up of the SL spectrumcontribute is shown at the14% bottom of the total FUV luminosity ( 8 1036 W) left. The SH and LH spectra are shown at the top right and thePAH bottom and right, Br respectively.α emission. Some The residual 3.3 µ ordermPAHhasafluxof tilting is evident in the two reddest ordersofLH.All% ≈ × 16 2 in II Zw 096 (uncorrected for reddening). spectra are shown in the rest frame. (1.77 0.12) 10− Wm− and an EQW of 0.08 µm. The Brα line± flux is× (0.33 0.06) 10 16 Wm 2. In Figure 8,wepresentanFUV–optical(F140LP F435W ± × − − versus F435W F814W) color–color diagram of the− clusters. originates from source C+D. When this is done, we derive a − 3.6. Bright Star Clusters in the Remnant We additionally show the colors of a number of diffuse emission ratio of FPAHtot /FIR 0.02. Since the continuum emission at regions throughout the merger system. For the diffuse emission the overlap wavelengths∼ of SL and LL has a difference less than The high-resolution HST imaging data show that a large regions, we use 0$$.5–0$$.8radiuscircularapertures,dependingon 10%, dust emission in our low-res spectrum is dominatednumber of by bright star clusters surround the II Zw 096 nuclei the location in the FUV image to avoid crowded regions. source C+D. and are present throughout the merging disks (see Figure 2). Two potential explanations for the observed color distribution From the B- and I-band images, we identify 128 clusters with are considered. First, if all the clusters are coeval, then their 3.5. AKARI Near-infrared Spectroscopyapparent magnitudes of 18.3mag ! mB ! 26.7mag(Vega) different colors result from variable extinction. In other words, and 17.9mag! mI ! 25.7mag(Vega)with80%completeness those clusters or diffuse emission regions which have colors of The spectrum of sources C and D taken withlimitsAKARI of 26.0is mag and 25.5 mag, respectively. Of these, 97 and F435W F814W 1.0magorF140LP F435W 0.0mag − " − " well separated from that of the other sources (A88 and are detected B) in in the FUV and H-band images, respectively. are heavily reddened (AV " 3 mag). When we de-redden an II Zw 096, although they are all in the Np aperture, because the direction of spectral dispersion is nearly perpendicular to the vector separating C, D, and A. The AKARI near-infrared spectrum of source C+D is shown in Figure 7.Strong3.3 µm PAH emission, 3.4 µmemissionfromaliphatichydrocarbon grains, and Brα emission lines are detected. The line fluxes are (2.37 0.14) 10 16 Wm 2 and (0.50 0.05) 10 16 Wm 2 ± × − − ± × − − for the PAH and the Brα,respectively,afterfittingbothwith Figure 7. AKARI near-infrared spectrum of sources C and D. The wavelength aGaussianfunction(seeTables3 and 4). The EQW of the scale is given in the rest frame. 3.3 µmPAHis0.14 µm. Because the AKARI spectral resolution between 2.5and5µmisonlyR 120, the Brα recombination line is not resolved. ∼ For all clusters, we use a 0$$.1radiuscircularapertureforthe The spectrum of source A (not shown) also exhibits 3.3 µm HST photometric measurements. We estimate that the clusters contribute 14% of the total FUV luminosity ( 8 1036 W) PAH and Brα emission. The 3.3 µmPAHhasafluxof % ≈ × 16 2 in II Zw 096 (uncorrected for reddening). (1.77 0.12) 10− Wm− and an EQW of 0.08 µm. The Brα line± flux is× (0.33 0.06) 10 16 Wm 2. In Figure 8,wepresentanFUV–optical(F140LP F435W ± × − − versus F435W F814W) color–color diagram of the− clusters. − 3.6. Bright Star Clusters in the Remnant We additionally show the colors of a number of diffuse emission regions throughout the merger system. For the diffuse emission The high-resolution HST imaging data show that a large regions, we use 0$$.5–0$$.8radiuscircularapertures,dependingon number of bright star clusters surround the II Zw 096 nuclei the location in the FUV image to avoid crowded regions. and are present throughout the merging disks (see Figure 2). Two potential explanations for the observed color distribution From the B- and I-band images, we identify 128 clusters with are considered. First, if all the clusters are coeval, then their apparent magnitudes of 18.3mag ! mB ! 26.7mag(Vega) different colors result from variable extinction. In other words, and 17.9mag! mI ! 25.7mag(Vega)with80%completeness those clusters or diffuse emission regions which have colors of limits of 26.0 mag and 25.5 mag, respectively. Of these, 97 and F435W F814W 1.0magorF140LP F435W 0.0mag − " − " 88 are detected in the FUV and H-band images, respectively. are heavily reddened (AV " 3 mag). When we de-redden an K. Iwasawa et al.: C-GOALS survey

UGC8387 UGC8696 F14348−1447 F14378−3651 − 1 0 s − 2 cm 1 − 1 0 E E E 1 F F F erg keV − 14 11 / 10 E F 0.1 11 0.5 1 2 5 0.5 1 2 5 0.1 0.5 1 2 5 0.5 1 2 5 Energy (keV) Energy (keV) Energy (keV) Energy (keV) 0 VV340 N 10 VV705 N F15250+3608 UGC9913 0 1 E E E E F F F F 1 11 11 0.1 0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 Energy (keV) Energy (keV) Energy (keV) Energy (keV)

ESO69−IG6 N NGC6240 0 F17132+5313 0 F17207−0014 10 100 E E E E F F F F 11 11 1 10

0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 Energy (keV) Energy (keV) Energy (keV) Energy (keV)

F18293−3413 ESO593−IG8 10 F19297−0406 19542+1110 0 0 1

ChandraE X-ray E E E F F F F 1 Embedded Star 11

• 11 Formation & 0.5 2 5 20 0.1 0.1 AGN0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 Energy (keV) Energy (keV) Energy (keV) Energy (keV)

ESO286−IG19 21101+5810 ESO239−IG2 CGCG448−020 10 0 0 1 E E E E F F F F 1 11 11 0.1

0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 Energy (keV) Energy (keV) Energy (keV) Energy (keV)

0 F22491−1808 ESO148−IG2 ESO148−IG2 N ESO148−IG2 S 0 1 1 E E E E F F F F 5 kpc 11 11 (Inami et al. 2010) 0.1 0.1

0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 0.5 1 2 5 Energy (keV) Energy (keV) Energy (keV) Energy (keV)

Fig. 4. continued.

A106, page 11 of 50 JVLA • Embedded Star Formation & AGN

5 kpc (Barcos et al. 2013) ALMA 1.6μm CO(6→5) • Dust continuum distribution • star-forming molecular gas tracers

Cycle 0 Stay Tuned

(Stierwalt et al. 2013) Isolated System? Mid-Stage - progenitors intact

NGC 5257/8 Mid-Stage - overlapping disks, double AGN + outﬂow

(Mazzarella et al. 2012) Late Stage - single nucleus, weak AGN NGC 2623

(Evans et al. 2008) The utility of IR/submm/radio emission

• assessment of the ionization mechanisms/source(s) in obscured regions and the optical depth towards these regions • assessment of the physical state in molecular clouds • measurement of the physical sizes of active regions and thus the IR surface brightness • measurement of the kinematics of the gas, and thus a assessment of rotation (e.g. for dynamical mass estimated) and importance of infall/outﬂow in obscured regions

...at wavelengths that are energetically important SED as a function of telescope λ coverage

JWST 11 10 10 10

ALMA

JVLA Spitzer

Herschel Mrk 231

Arp 220

•JWST, ALMA and JVLA provide sub-” resolution The utility of IR/submm/radio emission

1. Spitzer: AGN/Starburst diagnostics (see Veilleux’s talk) 2. Herschel: the [C II]/FIR decrement 3. Herschel: CO SLED 4. ALMA: CO(6→5) emission (1) AGN-Starburst diagnostics: ISO data high/low ionization → high/low

←destruction by AGN

(Genzel et al. 1998) AGN-Starburst diagnostics

ISO

(Genzel et al. 1998)

FIG. 2.ÈISOPHOT-S spectra (j/*j D 90) of 15 ULIRGs. Observed wavelengths are shown at the top and rest wavelengths at the bottom of each spectrum. Flux densities (not including the correction for chopped mode of a factor of D1.4) are in janskys. Error bars as provided by PIA(° 2.1) are shown for UGC 5101. AGN-Starburst diagnostics

ISO

Spitzer

(Genzel et al. 1998; Armus et al. 2007)

Fig. 1.—IRS Short-Low and Long-Low spectra of the 10 BGS ULIRGs in order of decreasing 25 m ﬂux density. Prominent emission features and absorption bands (the latter indicated by horizontal bars) are marked on representative spectra. Not all features are marked on all spectra; see Tables 3 and 4 for measured features. Expanded views of the 5Y14 m regions (Short-Low) of each spectrum are shown at the end. AGN-Starburst diagnostics

PAH

Spitzer

absorption by ionizing source •E.g., Strength of PAH anti-correlated with AGN strength (Armus et al. 2007)

We also detect absorption features from gas-phase interstellar We also detect absorption features from gas-phase interstellar C2H2 and HCN at 13.7 and 14.0 m, respectively, in 4 of the 10 C2H2 and HCN at 13.7 and 14.0 m, respectively, in 4 of the 10 BGS ULIRGs, namely Mrk 231, Arp 220, IRAS 08572+3915, BGS ULIRGs, namely Mrk 231, Arp 220, IRAS 08572+3915, and IRAS 15250+3609. Both absorption features are detected and IRAS 15250+3609. Both absorption features are detected in all galaxies, although the HCN line in Mrk 231 is uncertain in all galaxies, although the HCN line in Mrk 231 is uncertain due to a continuum ‘‘jump’’ just redward of the absorption. These due to a continuum ‘‘jump’’ just redward of the absorption. These features are labeled in Figure 3. Weak C2H2 absorption may features are labeled in Figure 3. Weak C2H2 absorption may also be present in IRAS 05189 2524 andMrk 273, but in these also be present in IRAS 05189 2524 andMrk 273, but in these galaxies the HCN line is notÀ detected. Fluxes and equivalent galaxies the HCN line is notÀ detected. Fluxes and equivalent widths (EQWs) for the C2H2 absorption-line range from 5Y10 ; widths (EQWs) for the C2H2 absorption-line range from 5Y10 ; 21 2 1 21 2 1 10À W cmÀ sÀ and 0.002Y0.012 m, respectively, with IRAS 10À W cmÀ sÀ and 0.002Y0.012 m, respectively, with IRAS 15250+3609 having the deepest C2H2 absorption and Mrk 231 15250+3609 having the deepest C2H2 absorption and Mrk 231 having the shallowest absorption among the four ULIRGs. Fluxes having the shallowest absorption among the four ULIRGs. Fluxes and EQWs for the HCN line are smaller, in the range 2Y6 ; and EQWs for the HCN line are smaller, in the range 2Y6 ; 21 2 1 21 2 1 10À WcmÀ sÀ and 0.002Y0.007 m, respectively, again with 10À WcmÀ sÀ and 0.002Y0.007 m, respectively, again with IRAS 15250+3609 showing the deepest absorption. The strengths IRAS 15250+3609 showing the deepest absorption. The strengths of the C2H2 and HCN features, as well as the measured 9:7 are of the C2H2 and HCN features, as well as the measured 9:7 are listed for the BGS ULIRGs in Table 5, below. The strengths of listed for the BGS ULIRGs in Table 5, below. The strengths of these features, when fit to models that vary the excitation tem- these features, when fit to models that vary the excitation temperature and column density of the gas, can be used to constrain perature and column density of the gas, can be used to constrain the temperature of the gas and, in some cases, its chemical evo- the temperature of the gas and, in some cases, its chemical evolutionary state (e.g., Lahuis & van Dishoeck 2000; Boonman et al. lutionary state (e.g., Lahuis & van Dishoeck 2000; Boonman et al. 2003). A detailed analysis of these lines in ULIRGs will be pre- 2003). A detailed analysis of these lines in ULIRGs will be presented in a future paper (Lahuis et al. 2007). However, the posi- sented in a future paper (Lahuis et al. 2007). However, the positions of the line centroids at 13.7 and 14.0 malreadysuggest tions of the line centroids at 13.7 and 14.0 malreadysuggest that a warm gas component of 200Y400 K dominates the ab- that a warm gas component of 200Y400 K dominates the ab- Fig. 2.—Normalized low-resolution IRS spectra. Spectra are normalized at Fig. 2.—Normalized low-resolution IRS spectra. Spectra are normalized at (a)restframe60m, and (b)restframe25m. (a)restframe60sorption in Mrkm, and 231, (b)restframe25 Arp 220, IRASm. 08572+3915, and IRAS sorption in Mrk 231, Arp 220, IRAS 08572+3915, and IRAS 15250+3609. 15250+3609. 3.2. Emission Lines 3.2. Emission Lines Figure 2, normalized at 60 m(Fig.2a)and25m(Fig.2b). In Figure 2, normalized at 60 m(Fig.2a)and25m(Fig.2b). In both normalizations, IRAS 08572+3915 has the flattest (bluest) bothThe normalizations, high-resolution IRAS (SH 08572+3915 and LH) ULIRG has the spectraflattest (bluest) are domi- The high-resolution (SH and LH) ULIRG spectra are domi- and Arp 220 has the steepest (reddest) spectrum of the BGS ULIRG andnatedby Arp 220 unresolved has the steepest atomic (redd fine-structureest) spectrum lines of the of BGS Ne, O, ULIRG Si, and natedby unresolved atomic fine-structure lines of Ne, O, Si, and sample. There is nearly a factor of 50 (25) spread among the sample.S covering There a largeis nearly range a factorin ionization of 50 (25) potential spread (see among Figs. 3 the and S covering a large range in ionization potential (see Figs. 3 and ULIRGs at 5.5 mwhennormalizedat60m(25m). ULIRGs4 and Tables at 5.5 3 andmwhennormalizedat60 4). Ratios of the fine-structurem(25 linesm). can be used 4 and Tables 3 and 4). Ratios of the fine-structure lines can be used

