FAR-INFRARED LINE SPECTRA OF ACTIVE GALAXIES FROM THE HERSCHEL/PACS SPECTROMETER: THE COMPLETE DATABASE

, , , , , and

Published 2016 October 12 © 2016. The American Astronomical Society. All rights reserved.
, , Citation Juan Antonio Fernández-Ontiveros et al 2016 ApJS 226 19 DOI 10.3847/0067-0049/226/2/19

0067-0049/226/2/19

ABSTRACT

We present a coherent database of spectroscopic observations of far-IR fine-structure lines from the Herschel/Photoconductor Array Camera and Spectrometer archive for a sample of 170 local active galactic nuclei (AGNs), plus a comparison sample of 20 starburst galaxies and 43 dwarf galaxies. Published Spitzer/IRS and Herschel/SPIRE line fluxes are included to extend our database to the full 10–600 μm spectral range. The observations are compared to a set of Cloudy photoionization models to estimate the above physical quantities through different diagnostic diagrams. We confirm the presence of a stratification of gas density in the emission regions of the galaxies, which increases with the ionization potential of the emission lines. The new [O iv]${}_{25.9\mu {\rm{m}}}$/[O iii]${}_{88\mu {\rm{m}}}$ versus [Ne iii]${}_{15.6\mu {\rm{m}}}$/[Ne ii]${}_{12.8\mu {\rm{m}}}$ diagram is proposed as the best diagnostic to separate (1) AGN activity from any kind of star formation and (2) low-metallicity dwarf galaxies from starburst galaxies. Current stellar atmosphere models fail to reproduce the observed [O iv]${}_{25.9\mu {\rm{m}}}$/[O iii]${}_{88\mu {\rm{m}}}$ ratios, which are much higher when compared to the predicted values. Finally, the ([Ne iii]${}_{15.6\mu {\rm{m}}}\,$+ [Ne ii]${}_{12.8\mu {\rm{m}}}$)/([S iv]${}_{10.5\mu {\rm{m}}}\,$+[S iii]${}_{18.7\mu {\rm{m}}}$) ratio is proposed as a promising metallicity tracer to be used in obscured objects, where optical lines fail to accurately measure the metallicity. The diagnostic power of mid- to far-infrared spectroscopy shown here for local galaxies will be of crucial importance to study galaxy evolution during the dust-obscured phase at the peak of the star formation and black hole accretion activity ($1\lt z\lt 4$). This study will be addressed by future deep spectroscopic surveys with present and forthcoming facilities such as the James Webb Space Telescope, the Atacama Large Millimeter/submillimeter Array, and the Space Infrared telescope for Cosmology and Astrophysics.

Export citation and abstract BibTeX RIS

1. INTRODUCTION

Rest-frame mid- to far-infrared (IR) spectroscopy is a unique tool to study dust-enshrouded galaxies and, in particular, to disentangle the emission produced by star-forming activity from emission generated by accretion onto supermassive black holes in the nuclei of active galaxies (e.g., Spinoglio & Malkan 1992). Optical/UV lines provide access only to relatively unobscured gas, hampering our ability to investigate the physics in obscured regions. The extinction is, however, negligible in the mid- to far-IR range, which contains several lines that provide information regarding the physical conditions even for heavily obscured gas (see Tables 1 and 2). Therefore, mid- to far-IR lines are the key to investigate not only "star formation" in galaxies, but, in general, all the processes in the majority of galaxies that occur in a dust-embedded phase, and thus are hidden from optical studies. In the dark side of star formation and active galactic nucleus (AGN) activity, we find critical aspects of these phenomena: e.g., deeply buried active nuclei, the onset of the star formation, and AGN feeding and feedback through gas inflow and outflow. These processes have a major impact on galaxy evolution but cannot be addressed from an optical/UV perspective.