AGN-Starburst diagnostics

low-ionization [Ne II]

high-ionization [Ne V]

Fig. 3.—IRS SH spectra of the 10 BGS ULIRGs. Prominent•E.g., spectral [Ne features V] are / marked. [NeFig. 3.—IRS II] correlated SH spectra of the 10 BGSwith ULIRGs. AGN Prominent strength spectral features are marked. (Armus et al. 2007) Spitzer IRS: AGN vs. starburst diagnostics The Astrophysical Journal,730:28(11pp),2011March20 Petric et al.

Figure 1. Mid-infrared [Ne v]/[Ne ii]vs.6.2µmPAHEQWexcitationdiagram. Figure 2. Mid-infrared [O iv]/[Ne ii]vs.6.2µmPAHEQWexcitationdiagram. The red triangles are [Ne v]detectionswhiletheblackarrowsareupperlimits. Symbols have the same definition as in Figure 1 except that red triangles indicate •Mid-IR diagnosticsIn all cases 1 σused:error bars are [NeV]/[NeII], shown. The black circles indicate ULIRGs[OIV]/[NeII], where [O iv]detections. 6.2 μm the filled symbols show detections and empty symbols show upper limits. The (A color version of this figure is available in the online journal.) PAH EQW & blueMIR empty starscontinuum mark upper limits for SB ratios galaxies from Bernard-Salas(e.g., Veilleux et al. et al. 2009) (2009). The solid black lines indicate the fractional AGNs and SB contribution to the MIR luminosity from the [Ne v]/[Ne ii](vertical)and6.2µmPAHEQW •Punchline 1: Based(horizontal) assumingon athe simple linearMIR: mixing 10% model. In eachof case, local the 100%, LIRGs have 50%, 25%, and 10% levels are marked. The 100% level is set by the average detected values for the [Ne v]/Ne ii] and 6.2 µmEQWamongAGNsandSBs, dominant AGNrespectively, as discussed in Armus et al. (2007). The blue line traces where the summed SB and AGN contribution equals 100%. For most LIRGs the [Ne v]/ •Punchline 2: AGN[Ne ii] ratio suggestsare thatresponsible the AGN contribution to the for nuclear MIR12% luminosity of LIR of LIRGs in is below 10%. the local Universe(A color version of this figure is available in the online journal.) empirically it has been shown that emission line ratios of [Ne v]/ •Is the mid-IR [Nea sufficientlyii] ! 0.75 and [O iv]/ [Nelongii] ! 1 .wavelength75 indicate that more than to diagnose AGN/ 50% of the nuclear MIR emission is produced by an AGN (e.g., starburst fractions?Armus et al. 2007 and references therein). Three LIRG nuclei have(Petric [Ne v]/ [Neet iial.] ! 2011;0.75, implying see Veilleux’s that only 1% talk of LIRGs for more in-depth discussion) have more than 50% of their MIR emission powered by an AGN (see Figure 1). Figure 3. Mid-infrared diagnostic diagram, first used by Laurent et al. (2000)and For comparison with the LIRGs, nine SB galaxies are in- later modified for Spitzer IRS by Armus et al. (2007), comparing the integrated cluded in the figures using data from Bernard-Salas et al. continuum flux from 14 to 15 µm, the integrated continuum flux from 5.3 to 5.5, (2009): NGC 660, NGC 1222, IC 342, NGC 1614, NGC 2146, and the 6.2 µmPAHflux.Thethreevertices,labeledasAGN,Hii, and PDR, NGC 3256, NGC 3310, NGC 4088, NGC 4676, NGC 4818, represent the positions of 3C273 from Weedman et al. (2005), and M17 and NGC 7023 from Peeters et al. (2004). These vertices were chosen to facilitate NGC 7252, and NGC 7714. Three of these galaxies are LIRGs comparison with ULIRGs as presented in Armus et al. (2007)andBrandletal. while the rest have lower IR luminosities. These particular ob- (2006). The red lines from left to right indicate a 90%, 75%, and 50% fractional jects were used by Armus et al. (2007)todeterminethezero AGN contribution to the nuclear MIR luminosity, respectively. GOALS sources points in the ionized gas diagnostics. Data for seven well-studied are shown as open circles. ULIRGs from the RBGS sample analyzed in Armus et al. (2004, (A color version of this figure is available in the online journal.) 2007)arealsopresentedforcomparison:Mrk231,Arp220, IRAS 05189 2524, Mrk 273, IRAS 08572+3915, UGC 5101, [Ne v]/[Ne ii]and[Oiv]/[Ne ii]fluxratiosforasourceinwhich and Mrk 1014.− the AGN contributes 100% of the MIR luminosity. In addition The [O iv]25.890µmlineisdetectedin120nuclei(53%of differential extinction of the [Ne v]and[Oiv]linesoriginating from the obscured region around the nucleus versus the [Ne ii] the sources). Four LIRGs have ratios of [O iv]/[Ne ii] ! 1.75 implying that only 1% of LIRGs have more than 50% of their line originating from the SB (likely more extended) may lead to MIR emission powered by an AGN. The ratios of [O iv]/ an underestimation of the contribution of the AGN to the MIR [Ne ii] range between 0.002 and 5.5 and have a median of emission. 0.03 and a mean of 0.24 with a dispersion of 0.74 (see 3.2. Dust Diagnostics Figure 2). For several sources, as observedinsimilar investigations (e.g., For the sample of LIRGs in GOALS the 6.2 µm PAH EQWs Farrah et al. 2007), the estimated contributions to the MIR range between 0.01 and 0.94 µmwithameanof0.47anda luminosity from star formation and an AGN do not add up median of 0.53. About 16% of LIRG nuclei have 6.2 µmPAH to 100%. In Figures 1 and 2,thesesourcesarelocatedeither EQW below 0.27 µm, less than one-half of the value seen in below or above the mixing line between the diagnostics, that is local SB systems (Brandl et al. 2006). This suggests that 16% as the PAH EQW drops the [Ne v]/[Ne ii]shouldgoup,and of LIRGs have an AGN which dominates the MIR emission. galaxies should follow this line. However, many sources fall Figure 3 adopted from Laurent et al. (2000)andArmusetal. well below this mixing line. We suggest that this is partially due (2007) shows the 15–5.5 µmcontinuumfluxratiosversusthe to uncertainties in the AGN zero points, that is the values of the 6.2–5.5 µmcontinuumfluxratiosfortheLIRGs.Thecontinuum

5 772 (2)LUHMAN The ET[C AL. II] 158μm/FIR decrementVol. in 594 U\LIRGs size effects in the nearby starburst nuclei, further measurements are warranted. Why might high-U [=È(H)/nec] conditions exist in ULIRGs and what possible configurations might we envisage? Given the evidence for relatively higher densities in ULIRGs, the high-U conditions likely arise from high È(H) due to high Lyman continuum luminosities impinging on material at small radii, rather than from low electron densities. These conditions could arise in highly compact H ii regions, or they could arise from differences in the global conditions in ULIRGs compared to luminous IR galaxies, such as might be present in very compact luminous starburst regions where the ionized regions of large numbers of individual H ii regions merge and are possibly sur- rounded by a global PDR. Here, we might expect high U to be accompanied by high G0. For the most extreme high-U cases, such as might be present in buried or partially obscured AGNs,• theL[C dust II] is ~ 1%Fig. of9.—Ratio LFIR ofin [C normal,ii] intensity tostar-forming FIR flux plotted as agalaxies function of IR emission at the dust sublimation radius must peak at near- luminosity for the present sample (circles)andforthosegalaxiesinthe •Major coolantMalhotra in et al.the (2001) neutral sample (diamonds ISM)thatappearintheIRAS Bright IR wavelengths (Dudley & Wynn-Williams 1997). In this Galaxy Sample. Sources plotted as open symbols may have both con- case, radiative transfer effects could produce excess•ionization FIR tinuum potential and line emission = 11.26 that extend eV beyond the LWS beam (see text). The emission arising from a central source that is not associated sample medians calculated (ignoring upper limits) for the filled(e.g., diamond Luhman et al. 2003) with the normal PDR tracers in these galaxies and may suf- symbol galaxies and for the filled circle galaxies with LIR greater than or equal to 1012 L are plotted by solid lines. fer higher extinction than the PDR components. This explanation may be responsible for the observed FIR molecular absorption seen toward ULIRGs (Fischer et al. deficit hold, it may become a driver for more ambitious 1999). Such entities could be present together with a compo- future surveys. nent of more normal and possibly less obscured PDR and In Figure 9, the ratio of [C ii]toFIRfluxisplottedagainst starburst regions. These PDR and starburst regions could 8–1000 lm IR luminosity for both the sample considered dominate the line emission and thus mask the presence of an here and the Malhotra et al. (2001) sample. For galaxies not obscured central source. In this paper, we do not try to listed in Table 2, we take the IR luminosities from the IRAS ascertain quantitatively which of these scenarios is Bright Galaxy Sample (Soifer et al. 1989) or its extension dominant, although pioneering work along these lines by (Sanders et al. 1995). As a precaution, we have plotted as Laurent et al. (2000) has attempted to quantify separate open symbols sources that are listed as resolved at 25 or 60 PDR, H ii region, and AGN contributions to the mid-IR lmorboth,althoughwehavemadeuseoftheLWSdatato spectral signatures of galaxies. Rather, one of the aims of estimate the FIR flux where appropriate in the present sam- this paper is to quantify the observational evidence that ple. It is of interest that for the strongest upper limit in this high-U effects are important and to emphasize the need for plot (NGC 4418), the power source of the FIR is considered further observational and modeling efforts. to be a buried AGN based on its mid-IR spectral properties (Roche et al. 1986; Dudley & Wynn-Williams 1997; Spoon et al. 2001). The data in Figure 9 suggest a break occurring 3.6. Implications for More Distant Sources near 1012 L ,althoughadditionaldata,saywiththe As discussed in Paper I, an important implication of the Herschel and Astro-F space telescopes, are required to rule [C ii] deficit in ULIRGs is that the usefulness of the [C ii] line out a continuing decline at higher luminosities and to better in cosmological work could be diminished if cosmological define the sharpness of the break. sources resemble this local sample. As a result of the present Since the [C ii]158lm line is one of the brightest lines in work, we can begin to provide an indication of what this normal and starburst galaxies (and ULIRGs) in the wave- means quantitatively. In the present sample, only 3 of 15 length range where dust extinction plays a lesser role, its ULIRGs resemble normal and starburst galaxies in the usefulness as an astrophysical diagnostic remains undimin- ratio of [C ii]-to-dust continuum, so that an experiment that ished notwithstanding the cautionary implications of the could just detect a cosmological ultraluminous IR source, present work. In the context of the high-U hypothesis dis- such as those detected in continuum using SCUBA with cussed above, sources with low dust abundance relative to normal [C ii]properties,wouldneedtoobservefivesources hydrogen would have to achieve an even higher U to pro- for one [C ii]detection.Specifically,forthehigh-redshift duce a similar diminution of the [C ii] line intensity relative sources considered by Stark (1997), the per season detection to the continuum. For such scenarios, we might expect the rate estimate would be reduced from 200 to 40. Such a detec- break suggested in Figure 9 to migrate to higher luminosi- tion rate should still provide useful information for under- ties if U and L are generically correlated and enrichment is a standing star formation as a function of redshift (e.g., gradual function of cosmic time in individual sources. If Madau, Pozzetti, & Dickinson 1998), since lower limits to such is the case, then mid- and far-infrared fine-structure the massive star formation rate can be calculated directly lines, which are preferred ISM abundance diagnostics (e.g., from the measured [C ii] intensity. Given the presence of Simpson et al. 1998, 1995), as well as the [C ii]158lm line upper limits to the [C ii]fluxinoursample,itisnotpossible itself, will be easily observed in moderate-luminosity cosmo- to say what would be required to detect the entire sample at logical ultraluminous IR sources and can provide a check cosmological distances, but it is clear that should the [C ii] on this scenario. These could be the majority of such sources [C II]/FIR vs. the star formation efficiency The Astrophysical Journal Letters,728:L7(5pp),2011February10 Gracia-Carpio´ et al.

−1 10 [C II] 158µm [O I] 145µm [N II] 122µm [O III] 88µm FIR FIR FIR FIR

10−2

10−3 [CII]/FIR 10−4

10−5

−1 1 10 100 1000 10 [O I] 63µm [N III] 57µm [O III] 52µm LFIR/M(H2) = star formation efﬁciency → Line flux / FIR FIR FIR FIR

• [CII]/FIR ↓ as 10the−2 star formation efﬁciency ↑ Similar decrements are seemHII galaxy for other forbidden submm lines • LINER 10−3 Seyfert & QSO (Gracia-Carpio et al. 2011) Blue Compact Dwarf Unclassified IRAS 17208−0014 −4 10 High−z galaxy

NGC 4418 Arp 220 10−5

100 101 102 103 100 101 102 103 100 101 102 103 100 101 102 103

LFIR / MH2 [ LSun / MSun ]

Figure 2. Global line to FIR continuum ratios in galaxies with different optical activity classifications as a function of LFIR/MH2 . PACS observations are represented 1 with squares. Open symbols indicate 3σ upper limits to the line flux. All the line to FIR ratios start to decrease at LFIR/MH2 80 L M − (vertical lines). The ∼ " " 1 horizontal lines indicate the approximate minimum values for the different line to FIR ratios observed in galaxies with LFIR/MH < 80 L M − . 2 " " (A color version of this figure is available in the online journal.)

Our main result is that we now find line deficits in both the PDR not be true for some galaxies like Mrk 231, but even an AGN and the H ii lines. All the line to FIR ratios start to decrease contribution of 50% to LFIR will fall short of explaining the 1 at LFIR/MH2 80 L M − .Thisresultalsodoesnotseem factor of 10 decrease in some of the line to FIR ratios. to depend on∼ the optical" activity" classification of the galaxies. In the next section, we show with quantitative modeling that We note, however, that the integrated emission of the [O i]and high ionization parameters can indeed account for most of the [O iii]linesinsomeSeyfertgalaxiesmighthaveasignificant observed line deficits. The key factor is that in dusty star-forming contribution from X-ray dominated regions close to the active regions the fraction of luminosity absorbed by dust particles in galactic nucleus (AGN; e.g., Maloney et al. 1996). ionized as well as neutral gas increases with U. The [O i]63µmlinewasdetectedinabsorptionatsome velocities in NGC 4418, Arp 220, and IRAS 17208-0014. The 4. COMBINED H ii AND PDR LINE MODELS [O i]63µm/FIR ratios used in Figure 2 correspond to the Abel et al. (2009)usedthespectralsynthesiscodeCloudy emission components only and should be regarded as lower (Ferland et al. 1998)tomodelthe[Cii]/FIR ratio in clouds limits. This will reduce the deficit observed for this line to some exposed to different radiation fields. They found that they could extent, but will not eliminate it, as other galaxies (e.g., Mrk 231; reproduce the [C ii] line deficits observed in some local LIRGs Fischer et al. 2010)withnoclearabsorptionfeaturesintheir and ULIRGs, as well as their warmer IRAS S60 µm/S100 µm line profiles do still have lower [O i]63µm/FIR ratios than colors, by increasing the value of the ionization parameter at the galaxies with L /M < 80 L M 1.The[Oi]145µm line FIR H2 − surface of the clouds. As U increases, the H ii region is extended is less likely to be affected by this" kind" of absorption and also to higher AV into the cloud. As a result, a larger fraction of the shows a deficit at high L /M .Thereisnosignofabsorption FIR H2 UV photons are absorbed by the dust in the ionized region and by foreground material in the other fine structure lines at the reemitted in the form of infrared emission. The net effect is that current PACS velocity resolution ( 100–250 km s 1). ∼ − the fraction of UV photons available to ionize and excite the In summary, our observations show that the line to FIR gas is reduced at high U,decreasingtherelativeintensityofthe continuum ratio drops by a factor of 3 to 10 above L /M FIR H2 fine structure lines compared to the FIR continuum (see also 1 ∼ 80 L M − in all the far-infrared fine structure lines we Voi t 1992). Since the number of UV photons per dust particle " " have observed. This drop seems to be a universal feature of increases with U,thedusttemperature(andtheS60 µm/S100 µm galaxies at different redshifts and with different optical activity color) increases. classifications. Neutral and ionized gas tracers are affected, We have extended the analysis in Abel et al. (2009)toinclude as are lines with different rest-frame wavelengths, intrinsic all the fine structure lines observed with PACS. We used Cloudy optical depths, and critical densities. This eliminates most to simultaneously calculate the emission from the molecular, of the proposed explanations for the [C ii]deficit,leaving neutral, and ionized gas assuming pressure equilibrium in the the dependence on the LFIR/MH2 ratio as the most plausible clouds. The ionizing source was represented by a non-LTE dominant factor driving the deficits. We have assumed in this CoStar stellar atmosphere (Schaerer & de Koter 1997)withT ∗ = Letter that LFIR is mostly due to star formation. This might 36,000 K, and the calculations were stopped at AV 100 mag = 3 The Astrophysical Journal Letters,765:L13(6pp),2013March1 Zhao et al.

low [C II]/FIR galaxies have high ionization parameters (U) Does [C ii] Emission Trace Current SF in LIRGs? First GOALS Results FromHerschel/PACS 5

sible to disentangle. As mentioned in §2.3, the angular size of a PACS spaxel is roughly similar to that of the aperture used to ex- tract the Spizer/IRS spectra of our galaxies. Because the PACS beam is under-sampled at 160 µm(FWHM∼ 12!! compared with the 9.6!! size of the PACS spaxels), and most of the sources in the sample are unresolved at 24 µm in our MIPS images (which have a similar angular resolution as PACS at ∼ 80 µm), an aperture correction has to be performed to the spectra extracted from the emission-peak spaxel of each galaxy to obtain their total nuclear fluxes. This was the same procedure employed to obtain the mid-IR IRS spectra of our LIRGs. The nominal, wavelength-dependenthigh ionization aperture / low correction ionization function → warmer dust → provided by HIPE v8.0 works optimally when the source is exactly positioned at the center of a given spaxel. How- ever, in some occasions the pointing of Herschel is not Figure 3. Ratio of [N ii]205µmtoIRluminosityisplottedagainst((a)and(d))IRluminosity,(b)IRAS 60 µm/100 µmcolor,and(c)[Oiii]88µm/[N ii]205µm accurate enough to achieve this and the target can be flux ratio. The symbols are the same!! as in Figure 2 except that here larger circles represent ULIRGs. The symbols enclosed by an additional red circle in panels (a) andslightly (b) are the misplaced same sources as! shown3 (up in panels to (c)1/3 and of (d). a Equation spaxel) (1 from) is shown the by the dashed line in panel (a). The ordinate of panel (d) is a nominal value after thecenter. correction for In the these hardness cases, effect the (using flux the of equation the line shown might in this panel). be under- Mrk 231 is outside the plotted range of panel (d). (A colorestimated. version of this We figure explored is available whether in the online this journal.) effect could be corrected•... byas measuring inferred the position from of thethe source [OIII] within the/ [NII] spaxel. However, some LIRGs in our sample show low ++ onlysurface-brightness•...and5eVhigherthanthatneededfortheformationofN dust extended temperature emission, either because, of the real scatter in the L[N ii]–LIR relation might probably be even andtheir thus∼ theproximity [O iii]/[N and/orii]lineratiosareevenstrongerindicator merger nature, or simply because smaller for these galaxies. of the hardness. gas and Furthermore, dust emission this are ratio spatially is insensitive decoupled. to the gas This, To obtain a direct view of how much the scatter can be re- densitycombined (Rubin with1984 the). Except spatial for sub-sampling M82 (Duffy et of al. the1987 PACS/IFS)and duced, we also display the hardness-corrected L[N ii]/LIR,which Mrkdetector 231 (Fischer and et the al. poor2010), S/N all of of the some [O iii sources] 88 µmfluxesare prevented us was calculated according to the fitted L[O iii]/L[N ii]–L[N ii]/LIR adopted(Zhaofrom from obtaining et Brauher al. 2013; an et al.accurate (2008 Diaz-Santos). measurement et of al. the 2013; spatial seerelation, also for Abel the [O iii et]sampleinFigure al. 2009; Gracia-Carpio3(d) (arbitrarily scaled). 2011) Figureposition3(c) and shows width the of relation the [C betweenii]emissionandthereforeL[O iii]/L[N ii] and Comparing to Figure 3(a), one can see that the scatter is reduced L[Nfromii]/LIR obtaining.Clearly,thereexistsagoodcorrelationsuchas a more refined aperture correction.L[N ii] Thus,/ significantly (from 0.26 dex to 0.11 dex by excluding Mrk 231, LIRwedecreases performed as L[O onlyiii]/L the[N ii] nominalincreases. aperture There is only correction one out- pro- which is still an outlier due to that the hardness correction has lier,vided the ULIRG by HIPE. Mrk 231, which shows both small L[N ii]/LIR little effect on it, and is outside the plotted range). and LThe[O iii]/L individual[N ii].ExcludingMrk231,andusinganordinary far-IR luminosities of galaxies belong- All of the seven ULIRGs in our sample show some deficit least-squaresing to a LIRGlinear fit system to the data, formed we obtain by of log( twoL[N orii]/L moreIR) com- in L[N ii] relative to LIR.Fouroftheseven,includingArp220 = ( 3ponents.63 0.06) were + ( derived0.61 0 in.06) a similarlog(L[O iii manner]/L[N ii]). to The the Spear- method and Mrk 231, were observed by ISO LWS, but no [O iii]88µm man−used rank± to correlation calculate− coefficient their± L ρ of(see this§ correlation2.1). In this is case,0.93 we emissionFig. was 1.— detectedThe ratio at the of [C3σiilimit]157.7 (e.g.,µm to Brauher far-IR et luminosity a. 2008), (upper IR ii at ascaled>5σ level the ofintegrated significance.IRAS far-IR luminosity of− the sys- despitepanel) all showing and [C ]157.7 someµ AGNm to characteristics. monochromatic These luminosity galaxies at 158 µm − (bottom panel) as a function of the Sν 63 µm/Sν 158 µm continuum Galaxies42 plotted.5 122. in5 µ Figurem 3(c) (hereafter the [O iii]sample) also showflux density very warm ratio FIR for individualcolors (see galaxies Figure 3 in(b)). the Therefore, GOALS sample. tem (LFIR ,asdefinedinHelouetal.1985)with are also marked with red circles in Figures 3(a) and (b), which veryCircles dusty H ofii diregionsfferent with colors high indicateU might the beLFIR neededof galaxies to explain (see color the ratio between the continuum flux density of each indi- 12 indicate that the hardness variation of the heating radiation field the deficitbar). of Red [N diamondsii] emission mark in these galaxies galaxies with L (e.g.,IR ≥ Fischer10 L! et, al. ULIRGs. vidual galaxy evaluated at 63 µminthePACSspectrum These two plots show that the decrease of the L /L ratio among galaxies can largely account for the scatterii shown in 1999;Abeletal.2009;Draine2011). [C II] FIR (measured in the same spaxel as the [C12 ]line),andthe with warmer far-IR colors seen in our LIRGs is primarily caused the LIR–L[N ii] relation, at least for LIR < 10 L galaxies. total IRAS 60 µmfluxdensity. # by a significant decrease of gas heating efficiency and an increase Actually, this is also supported by the fact that the scatter of of warm dust emission.4. SUMMARY less luminous galaxies4. RESULTS (0.2 dex for AND Brauher DISCUSION et al. sample; thus probably less active in star formation and lower )issmaller In this Letter we have presented our initial results of the [N ii] 4.1. The [C ii]/L Ratio: Dust HeatingU and Cooling than that of our luminousFIR sample (0.28 dex). Keeping in mind 205 µmemissionforasampleof70(U)LIRGsobservedby that theThe conversion far-IR fine from structure [N ii]122µ linem flux emission to [N ii]205 in normalµmflux star- the HerschelmainlySPIRE because FTS. of Combining two main with reasons: the ISO (1)and thisIRAS way we mayforming contribute galaxies largely as to well the scatter as in ofthe Brauher extreme et al. environments sample, data,are we able have plot studied data the from correlation individual between galaxiesL[N ii] insteadand LIR, of be- hosted by ULIRGs has been extensively studied for the ing constrained by the spatial resolution of IRAS,which past two decades. A number of works based on ISO data5 would force us to show only blended sources; (2) by using already suggested that the relative contribution of the the 63/158 µmratioweareprobingalargerrangeofdust [C ii]157.7 µm line to the cooling of the ISM in PDRs as temperatures within the starburst (T ∼ 50 → 20 K); compared to that of large dust grains, as gauged by the with the colder component probably arising from regions far-IR emission, diminishes as galaxies are more IR lumi- located far from the ionized gas-phase, and closer to the nous (Malhotra et al. 1997; Luhman et al. 1998; Brauher PDRs where the [C ii]emissionoriginates.Forreference, et al. 2008). we show the relation between the PACS 63/158 µmand IRAS 60/100 µmcolorsintheAppendix.TheULIRGs 4.1.1. The Average Dust Temperature of LIRGs in the GOALS sample (red diamonds) have a median −4 −4 Figure 1 (upper panel) shows the L /L L[C II]/LFIR =6.5 × 10 ,ameanof7.0(± 0.9) × 10 , [C II]157.7 µm FIR −4 ratio for the GOALS sample as a function of the far-IR and a standard deviation of the distribution of 4 × 10 . PACS Sν 63 µm/Sν 158 µmcontinuumfluxdensityratio. LIRGs span two orders of magnitude in L[C II]/LFIR, We chose to use this PACS-based far-IR color in the x- from ∼ 10−2 to ∼ 10−4,withameanof3.5 × 10−3 −3 axis instead of the more common IRAS 60/100 µmcolor and a median of 2.8 × 10 .TheL[C II] ranges from low [C II]/FIR galaxies1140 are optically thickBRANDL ET AL. Vol. 653 6 D´ıaz-Santos et al.