Table 1.  Fine-structure Lines in the Mid- to Far-IR Range

Line λ ν I.P. E ${n}_{\mathrm{cr}}$ Spec. Res. Ang. Res. # of AGN # of SB # of DW
  (μm) (GHz) (eV) (K) (${\mathrm{cm}}^{-3}$) (km s−1) (arcsec)      
[S iv]2P${}_{3/2}{\mbox{--}}^{2}$ P${}_{1/2}$ 10.51 28524.50 34.79 1369 5.39 × 104 ∼500 $4.7\times 11.3$ 112 19 30
[Ne ii]2P${}_{1/2}{\mbox{--}}^{2}$ P${}_{3/2}$ 12.81 23403.00 21.56 1123 7.00 × 105 ∼500 $4.7\times 11.3$ 172 20 21
[Ne v]3P${}_{2}{\mbox{--}}^{3}$ P1 14.32 20935.23 97.12 1892 3 × 104 ∼500 $4.7\times 11.3$ 115 0 1
[Ne iii]3P${}_{1}{\mbox{--}}^{3}$ P2 15.56 19266.87 40.96 925 2.68 × 105 ∼500 $4.7\times 11.3$ 172 19 29
[S iii]3P${}_{2}{\mbox{--}}^{3}$ P1 18.71 16023.11 23.34 769 2.22 × 104 ∼500 $4.7\times 11.3$ 129 20 26
[Ne v]3P${}_{1}{\mbox{--}}^{3}$ P0 24.32 12326.99 97.12 596 5.0 × 105 ∼500 $11.1\times 22.3$ 90 0 0
[O iv]2P${}_{3/2}{\mbox{--}}^{2}$ P${}_{1/2}$ 25.89 11579.47 54.94 555 104 ∼500 $11.1\times 22.3$ 147 12 17
[S iii]3P${}_{1}{\mbox{--}}^{3}$ P0 33.48 8954.37 23.34 430 7.04 × 103 ∼500 $11.1\times 22.3$ 110 20 21
[Si ii]2P${}_{3/2}{\mbox{--}}^{2}$ P${}_{1/2}$ 34.81 8612.25 8.15 413 3.4 × 105a, 103 ∼500 $11.1\times 22.3$ 100 20 20
[O iii]3P${}_{2}{\mbox{--}}^{3}$ P1 51.81 5787.57 35.12 441 3.6 × 103 ∼105 9.4b 19 1 0
[N iii]2P${}_{3/2}{\mbox{--}}^{2}$ P${}_{1/2}$ 57.32 5230.43 29.60 251 3.0 × 103 ∼105 9.4 32 5 3
[O i]3P${}_{2}{\mbox{--}}^{3}$ P1 63.18 4744.77 228 4.7 × 105a $\sim 86$ 9.4 115 20 31
[O iii]3P${}_{1}{\mbox{--}}^{3}$ P0 88.36 3393.01 35.12 163 510 ∼124 9.4 83 16 38
[N ii]3P${}_{2}{\mbox{--}}^{3}$ P1 121.90 2459.38 14.53 118 310 ∼290 9.4 79 17 8
[O i]3P${}_{1}{\mbox{--}}^{3}$ P0 145.52 2060.07 98 9.5 × 104a ∼256 10.3 66 8 12
[C ii]2P${}_{3/2}{\mbox{--}}^{2}$ P${}_{1/2}$ 157.74 1900.54 11.26 91 50, 2.8 × 103a ∼238 11.2 159 17 41
[N ii]3P${}_{1}{\mbox{--}}^{3}$ P0 205.3 1460.27 14.53 70 48 ∼297 ∼16.8 60 13 2
[C i]3P${}_{2}{\mbox{--}}^{3}$ P1 370.37 809.44 38.9 2.8 × 103a ∼536 ∼35.0 59 13 1
[C i]3P${}_{1}{\mbox{--}}^{3}$ P0 609.7 491.70 23.6 4.7 × 102a ∼882 ∼35.0 32 12 0

Notes. The columns correspond to the central wavelength, frequency, ionization potential, excitation temperature, critical density, spectral and spatial resolution of the data presented in this work, and the number of AGN, starburst, and dwarf galaxies with line detections above 3× rms. Critical densities and excitation temperatures are from Launay & Roueff (1977), Tielens & Hollenbach (1985), Greenhouse et al. (1993), Sturm et al. (2002), Cormier et al. (2012), and Farrah et al. (2013).

aCritical density for collisions with hydrogen atoms. bThe beam size for Herschel/PACS is dominated by the spaxel size ($9\buildrel{\prime\prime}\over{.} 4$) below $\sim 120\,\,\mu {\rm{m}}$.