7 9 ∼ 10 to 2 × 10 L!. Our results are consistent with ISO observations of a sample of normal and moderate IR-luminous galaxies presented in Malhotra et al. (1997) and further analyzed in Helou et al. (2001). The GOALS sample, though, populates a warmer far-IR color regime and extends the trend inferred from the ISO data more than one order of magnitude to higher LIR. Despite the increase in disper- −3 sion at 63/158 µm ! 1.25 or L[C II]157.7 µm/LFIR " 10 (basically in the ULIRG domain), the fact that we find the same tight trend independently of the range of IR luminosities covered by the two samples suggests that the main observable linked to the variation of the L[C II]/LFIR ratio is the average temperature of the dust Fig. 11.—Comparison of the average starburst template (dotted line)from Fig. 6 to the most extincted source within our sample, NGC 4945 (dashed line), (Tdust)ingalaxies. and to the ULIRG IRAS 08572+3915 (solid line), chosen as an extreme case of This interpretation agrees with the last physical sce- extinction from the sample of Spoon et al. (2006). The spectra have been nor- =ln(fobs/fcont)9.7μm nario described in the Introduction, in which an increase malized at 14.5 m. of the ionization parameter, ,wouldcausethefar- Fig. 2.— L[C II]157.7 µm/LFIR ratio for individual galaxies in UV radiation from the youngest stars to be less efficient starburst galaxies in Figure 10 do not show evidence for a similar in heating the gas in those galaxies. At the same time, the GOALS sample as a function of the strength of the 9.7 µm silicate absorption feature, S9.7 µm, measured with Spitzer/IRS. emission structure. The 13Y35 m spectra, from NGC 1222 (top) dust grains would be on average at higher temperatures Galaxies are color-coded as a function of the Sν 63 µm/Sν 158 µm optical depth ← increasing to NGC 4945 at the bottom, show an increasingly pronounced due to the larger number of ionizing photons per dust ratio, with values ranging from 0.10 to 4.15 (see Figure 1). The depression, peaking at 18.5 m, signaling increasingly strong sili- particle available in the outer layers of the H ii regions, solid• line [CII]/FIR represents an outlier-resistant ↓ as optical fit to the bulkdepth of the star- ↑ cate absorption. The latter result is in full agreement with the trend forming LIRG population with S9.7 µm ≥−2. The dotted lines are found for the 9.8 m silicate feature. close to the PDRs. Indeed, the presence of dust within the ± 1 σ uncertainty. H ii regions has been recently observed in several star- Among all the galaxies classified as starbursts, NGC 4945 forming regions in our Galaxy (Paladini et al. 2012). (bottom spectrum in Fig. 10) is a ‘‘special case,’’exhibiting by far the strongest dust obscuration to its nuclear region (Spoon Both effects combined can explain the wide range of et al. 2000). Based on the IRS spectrum, the apparent optical IR emission. The formal fit (solid line) can be expressed Fig. 10.—IRS spectra of Mrk 52, NGC 7714, NGC 1222, NGC 7252, NGC L[C II]157.7 µm/LFIR ratios and far-IR colors we observe in as: 2146, NGC 520, NGC 3628, and NGC 4945 arranged to illustrate the gradual depth in the 9.8 m silicate feature is at least 4 and may be higher, the most warm LIRGs. Variations in nH, though, could (Diaz-Santoseffect et of increasing al. 2013; silicate absorption spectra from top - toBrandl bottom. The fluxet densitiesal. 2006)depending on the choice of the local continuum. Apart from be responsible for the dispersion in L[C II]/LFIR that we Fk have been arbitrarily scaled for better comparison. strong amorphous silicate absorption, the line of sight also reveals see at a given 63/158 µmratio. log(L /L )=−2.30(±0.02) + 0.82(±0.05) S the presence of a 23 m absorption feature. Following the analysis [C II] FIR 9.7 µm of deeply obscured lines of sight toward ULIRG nuclei (Spoon To further support these findings, the bottom panel of (1) log space, i.e., the IRS estimated infrared luminosities agree Figure 1 shows that the drop in the ratio of [C ii]157.7 µm et al. 2006), we attribute the 23 m feature to crystalline silicates within 23% with LIR. This is much more accurate than the esti- (forsterite). Their detection in NGC 4945 suggests that crystal- luminosity to the monochromatic continuum at ∼ 160 µm with a dispersion in the y-axis of 0.30 dex. mates by Chary & Elbaz (2001) and Takeuchi et al. (2005). The line silicates are perhaps a more common component of the ISM is smaller by an order of magnitude than the correspond- Within the context described in the previous section, excellent correlation suggests that, at least for a homogeneous and not just limited to ULIRG nuclei. sample of starburst galaxies, F and F can be used to ing drop in L[C II]/LFIR (albeit with a larger dispersion). the contrast between the inner layer of dust that is being 15 m 30 m Figure 11 compares the average starburst template from Fig- accurately derive LIR. This implies that the decrease of the L[C II]/LFIR ratio heated by the ionizing radiation to T ! 50 K and that ure 6 to NGC 4945, the most extincted source within our sample. For comparison we also show the heavily embedded ULIRG with warmer far-IR colors seen in our LIRGs is primar- of the cold dust at T " 20 K emitting at λ ! 150 µm 4.3. ALarge Variety in Silicate Absorption ily caused by a significant increase in warm dust emission would create both: (1) the silicate absorption seen at IRAS 08572+3915 from Spoon et al. (2006). The usually rather (peaking at λ " 60 µm) that is not followed by a propor- Figure 10 shows, from top to bottom, a series of starburst spec- shallow 18 m silicate band reduces the continuum by a large 9.7 µmduetothelargertemperaturegradientbetween tra, sorted by increasing absorption of the 9.8 m silicate reso- factor compared to the average starburst spectrum. While NGC tional enhancement of [C ii]emission,contrarilytowhat the two dust components and (2) the increasingly higher nance. Since the peak of the resonance coincides with a minimum 4945 has a similarly strong silicate absorption feature as the is normally seen in star-forming galaxies with lower IR 63/158 µmratiosseeninFigure1duetotheprogres- between the 7Y9and11Y13 mPAHemissioncomplexes,theef- ULIRG IRAS 08572+3915, the starburst also shows strong PAH luminosities. This suggest that the [C ii]157.7 µmline sively larger amount of dust mass that is being heated to fect of silicate absorption only becomes apparent toward the lower complexes near 16.4 and 17.1 m, which are absent in the ULIRG is not sensitive to the youngest stars responsible for the higher temperatures. This scenario is consistent with the half of the plot. Nevertheless, the figure strikingly illustrates the spectrum. Crystalline silicates, causing the features at 16 and 19 m warm dust emission. We will elaborate on this point in physical properties of the ISM found in the extreme envi- strong effect of amorphous silicates on the overall 5Y38 mspec- in IRAS 08572+3915, appear to be absent in the starburst tem- §4.2. ronments of ULIRGs, in which the fraction of total dust tral shape and the large variations even within one class of objects. plate, although they are more difficult to discern at our spectral luminosity contributed by the diffuse ISM decreases sig- Figure 10 shows that the 10 m trough can in fact become a dom- resolution given the presence of PAH features and the [S iii] line inating feature of the spectral shape of starbursts. in that range. 4.1.2. ALinkBetweenMid-IRDustObscurationand nificantly, and the emission from dust at T ∼ 50 − 60 K Far-IR Re-emission The top two spectra also show a broad emission structure be- While NGC 4945 shows by far the most extreme silicate ab- arising from optically-thick ”birth clouds” (with ages ginning at about 16 mandextendingto21Y25 m. The shape sorption within our sample, it is not an exotic object but defines − The strength of the 9.7 µmsilicatefeatureisdefinedas " 107 8 Myr) accounts for ! 80% of their IR energy out- of this feature is consistent with that of an 18 msilicateemis- the endpoint of a sequence of increasing optical depths. Recently, obs cont cont obs put (da Cunha et al. 2010). Furthermore, our findings sion feature, which would make these the first detections of this Dale et al. (2006) reported on the Spitzer SINGS survey of 75 S9.7 µm ≡ ln(fλP /fλP ); with fλP and fλP being the un-obscured and observed continuum flux density mea- are also in agreement with recent results showing that the feature in starburst galaxies. However, the emission structure nearby galaxies. The typical target in their sample shows only could also be due to the C C Cin-planeandout-of-planebending modest dust obscuration, consistent with an AV 1foreground sured in the mid-IR IRS spectra of our LIRGs and eval- increase of the silicate optical depth in LIRGs is related À À with the flattening of their radio spectral index (1.4 to modes of PAHs (Van Kerckhoven et al. 2000). The identification screen and a lack of dense clumps of highly obscured gas and dust. uated at the peak of the feature, λP ,normallyat9.7µm of this feature will be discussed in a future paper. The remaining Unfortunately, we cannot distinguish between the local and global (see Stierwalt et al. 2013a for details on how it was cal- 8.44 GHz) due to an increase of free-free absorption, sug- culated in our sample). Negative values indicate absorp- gesting that the dust obscuration must largely be origi- tion, while positive ones indicate emission. By definition, nated in the vicinity of the starburst region (Murphy et S9.7 µm measures the apparent optical depth towards the al. 2013). warm, mid-IR emitting dust. Figure 2 shows that there There are a few galaxies that do not follow the cor- is a clear trend (r =0.65,pr =0;κ =0.47) for LIRGs relation fitted in Figure 2, showing very large silicate with stronger (more negative) S9.7 µm to display smaller strengths (S9.7 µm < −2) and small L[C II]/LFIR ratios −3 L[C II]/LFIR ratios, implying that the dust responsible (< 10 )typicalofULIRGsor,ingeneral,warmgalax- for the mid-IR absorption is also accountable for the far- ies (see color coding or Figure 1). We would like to note low [C II]/FIR galaxies have compact IR- emitting regions 8 D´ıaz-Santos et al. Fit to “SB-dominated” U/LIRGs

low PAH EQW - AGN-dominant U/LIRGs

• [CII]/FIRFig. ↓ 4.— as Lthe[C II]157 compactness.7 µm/LFIR ratio asof a the function MIR of theemission lumi- ↑ nosity surface density at 15 µm, Σ15 µm for individual galaxies in the GOALS sample. This Figure is the same as Figure 3 but color coded as a function of the 6.2 µmPAH EW to highlight(Diaz-Santos et al. 2013) those sources whose mid-IR emission might indicate the presence of an AGN (see text for details). The solid line is a ﬁt to pure star-forming LIRGs only, deﬁned as to have 6.2 µmPAH EWs ≥ 0.5 µm.

not caused by a rise of AGN activity but instead is a fundamental property of the starburst itself. It is only in −3 the most extreme cases, when L[C II]/LFIR < 10 ,that the AGN could play a significant role. In fact, power- ful AGN do not always reduce the L[C II]/LFIR ratio, as shown also in Sargsyan et al. (2012). Stacey et al. (2010) also find that the AGN-powered sources in their high- redshift galaxy sample display small L[C II]/LFIR ratios. However, they speculate that except for two blazars, the Fig. 3.— L[C II]157.7 µm/LFIR ratio as a function of the lumi- deficit seen in these sources could be caused compact, nu- nosity surface density at 15 µm, Σ15 µm (top), and the fraction of clear starbursts (with sizes less than 1 − 3kpc)perhaps extended emission at 13.2 µm, FEE13.2 µm (bottom), for individual galaxies in the GOALS sample. Galaxies are color-coded as a func- triggered by the AGN. We will return to this discussion tion of the 63/158 µm ratio (top) and the strength of the silicate in §4.2.2. feature, S9.7 µm (bottom). A linear fit to the datapoints without The result obtained above also implies that the limits is shown in the top panel as a solid line. The dotted lines [C ii]157.7 µmlinealoneisnotagoodtraceroftheSFR are the ± 1 σ uncertainty.We note that the 15 µ luminosities are measured within the Spitzer/IRS LL slit while the 13.2 µm sizes in most LIRGs since it does not account for the increase were obtained from the SL module. of warm dust emission (Figure 1) seen in the most compact galaxies that is usually associated with the most recent starburst. In Figure 4 we fit those sources with able to destroy a significant fraction of the smallest PAH 6.2 µmPAH EWs ≥ 0.5 µmtoprovidearelationbe- molecules (Voit 1992; Siebenmorgen et al. 2004). In par- tween the L[C II]/LFIR ratio and the Σ15 µm for pure star- ticular, a galaxy is regarded as mid-IR AGN-dominated forming LIRGs. The analytic expression of the fit is: when its 6.2 µmPAH EW ! 0.3, and it is classified as apurestarburstwhen6.2µmPAH EW " 0.5(although these limits are not strict). Sources with intermediate L[C II] log( )=1.87(± 0.39) − 0.47(±0.05) × log(Σ15 µm) values are considered composite galaxies, in which both LFIR starburst and AGN may contribute significantly to the (3) mid-IR emission. with a dispersion in the y-axis of 0.20 dex. The slope and intercept of this trend are indistinguishable (within 4.2.1. [C ii]157.7 µm Deficit in Pure Star-Forming LIRGs the uncertainties) from those obtained in Eq. (2), fur- Figure 4 shows the same data as Figure 3 (top) but ther supporting the idea that the influence of AGN activity is negligible among IR-selected galaxies with color-coded as a function of the 6.2 µmPAH EW.Aswe −3 −2 can see, when only pure star-forming galaxies are con- 10 >[CII]<< Molecular PAH e- Ho Cloud >>[CII]<< PAH Ho [CII] PDR PAH Ho [CII] Ho HII Region H+ H+ o H PAH o H+ H+ H [CII] H+ H+ H+ H+ H+ H+ H+ FIR H+ H+ H+ UV

•As U ↑, the gas opacity in the HII regions ↓ •chance of UV photons → dust → FIR photons high, especially if the starburst region is optically thick U/LIRGs with high SSFRs

Main Does [C ii]EmissionTraceCurrentSFinLIRGs?FirstGOALSResultsFroStarbursts mHerschel/PACS 11 Sequence lower than the value predicted by the fit to our local galaxy sample. Interestingly, both SMGs lie in the parameter space where most of the mid-IR identified AGN are located (red stars). This may suggest that these two galaxies could harbor AGN or unusually week [C ii]emission for their normalized SSFR. Eq. 3 can also be used to predict the mid-IR luminosity surface density of star-forming galaxies at high redshifts for which the [C ii]andfar-IRluminosityareknown, such as those which may be found in future spectroscopic surveys with CCAT. Moreover, if a measurement of their rest-frame mid-IR luminosity is also available, this correlation can be further used to estimate the physical size of the star-forming region in the mid-IR. This is particularly useful for Herschel detections of z > 3 sources in deep fields that have been already observed. In this cases, the = SFR/M* predicted mid-IR size of galaxies could be compared with ... form their stars in warm, compact, optically thick regionsdirect measurements of the size of their far-IR emitting Fig. 6.— L[C II]157.7 µm/LFIR ratio versus LIR/M! normalized region as observed with ALMA on physical scales simi- ∼ " −1 to the LIR/M! of MS galaxies at z 0(SSFRMS 0.09 Gyr(Diaz-Santos), lar et to al. those 2013) we are probing in our GOALS LIRGs with for individual galaxies in the GOALS sample. Galaxies are color- PACS. coded as a function of their 6.2 µm PAH EW. Colored circles indicate sources for which an EW is available. Squares indicate lower Finally, because GOALS is a complete flux-limited limits. Small black circles are sources for which there is no infor- sample of local LIRGs, we are able to predict the con- mation. The solid line is a fit to pure star-forming LIRGs only, tamination of sources hosting AGNs in future large-scale excluding those sources that may harbor an AGN that could dom- surveys with both [C ii]andfar-IRmeasurements.In inate their mid-IR emission (6.2 µmPAH EWs < 0.5 µm; see Fig- ure 5). The dotted lines are the ± 1 σ uncertainty. The dashed line Table 1 we provide the percentages of galaxies with mid- indicates the SSFR of MS galaxies. The dotted-dashed line repre- IR detected AGNs classified in different L[C II]/LFIR and sents 3 × SSFRMS, the limit above which galaxies are considered S 63 µm/S 158 µm bins. The values provided in the Ta- to be starbursting. ν ν ble were computed using two conditions for the detection of the AGN that serve as upper and lower limits for the estimated fractions (columns (2) and (3)). The first was SSFR; see, e.g., Graciá-Carpio et al. 2011) then this based on the 6.2 µmPAH EW only and the second re- equation can be applied to predict the [C ii]luminos- quired of an additional mid-IR diagnostic to classify the ity of star-forming galaxies at any redshift for which a galaxy as harboring an AGN (see above). For exam- measurement of the (far-)IR luminosity and stellar mass ple, we predict an AGN contamination of up to 65% for −4 are available as long as their SSFR is normalized to the L[C II]/LFIR < 5 × 10 ,whichimpliesthatatleast1/3 SSFR at that z to account for the increase in gas mass MS of IR-selected sources with extremely low L[C II]/LFIR fraction (Mgas/M!)andthereforeSSFRofMSgalaxies ratios will be powered by starbursts. Moreover, at the at higher redshifts (Daddi et al. 2010a; Magdis et al. −4 levels of L[C II]/LFIR ≥ 5 × 10 or 63/158 µm < 2, the 2012). Conversely, in future large IR surveys like those AGN fraction is expected to be ≤ 20%. projected with the Cornell-Caltech Atacama Telescope (CCAT), this relation could be used for estimating the 6. CONCLUSIONS SSFR or stellar mass of detected galaxies when their [C ii] We obtained new Herschel/PACS [C ii]157.7 µmspec- and far-IR luminosities are known. troscopy for 200 LIRG systems in GOALS, a 60 µmflux- As an example, in Figure 6 we show two interme- limited sample of all LIRGs detected in the nearby Uni- diate redshift z ∼ 1.2galaxiesfromStaceyetal. verse, which are the most likely local counterparts of the (2010) (SMMJ123633 and 3C368) and three high redshift dominating IR-selected population of star-forming galax- z ∼ 4.5sub-millimetergalaxies(SMGs)(LESS61.1and ies at high redshift. A total of 240 individual galaxies LESS 65.1: Swinbank et al. 2012; LESS 033229.3: De where observed in [C ii]. We combined this information Breuck et al. 2011; Wardlow et al. 2011) marked as black together with Spitzer/IRS spectroscopic data to provide squares and diamonds, respectively. To calculate the the context in which the observed [C ii]luminositiesand SFR of the SMGs we assumed that their L " 2 × L . IR FIR [C ii]/L ratios are best explained. We have found the We estimated the SSFR at the redshift of each galaxy FIR MS following results: in the same manner as described above. However, since Eq. 13 from Elbaz et al. (2011) is calibrated only up to • The LIRGs in GOALS span two orders of mag- z ∼ 3, we normalized the SSFR of the SMGs by the nitude in L /L , from ∼ 10−2 to 10−4, SSFR at this redshift (" 4.6Gyr−1). As we can see, [C II] FIR MS with a median of 2.8 × 10−3.ULIRGshavea three galaxies follow the correlation suggesting that they −4 median of 6.5 × 10 .TheL[C II] range from are starbursting sources with L[C II]/LFIR ratios consis- 7 9 tent with their normalized SSFRs. On the other hand, ∼ 10 to 2 × 10 L" for the whole sample. The the two remaining high-z SMGs show significantly lower L[C II]/LFIR ratio is correlated with the far-IR L[C II]/LFIR ratios than the average of LIRGs for the Sν 63 µm/Sν 158 µmcontinuumcolor.Wefindthat same normalized SSFR. In particular, one of them dis- all galaxies follow the same trend independently play a L[C II]/LFIR more than an order of magnitude of their LIR,suggestingthatthemainobservable A&A 533, A119 (2011) U\LIRGs with high SSFRs

radio calibrated Fig. 16. Left:SFR–M correlation at z 0. Galaxies classified as compact are marked with large filled red dots. Solid line: fit to the main sequence ∗ 9 ∼ sizes SFR–M relation: SFR M!/[4 10 M ]. Dotted lines: 16th and 84th percentiles of the distribution around the sliding median (0.26 dex). Right: relation∗ of the sSFR and∝ IR surface× brightness% of galaxies for which a radio size was estimated. The vertical dashed line illustrates the threshold above which galaxies have been classified as compact. The solid and dashed red lines are a fit to the sliding median of the relation (Eq. (11)) and its 68% dispersion. ... form their stars in warm, compact, optically thick regions (Elbaz et al. 2011)

RSB,whichmeasurestheexcessinsSFR of a star-forming galaxy (which we label its “starburstiness”), as deﬁned in Eq. (10):