Download table as:  ASCIITypeset image

Table 2.  Summary of the Line Ratios Discussed in the Paper and Their Associated Diagnostic

Line Ratio Diagnostic Parameter range
[N ii]${}_{205/122}$ Density of ionized gas $1\mbox{--}{10}^{3}\,{\mathrm{cm}}^{-3}$
[O iii]${}_{88/52}$ Density of ionized gas $10\mbox{--}{10}^{4.5}\,{\mathrm{cm}}^{-3}$
[S iii]${}_{33.5/18.7}$ Density of ionized gas $10\mbox{--}{10}^{5}\,{\mathrm{cm}}^{-3}$
[Ne v]${}_{24.3/14.3}$ Density of ionized gas $100\mbox{--}{10}^{6}\,{\mathrm{cm}}^{-3}$
[S iv]${}_{10.5}$/[S iii]${}_{18.7}$ Ionization parameter $34.8\mbox{--}23.3\,\mathrm{eV}$
[O iii]88/[O i]63 Ionized/neutral gas ratio
[O iii]88/[N ii]122 Ionization parameter $35.1\mbox{--}14.5\,\mathrm{eV}$
[Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ Ionization parameter $41.0\mbox{--}21.6\,\mathrm{eV}$
[O iv]${}_{25.9}$/[O iii]88 Ionization parameter, AGN/Starburst $54.9\mbox{--}35.1\,\mathrm{eV}$
[O i]${}_{145/63}$ Temperature of neutral gas (PDR) $100\mbox{--}400\,{\rm{K}}$
[C i]${}_{609/371}$ Temperature of neutral gas (PDR) $20\mbox{--}100\,{\rm{K}}$
[C ii]158/[O i]63 PDR density
[C ii]158/[N ii]122 PDR/low-excitation ionized gas contribution
[N ii]122/[C i]371 Low-excitation ionized gas/neutral gas contribution
[N ii]122/[C i]371 Low-excitation ionized gas/neutral gas contribution
[O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ AGN/Starburst contribution
[O iv]${}_{25.9}$/([Ne ii]${}_{12.8}$+[Ne iii]${}_{15.6}$) AGN/Starburst contribution
[O iii]88/[N iii]57 Ionization parameter and metallicity $35.1\mbox{--}29.6\,\mathrm{eV}$
([Ne ii]${}_{12.8}\,+\,$[Ne iii]${}_{15.6}$)/([S iii]${}_{18.7}\,+\,$[S iv]${}_{10.5}$) Metallicity $\sim 0.07\mbox{--}2.5\,{Z}_{\odot }$

Download table as:  ASCIITypeset image

The potential of mid- to far-IR spectroscopy was already proven by the Infrared Space Observatory (ISO; Kessler et al. 1996), using the Short Wavelength Spectrometer (SWS; de Graauw et al. 1996) and the Long Wavelength Spectrometer (LWS; Clegg et al. 1996). Because mid- and far-IR fine-structure emission lines are sensitive to the physical conditions of the interstellar medium (ISM), these transitions can be used to investigate its different phases (e.g., Sturm et al. 2000; Negishi et al. 2001; Spinoglio et al. 2005) and characterize the primary spectrum of the ionizing radiation (e.g., Alexander et al. 1999). ISO enabled the development of the first IR diagnostics to separate the AGN, the ISM, and the stellar contributions (e.g., Sturm et al. 2002; Brauher et al. 2008), and to shed light on the nature of ultra-luminous IR galaxies (ULIRGs; Genzel et al. 1998).