RSB = sSFR/sSFRMS = τMS/τ [> 2forstarbursts], (10) where the subscript MS indicates the typical value for main sequence galaxies at the redshift of the galaxy in question. A starburst is defined to be a galaxy with RSB 2. 75% of the galax- ≥ 10 2 ies with compact star formation (ΣIR 3 10 L kpc− )have ≥ × % RSB > 2, hence are also in a starburst mode, and 93% of them have RSB>1. Conversely, 79% of the starburst galaxies are “compact”. Globally, starburst galaxies with sSFR > 2 sSFR have 11 2 ×( ) amedianΣIR 1.6 10 L kpc− ,hencemorethan5times higher than the∼ critical× IR surface% brightness above which galaxies are compact. The size of their star-forming regions is typi- cally 2.3 times smaller than that of galaxies with sSFRMS. The sSFR and ΣIR are correlated with a 0.2 dex dispersion 1 (Fig. 16-right) following Eq. (11), where sSFR is in Gyr− and 2 ΣIR in L kpc− : tot % Fig. 17. RSB = sSFR/sSFRMS versus IR8 (=LIR/L8)forz 0galaxies 4 0.33 ∼ sSFR = 1.81 [ 0.66, +1.05] 10− Σ . (11) (AKARI and GOALS samples). The red line is the best fit (plain) and − × × IR its 0.3 dex dispersion (dashed). Both parameters measure specific quantities related to the SFR: the sSFR is measured per unit stellar mass, while ΣIR is related tot 6. IR8 as a tracer of star formation compactness to the SFR (derived from LIR )perunitarea.Note,however,that it was not obvious a priori that these quantities should be corre- and “starburstiness” in distant galaxies lated, since the stellar mass of most galaxies is dominated by the old stellar population, whereas the IR (or radio) size used to de- In the previous section, we have defined two modes of star formation: rive ΣIR measures the spatial distribution of young and massive stars. – anormalmodethatwecalledtheinfraredmainsequence,in Because the starburstiness and the IR8 ratio are both en- which galaxies present a universal IR8 bolometric correction hanced in compact star-forming galaxies, they are also corre- factor and a moderate star formation compactness, ΣIR,and lated as shown by Fig. 17. The fit to this correlation is given in – astarburstmode,identifiedbyanexcessSFRperunitstellar Eq. (12): mass, hence sSFR, as compared to the typical sSFR of most 1.2 RSB = (IR8/4) . (12) local galaxies. The dispersion in this relation is 0.3 dex. Hence we find that it Galaxies with an enhanced IR8 ratio were systematically found is mostly compact starbursting galaxies that present atypically to be forming their stars in the starburst mode and to show a 10 2 strong IR8 bolometric correction factors, although there is not strong star formation compactness (ΣIR > 3 10 L kpc− ). asharpseparationofbothregimes,butinsteadacontinuumof In order to separate these two modes of star× formation% in dis- values. tant galaxies as well, we first need to define the typical sSFR

A119, page 16 of 26 RADIO AND MID-INFRARED PROPERTIES OF COMPACT STARBURSTSU\LIRGs5 with high SSFRs 1140 BRANDL ET AL. Vol. 653

4 MURPHY ET AL.

Figure 4. The infrared-derived star formation rate of each galaxy Figure 5. The specific star formation rate of each galaxy plot- plotted against its stellar mass. The size of each plotting symbol ted as a function of radio spectral index (generally) measured be- is a function of the galaxies radio spectral index, with smaller sizes tween 1.49 and 8.44GHz (see Table 1). Sources having 8.44GHz indicating flatter spectral indices. The solid and dashed lines in- 4.5 dicate the main-sequence and 1σ (0.26dex) dispersion for z ∼ 0 brightness temperatures larger than 10 K, indicating the presence of an AGN, are identified by filled circles. We additional 9 µm-detected AKARI galaxies as measured by Elbaz et al. (2011, Fig. 11.—Comparison of the average starburst template (dotted line)from −1 identify those sources categorized as being AGN dominated by SFR/M∗ ∼ 0.25 Gyr ). Each of the infrared-luminous galax- their low 6.2 µmPAHEQW(circleswithcrosses);note,eachof Fig. 6 to the most extincted source within our sample, NGC 4945 (dashed line), ies in this sample is located well above the main-sequence rela- the sources with an 8.44 GHz brightness temperature larger than and to the ULIRG IRAS 08572+3915 (solid line), chosen as an extreme case of tion. Sources having 8.44GHz brightness temperatures larger than 4.5 4.5 10 Kalsohasa6.2µmPAHEQW< 0.27 µm. Thesolidline extinction from the sample of Spoon et al. (2006). The spectra have been nor- 10 K, indicating the presence of an AGN, are identified by filled is a fit to the non-AGN sources, as identified by either their high malized at 14.5 m. circles. 8.44 GHz brightness temperature or low 6.2 µmPAHEQW,ofthe α−αMS α−αMS form sSFR = η(1−e )+sSFRMS e (see §4.2), which is optical depth most likely indicates that the dust obscu- forced to fit through the typical radio spectral index (αMS ∼ 0.8) −1 starburst galaxies in Figure 10 do not show evidence for a similar ration in these sources occurs in the compact starburst and specific star formation rate (sSFRMS ∼ 0.25 Gyr )ofz ∼ 0 itself, and not by extended dust located in the foreground star-forming galaxies (diamond). The corresponding error bars in- emission structure. The 13Y35 m spectra, from NGC 1222 (top) galaxy disks. dicate the 1σ dispersions of these values. optical depth ← increasing to NGC 4945 at the bottom, show an increasingly pronounced depression, peaking at 18.5 m, signaling increasingly strong sili- 4.2. Radio Spectral Indices and the Main Sequence of the sample increases as the radio spectral index flattens. While a general trend does exist, it is worth noting that cate absorption. The latter result is in full agreement with the trend Star-Forming Galaxies Figureα is 2. correlatedThe optical depth ofwith the 9.7µ msilicatefeatureplot-radio extent found for the 9.8 m silicate feature. theted entire against sample the radio lies spectral well index above (generally) the z measured∼ 0 main between se- In Figure 4 we plot the z ∼ 0 main sequence rela- quence.1.49 and The 8.44 minimum GHz (see andTable mean 1). Sources specific with star 8.44 formation GHz bright- Among all the galaxies classified as starbursts, NGC 4945 tion as given in Elbaz et al. (2011), which is defined ratesness among temperatures the sample larger are than a 10 factor4.5 K, indicatingof ∼2.5 and the presence 15 times of (bottom spectrum in Fig. 10) is a ‘‘special case,’’exhibiting by an AGN, are identified by filled circles. We additionally identify using a local sample of galaxies detected at 9 µm with largerthose than sources the categorizedz ∼ 0 main as being sequence AGN value,dominated respectively. by their low far the strongest dust obscuration to its nuclear region (Spoon the AKARI/Infrared Camera (IRC; Onaka et al. 2007). 6.2 µmPAHEQW(circleswithcrosses);note,eachofthesources et al. 2000). Based on the IRS spectrum, the apparent optical However, many sources have radio spectral indices con- Fig. 10.—IRS spectra of Mrk 52, NGC 7714, NGC 1222, NGC 7252, NGC Most galaxies lie along a linear relation between star ...with form an 8.44 GHz brightnesstheir temperature stars larger in than warm, 104.5 Kalso compact, optically thick regions sistent with normal star-forming galaxies (i.e., α ∼ 0.8; 2146, NGC 520, NGC 3628, and NGC 4945 arranged to illustrate the gradual depth in the 9.8 m silicate feature is at least 4 and may be higher, formation rate and stellar mass, such that the specific has a 6.2 µmPAHEQW< 0.27 µm. Thedottedlineisanordi- Condonnary least 1992). squares Thus, fit to all it galaxies seems while that the a dashedflat radio line ex spec-cludes effect of increasing silicate absorption from top(Murphy to bottom. The fluxet densitiesal. 2013)depending on the choice of the local continuum. Apart from ∗ star formation rate (sSFR = SFR/M ) is constant, being tralsources index identifieddoes indicate as possible that AGN. a source Both fits is above illustrate the a gen maineral Fk have been arbitrarily scaled for better comparison. strong amorphous silicate absorption, the line of sight also reveals sSFR ∼ 0.25 Gyr−1 with a dispersion that is a factor of sequence,trend for but sources galaxies to have having decreasing a silicate “normal” optical star-forming depthsastheir radio spectra steepen. The source to the lower left, having the presence of a 23 m absorption feature. Following the analysis 1.82. It is worth noting that the z ∼ 0 main-sequence re- radioanegativeradiospectralindex,isUGC08058(Mrk231) spectra (i.e., α ∼ 0.8) are not necessarily on the of deeply obscured lines of sight toward ULIRG nuclei (Spoon log space, i.e., the IRS estimated infrared luminosities agree lation reported by Elbaz et al. (2007) has a slope that is mainand sequence. is not included in the ordinary least squares given that et al. 2006), we attribute the 23 m feature to crystalline silicates we only have a lower limit for its silicate optical depth. slightly flatter than unity, being 0.77. If we instead adopt Since we know that typical z ∼ 0 star-forming galaxies within 23% with LIR. This is much more accurate than the esti- (forsterite). Their detection in NGC 4945 suggests that crystal- this relation for the z ∼ 0 main sequence, our conclu- lying on the main sequence have radio spectral indices mates by Chary & Elbaz (2001) and Takeuchi et al. (2005). The line silicates are perhaps a more common component of the ISM sions are not strongly affected. Like Elbaz et al. (2011), near α ∼ 0.8, it may be more appropriate to fit the data excellent correlation suggests that, at least for a homogeneous Figure 1. Top: The linear 8.44GHz size of each source plotted and not just limited to ULIRG nuclei. we decideagainst to chose the radio the spectral main-sequence index (generally) relation measured having between a 1.49 with something other than an ordinary least squares fit sample of starburst galaxies, F15 m and F30 m can be used to and 8.44 GHz (see Table 1). Bottom: The same as the top panel, Figure 11 compares the average starburst template from Fig- slope of unity as this is consistent with findings at z ∼ 1 that passes through this normalization point. The radio accurately derive LIR. (Elbaz etexcept al. 2007), that thez 8.44∼ 2 GHz (Pannella brightness et temperature al. 2009), isz now∼ plott3 ed brightness temperature, which is a measure of the surface ure 6 to NGC 4945, the most extincted source within our sample. against the radio spectral spectral index. As already shown by For comparison we also show the heavily embedded ULIRG (MagdisCondon et al. 2010), et al. (1991), and z there∼ 4 is (Daddi a general et trend al. 2009). for sources having brightness or “compactness” of a source, is given by the 4.3. ALarge Variety in Silicate Absorption −τff −τff IRAS 08572+3915 from Spoon et al. (2006). The usually rather We overplotflat spectral our indices sample to be galaxies small in and size and find have that large each 8.44 GHz expression Tb = Te(1 − e )+Tb(0)e ,whereTe is brightness temperatures. In both panels, sources with 8.44 GHz Figure 10 shows, from top to bottom, a series of starburst spec- shallow 18 m silicate band reduces the continuum by a large source liesbrightness more thantemperatures 1σ away larger from than the 104.5 mainK, indicating sequence, the pres- the thermal electron temperature and τff is the free-free as expectedence of for an starbursting AGN, are identified galaxies. by filled The circles. size We of additional each ly optical depth. The specific star formation rate has been tra, sorted by increasing absorption of the 9.8 m silicate reso- factor compared to the average starburst spectrum. While NGC plottingidentify symbol those is a function sources categorized of the radio as being spectral AGN index dominated of by shown to be related to the compactness (brightness tem- nance. Since the peak of the resonance coincides with a minimum 4945 has a similarly strong silicate absorption feature as the their low 6.2 µmPAHEQW(circleswithcrosses);note,eachof the source,the sources with smaller with an 8.44symbols GHz brightness indicating temperature flatter spec- larger than perature) of a source (e.g., Elbaz et al. 2011), and in Fig- between the 7Y9and11Y13 mPAHemissioncomplexes,theef- ULIRG IRAS 08572+3915, the starburst also shows strong PAH tral indices.104.5 Kalsohasa6.2 Sources havingµmPAHEQW the flattest< 0 spectral.27 µm. indices ure 2 we have found a relation between the radio spectral fect of silicate absorption only becomes apparent toward the lower complexes near 16.4 and 17.1 m, which are absent in the ULIRG tend to be furthest from the main sequence. index and the optical depth of the 9.7 µm silicate feature. half of the plot. Nevertheless, the figure strikingly illustrates the spectrum. Crystalline silicates, causing the features at 16 and 19 m In Figure3.2. 5Identifying we investigate Potential the AGN-Dominatedrelation between Sources the Thus, let us assume that the specific star formation rate strong effect of amorphous silicates on the overall 5Y38 mspec- in IRAS 08572+3915, appear to be absent in the starburst tem- distance ofIn these Figure galaxies 3 we look from at the the main 6.2 µm sequence PAH EQW, versus which is related to the radio spectral index with a function hav- tral shape and the large variations even within one class of objects. plate, although they are more difficult to discern at our spectral α−αMS α−αMS the compactnessmeasures the of the relative starburst strength as of measured the PAH by flux their with re- ing theFigure form 3. ofThe sSFR 6.2 µmPAHEQWplottedagainsttheradiospec- = η(1−e )+sSFRMS e , Figure 10 shows that the 10 m trough can in fact become a dom- resolution given the presence of PAH features and the [S iii] line radio spectralspect to index. the amount Since of we hot have dust adopted emission a at constant 6.2 µm, ver- wheretralα indexMS and (generally) sSFRMS measuredare the between typical 1.49 radio and spectral 8.44GHz (sin-ee Table 1). The horizontal lines indicate regions for which sources inating feature of the spectral shape of starbursts. in that range. specific starsus radio formation spectral rate index. to define The the 6.2zµmPAHEQWhas∼ 0 main se- dicesare and thought specific to be starAGN formationor star formation rates driven for basedgalaxies on their on m theea- The top two spectra also show a broad emission structure be- While NGC 4945 shows by far the most extreme silicate ab- quence,been the distance used to indicate from the the main presence sequence of an is AGN given since by they z ∼ sured0 main 6.2 µ sequence,mPAHEQW.Sourceshavingspectralindicessteeper respectively, and η is a normaliza- 8.44GHz ginning at about 16 mandextendingto21Y25 m. The shape sorption within our sample, it is not an exotic object but defines the specifictypically star have formation very small rate PAH itself. EQWs We (e.g., find Genzel a gen- et al. tionthan constantα . related! 0.6neverappeartohaveextremelylow6.2 to the maximum specific star for-µm 1 49GHz of this feature is consistent with that of an 18 msilicateemis- the endpoint of a sequence of increasing optical depths. Recently, eral trend1998; where Armus the et specific al. 2007). star Specifically, formation rate starburst among domi- mationPAH rate, EQWs which (i.e., "0.1 isµ determinedm). Sources with by 8.44 the GHz fit to brightness the data. tem- nated systems appear to have 6.2 µmPAHEQWsthat peratures larger than 104.5 K, indicating the presence of an AGN, sion feature, which would make these the first detections of this Dale et al. (2006) reported on the Spitzer SINGS survey of 75 are identified by filled circles. The source to the lower left, having feature in starburst galaxies. However, the emission structure nearby galaxies. The typical target in their sample shows only are ! 0.54 µm (Brandl et al. 2006), while AGN have anegativeradiospectralindex,isUGC08058(Mrk231). 6.2 µmPAHEQWsare" 0.27 µm. We find that the could also be due to the C C Cin-planeandout-of-planebending modest dust obscuration, consistent with an AV 1foreground 8.44GHz À À galaxies having steep spectral indices (i.e., α1.49GHz ! the AGN dominated systems, we find a significantly flat- modes of PAHs (Van Kerckhoven et al. 2000). The identification screen and a lack of dense clumps of highly obscured gas and dust. 0.6) never have extremely low 6.2µm PAH EQWs (i.e., ter mean spectral index of 0.45, albeit with a much larger of this feature will be discussed in a future paper. The remaining Unfortunately, we cannot distinguish between the local and global "0.1 µm). Additionally, all 3 sources identified as AGN scatter of 0.27. via their excessively large 8.44 GHz brightness tempera- 4. ture sit in the PAH-defined AGN region. DISCUSSION Using the 6.2 µm PAH EQW to split the sources up Taking archival high resolution 8.44 GHz data from the into starburst and AGN dominated systems, we find that literature, we have compared the radio and mid-infrared the mean radio spectral index for starburst dominated spectral properties for a sample of local (U)LIRGs. We systems is 0.74, with a standard deviation of 0.15. For now use these results to investigate how the dust is dis- (3) The CO Spectral Line Energy Distribution - Mrk 231 A&A 518, L42 (2010)

comparison with the SPIRE spectrum of the Orion bar PDR + + (Habart et al. 2010), which shows no trace of OH or H2O , and only weak CH+ emission, while in Mrk 231 the lines are only a factor 2 3 fainter than the CO lines. While enhanced cosmic ray ﬂuxes− in a starburst environment will increase the de- + + gree of ionization and hence the production of OH and H2O in a PDR, they do not elevate the temperatures to the level required to produce the highest CO lines (Meijerink et al. 2006). It is thus the combination of strong high-J CO lines and high OH+ + and H2O abundances that reveals X-ray driven excitation and chemistry in Mrk 231.

4. Outlook We have shown that the SPIRE spectrum of Mrk231 reveals both PDR and XDR emission lines, and made a separation of Fig. 2. Luminosities of CO lines from Mrk 231. Filled symbols repre- these components. A key goal of the HerCULES project will be •Transitions up to J = 14→sent13 measurements observed from the SPIRE FTS spectrum, while ground-based using this decomposition for a quantitative separation between measurements are denoted with open symbols. Coloured lines indicate star formation and black hole accretion as power sources for the •Model: dense gas clump intwo a model560 PDR pc components radius (red star-forming and green lines) and an XDRdisk compo- infrared luminosities of dusty galaxies. In the case of Mrk 231, nent (blue line). The sum of these three components is indicated by the this issue will be addressed in a forthcoming paper (Meijerink exposed to UV + X-rays fromblack line an and AGN fits the CO heating measurements. the In the legend,innern denotes 160 the pc. 3 et al., in prep.). Data will be obtained as part of the HerCULES number density of hydrogen nuclei (n = nH +2nH2 )incm− , G0 denotes 3 1 2 program for an additional 28 objects, which will enable us to put the incident UV flux in units of 1.6 10− erg s− cm− for the PDRs, × the results presented here on a statistically significant footing. and FX the incident X-ray flux for the XDR. The legend also indicates the relative emitting areas of(van the three der components. Werf et al. 2010) Acknowledgements. We thank Ewine van Dishoeck, Xander Tielens, and Thomas Nikola for useful discussions. We especially thank Ed Polehampton, Peter Imhof-Davies and Bruce SwinyardfortheirhelpwiththeFTSdatapro- the fraction of gas within 0.3pcfromanO5starinMrk231 cessing. JF thanks MPE for its hospitality. The Dark Cosmology Centre is to much less than the 50% required to produce the highest funded by the DNRF. The following institutes have provided hardware and soft- CO lines with a high excitation∼ PDR. This problem does not ware elements to the SPIRE project: University of Lethbridge, Canada; NAOC, exist for the XDR model, which predicts most of the dust to be Beijing, China; CEA Saclay, CEA Grenoble and LAM in France; IFSI, Rome, and University of Padua, Italy; IAC, Tenerife, Spain; Stockholm Observatory, cooler. Sweden; Cardiff University, Imperial College London, UCL-MSSL, STFC- RAL, UK ATC Edinburgh, and the University of Sussex in the UK. Funding for SPIRE has been provided by the national agencies of the participating coun- 3.2. XDR chemistry tries and by internal institute funding: CSA in Canada; NAOC in China; CNES, CNRS, and CEA in France; ASI in Italy;MECinSpain;StockholmObservatory The extraordinarily luminous emission from the molecular ions in Sweden; STFC in the UK; and NASA in the USA. Additional funding support + + H2O and OH reveals the chemical signature of an XDR. for some instrument activities has been provided by ESA. Assuming that the emission arises from a disk with 160 pc radius (as derived above), we can derive column densities in the References upper levels of the relevant transitions, which results in values 13 2 Nup 3.0 3.5 10 cm− for both species. Modeling the nu- Aalto, S., Spaans, M., Wiedner, M. C., & Hüttemeister, S. 2007, A&A, 464, 193 clear∼ molecular− × gas disk in Mrk 231 with a radius of 520 kpc and Boksenberg, A., Carswell, R. F., Allen, D. A., et al. 1977, MNRAS, 178, 451 9 Braito, V., Della Ceca, R., Piconcelli, E., et al. 2004, A&A, 420, 79 H2 mass of 2.2 10 M then results in lower limits to the to- Davies, R. I., Tacconi, L. J., & Genzel, R. 2004, ApJ, 613, 781 + ×+ $ 10 tal OH and H2O abundances relative to H2 of 2 10− in the Downes, D., & Solomon, P. M. 1998, ApJ, 507, 615 central 160 pc. Given the short radiative lifetimes∼ (×<60 s) of the Fixsen, D. J., Bennett, C. L., & Mather, J. C. 1999, ApJ, 526, 207 + + González-Alfonso, E., Smith, H. A., Ashby, M. L. N., et al. 2008, ApJ, 675, 303 upper levels involved, most H2O and OH molecules will be in the ground state, and total abundances will exceed these lower González-Alfonso, E., et al. 2010, A&A, 518, L43 Griffin, M. J., et al. 2010, A&A, 518, L3 limits by large factors. Such abundances require an efficient and Habart, E., et al. 2010, A&A, 518, L116 penetrative source of ionization in the molecular gas, since the Meijerink, R., & Spaans, M. 2005, A&A, 436, 397 production of OH+ is mainly driven by H+ + O O+ + Hfol- Meijerink, R., Spaans, M., & Israel, F. P. 2006, ApJ, 650, L103 + + + → + Meijerink, R., Spaans, M., & Israel, F. P. 2007, A&A, 461, 793 lowed by O + H2 OH + HandH + OH OH + H. In the models considered→ in Sect. 3.1,thekeyspeciesH→ + is 2 3orders Papadopoulos, P. P., Isaak, K. G., & van der Werf, P. P. 2007, ApJ, 668, 815 − Pilbratt, G. L., et al. 2010, A&A, 518, L1 of magnitude more abundant in the XDR model than in the high Poglitsch, A., et al. 2010, A&A, 518, L2 excitation PDR, and the OH+ abundance is larger by a compara- Sanders, D. B., Mazzarella, J. M., Kim, D.-C., Surace, J., & Soifer, B. T. 2003, + ApJS, 126, 1607 ble amount. A similar argument can be made for H2O ,whichis + + Spaans, M., & Meijerink, R. 2008, ApJ, 678, L5 formed by OH + H2 H2O + H. → + + Swinyard, B. M., et al. 2010, A&A, 518, L4 The extraordinary luminosity of the OH and H2O (and Taylor, G. B., Silver, C. S., Ulvestad, J. S., & Carilli, C. L. 1999, ApJ, 519, 185 + to a lesser extent CH )linesinMrk231isunderlinedbya Veilleux, S., Rupke, D. S. N., Kim, D., et al. 2009, ApJS, 182, 628

Page 5 is available in the electronic edition of the journal at http://www.aanda.org

Page 4 of 5 The Herschel-ALMAOverlap

Band 8

Band 9

Band 10 Mrk 231 (van der Werf etal. 2010) Early Science Full Science (4) ALMA observations of IIZw096 far-UV near-IR 1.6μm mid-IR

5 kpc

•>95% of the emission centered on a single, resolved source •If the molecular gas is distributed in a rotating disk, CO(6→5) 9 Mdyn = 3.3 x 10 Msun

(Inami et al. 2010; Stierwalt et al. 2013) ALMA observations of IIZw096 far-UV near-IR 1.6μm mid-IR

5 kpc

LIR size μIR Source -2 L⨀ pc L⨀ kpc IIZw096 7x1011 277x185 4.3x1012

Orion core 2x105 0.3 3x1012

→ CO(6 5) M82 3x1010 450x75 9x1011

(Inami et al. 2010; Stierwalt et al. 2013) No. 2, 2009 P CYGNI PROFILES TOWARD ARP 220 NUCLEI L105