The full exploitation of the mid-IR window was possible thanks to the InfraRed Spectrograph (IRS; Houck et al. 2004) on board the Spitzer Space Telescope (Werner et al. 2004). The study of large samples of galaxies including ULIRGs, Quasars, Seyfert galaxies, radio, and starburst galaxies (Brandl et al. 2006; Armus et al. 2007; Bernard-Salas et al. 2009; Veilleux et al. 2009; Tommasin et al. 2010) showed, e.g., the ability of the [Ne v]${}_{14.3,24.3\mu {\rm{m}}}$ and the [O iv]${}_{25.9\mu {\rm{m}}}$ lines, associated with high-density and high-excitation photoionized gas (${n}_{{\rm{e}}}\gtrsim {10}^{3}\,{\mathrm{cm}}^{-3}$, I.P. $\gtrsim \,54\,\mathrm{eV}$) to investigate (1) the inner part of the narrow-line region (NLR) and identify optically hidden AGNs (Armus et al. 2007), (2) the AGN contribution in ULIRGs (Veilleux et al. 2009), (3) the AGN nature of radio-galaxies (Haas et al. 2005; Leipski et al. 2009), (4) the power of the extinction-free density tracers in the 102–104 cm−3 range based on the [S iii]${}_{18.7,33.5\mu {\rm{m}}}$ and the [Ne v]${}_{14.3,24.3\mu {\rm{m}}}$ lines (Tommasin et al. 2008, 2010), (5) star formation tracers based, e.g., on [Ne ii]${}_{12.8\mu {\rm{m}}}$ and [Ne iii]${}_{15.6\mu {\rm{m}}}$ (Ho & Keto 2007), or polycyclic aromatic hydrocarbon (PAH) features (O'Dowd et al. 2009; Sargsyan & Weedman 2009), (6) the warm molecular gas traced by H2 lines (e.g., Tommasin et al. 2008; Baum et al. 2010), and (7) the extreme ultraviolet (EUV) spectra of AGN revealed by high-ionization fine-structure IR lines (Meléndez et al. 2011).

In the far-IR, the Herschel Space Observatory (Pilbratt et al. 2010) provided a large gain in spectroscopic sensitivity when compared to ISO, allowing us to access this spectral range for a larger number of galaxies in the nearby universe. The Photoconductor Array Camera and Spectrometer (PACS; Poglitsch et al. 2010) and the Spectral and Photometric Imaging REceiver Fourier-transform spectrometer (SPIRE; Griffin et al. 2010; Naylor et al. 2010) sampled the 50–210 μm and 200–670 μm ranges, respectively. This spectral region covers the main cooling lines of photo-dissociation regions (PDR), [O i]${}_{\mathrm{63,145}\mu {\rm{m}}}$ and [C ii]${}_{158\mu {\rm{m}}}$, which provide information on the cold neutral gas (Tielens & Hollenbach 1985), as well as the high-J CO transitions (Kamenetzky et al. 2014, 2015), allowing the investigation of the excitation of molecular gas, the fine-structure lines of [O i]${}_{63\mu {\rm{m}}}$ and [N ii]${}_{122\mu {\rm{m}}}$ as tracers of the IR luminosity and star formation rate in ULIRGs (Farrah et al. 2013), and the origin of [C ii]${}_{158\mu {\rm{m}}}$ emission, mainly associated with the PDR (Spinoglio et al. 2015 hereafter S15); X-ray dissociation regions (XDR; Spinoglio et al. 2012b).