Table 1 Observed Parameters of Arp 220

a Parameter East West All ftot 1 2 SCO(3 2) (Jy km s− )250301(7.0 1.1) 10 0.23 0.06 − 1 ± × 2 ± SHCO+(4 3) (Jy km s− ) 19 61 (1.1 0.2) 10 1.0 0.3 − 1 ± × 2 ± SHCO+(3 2) (Jy km s− ) 35 63 (1.0 0.1) 10 0.64 0.13 − ± × ± S0.87 mm (Jy) 0.18 0.49 0.68 0.10 1.0 0.2 ± ± S1.1mm (Jy) 0.07 0.20 0.27 0.03 0.82 0.15 ± ± CO(3–2) max Tb (K) 58 (66) 40 (47) + ··· ··· HCO (4–3) max Tb (K) 16 (24) 26 (34) + ··· ··· HCO (3–2) max Tb (K) 12 (17) 15 (21) ··· ··· 0.87 mm max Tb (K) 16 (23) 42 (50) ··· ··· 1.1 mm max Tb (K) 7 (12) 20 (25) 1 ··· ··· Vsys(radio, LSR) (km s )541515 5355 15 − ± ± ··· ··· Notes. We conservatively estimate flux-calibration 1σ errors to be 10% and 15% for 0.86 and 1.1 mm bands, respectively. Line fluxes are 1 1 + integrated over 4800–5900 km s− (5200–5900 km s− for HCO (3–2)) including both emission and absorption. Peak brightness temperatures are measured at or around each nucleus without correcting for beam dilution or missing flux. The first one is Rayleigh–Jeans Tb and the second in parentheses is Planck Tb. a Fraction of the flux (or flux density) recovered in our 0. 3 resolution data with respect to the total flux of Arp 220. The total fluxes are 106 23 ## ± Jy km s 1 for HCO+(4–3) (Greve et al. 2009),outflows: (3.0 0.6) 103 Jy km s 1 for CO(3–2) in thee.g., central 14 (Wiedner SMA et al. 2002), 152 26 Jyresults − ± × − ## ± km s 1 for HCO+(3–2) (this work; SMA subcompact array observations with the minimum projected baseline of 9 m), 0.7 0.1Jyfor0.87 − ± mm continuum (S08 and references therein), and 0.33 0.05 Jy for 1.1 mm continuum (this work; SMA subcompact configuration). ± L106 SAKAMOTO ET AL. Vol. 700

200CO data.Arp 220 This E proﬁle is consistent with the previous CO(2–1) 200 Arp 220 W (a) Arp 220 (b) Arp 220 + + + + proﬁleHCO in(4-3) Downes & Eckart (2007)exceptthattheabsorptionis HCO (4-3) 1 HCO (4-3) HCO (3-2) 100deeper in our CO(3–2) data. The deep absorption is almost at the 100 systemic0 velocity of the nucleus, marked with a dashed0.0 line in 0 0.0 -100Figure 2, as expected if the CO-absorbing gas is in-0.5 circular -100

0 1 σ 4500 4700 4900 1 σ 4500 4700 4900 rotation in the western nuclear disk. TheHCN(4-3) systemic Velocity velocity,-1.0 HCN(4-3) Velocity -200 1 -200 -0.5 Vsys(W) 5355 15 km s− ,wasestimatedfromthecentral ] ] -1 100velocityoftheCOproﬁleinthelow-resolution(0HCO+=(3-2) ± .5) data of S08,-1 100 HCO+(3-2) Dec. offset [arcsec] ## -1 measured at half-maximum. The HCO+(4–3) and (3–2) lines 0.33" x 0.28" 300 pc 0.45" x 0.34" 300 pc are0 also in absorption at the systemic velocity. In addition,0.0 both 0 0.0 + 1 -0.5 1 -0.2 HCO linesσ have absorption at blueshifted velocities while-1.0 most σ (c) Arp 220 (d) Arp 220 + 1 of their emission appears at redshifted velocities. The HCO (3– CO(3-2) 0.87 mm CO(3-2) CO(3-2) 4002) proﬁle shows multiple velocity components in absorption; 400

Intensity [mJy beam CO(3-2), 0.5" data Intensity [mJy beam CO(3-2), 0.5" data 1 the minima at 4890, 5220, and 5340 km s− are at 3.6σ 4.6σ, and 4.6σ,respectively.Absorptiondeeperthan3σ is seen down 0 200 1 200 to Vsys(W) 495 km s− . The eastern− nucleus has very asymmetric line profiles with respect0 to its systemic velocity, which we estimate to be 0 Dec. offset [arcsec] 0.0 0.0 -1 1 Vsys(E) 5415 15 km s− in the same way as for the western-0.5 -0.2 1 σ 1 σ 0.34" x 0.28" 0.34" x 0.28" 300 pc = ± -1.0 nucleus. In all the observed lines, absorption is predominantly -0.4 1 0 -1 1 0 -1 seen at or4500 blueward 5000 of the systemic 5500 velocity 6000 while 6500 most emission 4500 5000 5500 6000 6500 R.A. offset [arcsec] R.A. offset [arcsec] is in redshifted velocities,Velocity (radio, making LSR) the[km profiless-1] P Cygni type. The Velocity (radio, LSR) [km s-1] + 1 Figure 1. Integrated intensity maps of HCO (4–3), (3–2), and CO(3–2) linesFigure 2.deepestMolecular-line absorption spectra is at at the 45–75 two nuclei km s of− Arpbelow 220.V Continuumsys(E) and has is 2.8 beenσ , subtracted. The right ordinate is the fractional absorption depth fa with respect 1 + and 0.87 mm continuum. The line moment-0 maps were made using positiveto the continuum intensity of each nucleus,+ and can be converted to optical depth via τa log( fa). The 1σ noise is 32, 12, and 28 mJy beam for HCO (4–3), 4.5σ, and 3.0σ in HCO (4–3), (3–2), and CO(3–2), respectively. = − − − signals only. The lowest contour is at 2σ and the nth contour is at 9.5n,4.3(3–2),n, and CO(3–2), respectively. The CO(3–2) spectra overplotted in black are from 0"". 5resolutiondatainSakamotoetal.(2008). The systemic velocity that we 1 1 1.5 and 10.3n Jy beam− km s− in (a), (b), and (c), respectively, and at 14.4nestimated for each nucleus is shown by a dashed line. 1 3.3. Absorption Depth mJy beam− in (d). Crosses are at continuum peaks and filled ellipses show synthesizedgas beams. disk has an inner regionAbsorption (i.e., with negative intensity) significant in our continuum- outflow motion • subtracted data means absorption of the continuum, because of 3.7 (or >1.4 when allowing for 1σ error) toward the eastern diative transfer of both molecules, including radiative coupling there in conditions favorable for the transitions of high critical line self-absorption alone cannot produce negative intensity. In nucleus. The total line absorption τa dV integrated over the to the intense infrared continuum, will be explored in a future density. order to absorb continuum the excitation temperature+ of the and outer region dominatedvelocitiesforeground where τa gas! must0.1is,intheorderofHCO by be below rotation the brightness temperature(4–3), ofpublication. + ! 1 HCO (3–2),the background and CO(3–2), continuum. 75, 150, The and (deconvolved) 50 km s− westerntoward nucleus 3.2. P Cygni Line Profiles at Both Nuclei the eastern nucleus and 85, 82, and 15 km s 1 toward the 3.4. Gas Motion—Rotation and Outflow has Tb 90–160 K and a size of 50–80 pc− in 860 µmcontin- •Thedense line spectra at the absorbing two nuclei, in Figure 2,showboth gaswestern. deepuum The (S08;= absorption for with the spectral profiles energy varythe among distribution gas the lines, (SED) disk more of the two Rotation of gas around each nucleus is evident in the CO(3–2) emission and absorption. In particular, all the observed linesnotablynuclei toward see the Matsushita western nucleus. et al. 2009 There). The are high twoTb caveats,duetohighposition–velocity diagrams (Figure 3)asthenegativetopositive except CO(3–2) toward the western nucleus show blueshiftedto keepdust in mind. opacity First, in submillimeter, our observations makes provide line lowerabsorption limits againstshift of V Vsys across the nucleus. The CO emission ranges •source of accelerationto total absorption - mechanical because emission from the absorbingenergy gas from− supernovae1 absorption and redshifted emission, i.e., P Cygni profiles. This dust continuum more likely to occur (and deeper) than at longerfrom 5000 to 5900 km s− around the western nucleus. The 1 feature is absent in our calibrator spectra. itself andwavelengths. from ambient (i.e., nonabsorbing) gas in our beam high-velocity gas around 5800–5900 km s− is detected at >3σ The western nucleus has a narrow, deep absorption featuremasks at the absorption.The maximum Second, apparent our synthesizedoptical depth beam of each is larger absorption for linefor the first time. The fall-off of the rotation velocity away from or radiation pressure+ from an+ AGN 1 + 5340 km s− and emission at higher and lower velocities in ourHCO (3–2)is in than the range for HCO of 0.4–1.0(4–3) andexcept CO(3–2). the HCO (3–2) optical depththe nucleus agrees with the CO(2–1) observations by Downes & The absorbing column density NH2 is on the order of Eckart (2007)andsuggeststruncationofthemassdistribution 22 2 23 2 + 10 cm− for the CO absorption and 10 cm− for the HCO of the nucleus beyond a "100 pc core. if the absorbing gas is in 50 K LTE and abundances are Radial motion of gas, in addition to the rotation, is suggested 4 + 8 [CO/H2] 10− and [HCO /H2] 10− .Thehighabun- by the blueshifted absorption mentioned above. The absorbing ≈ + ≈ dance adopted for HCO is consistent with recent models of (Sakamotogas must be in front ofet the al. continuum 2009) core of each nucleus the photon-dominated region (PDR) and the X-ray-dominated and moving away from it toward us. Although absorption region (XDR) (Meijerink & Spaans 2005), as well as models of only tells us the gas motion along our lines of sight to the HCO+ enhancement in protostellar outflow shocks (Rawlings continuum nuclei, the detection of similar blueshifted absorption et al. 2004). The absorbing gas will have a depth on the order on both nuclei whose nuclear disks are misaligned suggests of 1–10 pc if its density is at the CO(3–2) critical density of that the gas motion is more likely a radial outflow in most 4 3 10 cm− .Theactualabsorbingmaterialcanbedistributed directions than noncircular motion only along our sight lines. over a longer length if it is inhomogeneous as in numerical The redshifted emission in the P Cygni profiles must then hydrodynamical simulations of multiphase gas obscuring active contain emission from gas on the far side of the continuum nuclei nuclei (e.g., Wada & Norman 2002). The column densities of and moving away from the nuclei and us. The other emission the absorbing gas calculated here have large uncertainties be- in the spectra must be from the gas rotating in the nuclear cause the absorption data provide lower limits, gas excitation disks. conditions are poorly known, and molecular abundances can The typical outflow velocity of the absorbing gas is probably 1 significantly vary. Still, it is possible for the absorbing column about 100 km s− because the deepest absorption in the P Cygni 1 density to take these values because they are smaller than the spectra is between (Vsys 45) and (Vsys 135) km s− . High- 25 2 total column density of the nuclei ( 10 cm− ; S08). The re- velocity components up− to 500 km s 1−are suggested toward ∼ − markable detection of vibrationally excited HCN in absorption the western nucleus by the∼ HCO+(3–2) absorption with large in Arp 220 (Salter et al. 2008)suggeststhattheexcitationof velocity offsets. The absorbing gas is turbulent or has a velocity HCO+ may also be more unusual than described by our sim- gradient across each continuum nucleus because absorption is ple analysis. More realistic models of the excitation and ra- also seen at slightly positive velocities with respect to systemic. Summary • U\LIRGs have complex substructures which vary as a function of wavelength • Mid-IR diagnostics has provided insight to the relative importance of starburst and AGN to the IR output of U/ LIRGs • At the high luminosity end, the star formation in U/LIRGs is compact and highly dust-enshrouded • ALMA will enable the study of the star-forming molecular gas and dust, the importance of starburst and AGN heating to the chemistry observed in molecular clouds, the extent of the star-forming regions, the feedback in the form of molecular outflows