The aim of the present study is to extend for the first time, with a statistical approach, the spectroscopic work done by Spitzer in the mid-IR to the longer wavelengths using Herschel. Taking advantage of the Herschel scientific archive, we extended the previous study presented in S15, limited to a sample of 26 Seyfert galaxies, to a total sample of 170 active galaxies, and a comparison sample of 20 starburst galaxies and 43 dwarf galaxies—the latter taken from the Dwarf Galaxy Survey (DGS; Madden et al. 2013). Of particular interest is the development of the diagnostics that will be exploited by future IR observatories, e.g., the James Webb Space Telescope (JWST; Gardner et al. 2006) and the SPace Infrared telescope for Cosmology and Astrophysics (SPICA; Swinyard et al. 2009) for galaxies up to the peak of the star formation rate density (SFRD) at $1\lt z\lt 4$, but also the Atacama Large Millimeter/submillimeter Array (ALMA) for far-IR rest-frame observations of galaxies at high redshift ($z\gtrsim 3$). The combination of mid- and far-IR lines expands the possible diagnostics to, e.g., the [O iv]${}_{25.9\mu {\rm{m}}}/$[O iii]${}_{88\mu {\rm{m}}}$ ratio (hereafter [O iv]${}_{25.9}/$[O iii]88) that has been proposed as a powerful diagnostic to discriminate AGNs from star formation activity (S15). The larger statistics in the present work allow us to perform a more robust analysis of the diagnostics already tested in S15, including new line ratios sensitive to metallicity and ionization (Nagao et al. 2011, 2012). We include in this work the results of the "compact sample" of 43 dwarf galaxies presented by Cormier et al. (2015) to investigate how strong star formation activity in low-metallicity environments can be well separated from either AGN and "normal" starburst galaxies through mid- and far-IR line ratios. A set of Cloudy photoionization models were developed using AGN and starburst galaxies as ionization sources in order to interpret the behavior of the observed line ratios and their dependence on density, ionization parameter, and metallicity.

The text is organized as follows. The selection of the sample of galaxies is explained in Section 2, Section 3 presents the data reduction and describes the catalogs taken from the literature and included in our study, the Cloudy photoionization models are described in Section 4, the results of this work are detailed in Section 5, and the main conclusions are summarized in Section 6. Due to the large data set and associated tables needed to show the results of this study, the majority of the latter appear in their complete form only in the online version of this journal.

2. SAMPLE OF GALAXIES

2.1. AGN Sample

We have assembled the largest sample of active galaxies with far-IR spectroscopy by Herschel/PACS that was possible from the Herschel scientific archive (Table 3), in spite of the fact that the Herschel mission did not observe, during its 3.5 years of operational life, any systematic and statistically complete sample of active galaxies.

Table 3.  Samples of AGN, Starburst, and Dwarf Galaxies Observed with the PACS Spectrometer

# Name R.A. (2000) decl. (2000) z Dist. RefDa Type ${L}_{2\mbox{--}10\mathrm{keV}}$ CT RefXb ${L}_{[\mathrm{Ne}{\rm{V}}]14.3}$ ${L}_{[\mathrm{Ne}{\rm{V}}]24.3}$ ${L}_{[{\rm{O}}{\rm{IV}}]25.9}$ Z RefZc Alt.
    (h:m:s) (d:m:s)   (Mpc)     ($\mathrm{erg}\,{{\rm{s}}}^{-1})$     ($\mathrm{erg}\,{{\rm{s}}}^{-1}$) ($\mathrm{erg}\,{{\rm{s}}}^{-1}$) ($\mathrm{erg}\,{{\rm{s}}}^{-1}$) (${Z}_{\odot }$)    
AGN Sample
1 Mrk334 00:03:09.6038 +21:57:36.8064 0.021945 96.6 29 S1.8 ...     41.16 41.09 41.22     UGC00006
2 Mrk938 00:11:06.5412 −12:06:27.6624 0.019617 81.8   S2 42.33 N 25 <40.24 <39.47 <39.72     NGC17, NGC34
3 IRAS00182-7112 00:20:34.7210 −70:55:26.2488 0.326999 1656.9   S2 43.90 Y 44 <41.63 <41.82 <42.06      
Starburst Sample
1 NGC253 00:47:33.0727 −25:17:18.9960 0.000811 3.66 11 H ii 39.96 N 43 <38.52 <39.07 39.39 0.72 ± 0.30 13 Sculptor
2 M74 01:36:41.7384 +15:47:00.8592 0.002192 9.46 23 H ii 38.12   26 <37.21 <37.69 <37.14 0.46 8 NGC628
3 NGC891 02:22:32.8464 +42:20:52.6740 0.001761 10.8 31 H ii 40.46   67 <36.75 <37.94 <38.16 1.6 6 UGC01831
Dwarf Sample
1 HS0017+1055 00:20:21.4000 +11:12:21.0000 0.018846 78.6   Dwarf ...     ... ... ... 0.09 ± 0.02 10  
2 Haro11 00:36:52.4544 −33:33:16.7652 0.020598 86.0   Dwarf 40.98 N 23 <39.67 ... 40.59 0.47 ± 0.01 10  
3 HS0052+2536 00:54:56.3647 +25:53:08.0052 0.045385 193.0   Dwarf ...     <40.52 ... <40.14 0.24 ± 0.14 10  
No Lines Detected by PACS
1 3C33 01:08:52.8818 +13:20:14.1684 0.0597 256.6   S1h 43.86 N 64 41.20 41.10 41.80 0.89 3 4C+13.07
2 PG1114+445 11:17:06.3950 +44:13:33.2076 0.143862 655.5   S1.0 43.95 N 51 ... <41.94 42.04      
3 ESO506-G27 12:38:54.5820 −27:18:28.1376 0.025024 104.8   S2 43.06 N 70 ... ... 40.70     AM1236-270

Notes. The columns correspond to the galaxy name, coordinates taken from 2MASS Point Source Catalog, redshift, redshift-independent distance, reference for the distance, spectral type, absorption-corrected $2\mbox{--}10\,\mathrm{keV}$ luminosity, Compton thick (Y: yes; N: no), reference for the X-ray luminosity, [Ne v]${}_{14.3\mu {\rm{m}}}$, [Ne v]${}_{24.3\mu {\rm{m}}}$, and [O iv]${}_{25.9\mu {\rm{m}}}$ luminosities, metallicity based on optical lines, reference for the metallicity, and alternative name(s).

aReferences for redshift-independent distances: (1) Adamo et al. (2012), (2) Cortés et al. (2008), (3) Dolphin & Kennicutt (2002). bReferences for X-ray absorption-corrected luminosities: (1) Awaki et al. (2000), (2) Ballo et al. (2011), (3) Bassani et al. (1999). cReferences for optical metallicities: (1) Contini (2012), (2) Davis et al. (2013), (3) Dors et al. (2015).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

To do so, we have selected all objects in the Véron-Cetty & Véron (2010) active galaxies catalog that are classified either as QSO or Seyfert galaxies that have been detected with a signal-to-noise ratio $({\rm{S}}/{\rm{N}})\gt 3$ in at least one of the Herschel/PACS spectral lines in Table 1. The far-IR spectra of these objects were completed with published fine-structure mid-IR emission lines from Spitzer/IRS detected with ${\rm{S}}/{\rm{N}}\gt 3$ (see Section 3.2), mostly in the high-resolution mode (hereafter HR) and in a few cases in the low-resolution mode (hereafter LR). These galaxies belong to the following original catalogs: the third and fourth Cambridge catalogs of radio sources (3C and 4C, respectively; Edge et al. 1959; Pilkington & Scott 1965), the European Southern Observatory ESO/Uppsala survey of the Southern Hemisphere (Holmberg et al. 1974), the Fairall catalog of galaxies (Fairall 1977), the New General Catalog and the Index Catalogs (NGC and IC, respectively; Dreyer 1888, 1895), the Infrared Astronomical Satellite (IRAS) point source and faint source catalogs (Helou & Walker 1995; Moshir et al. 1993), the Markarian catalog (Markarian et al. 1989), the Morphological Catalog of Galaxies (MCG; Vorontsov-Velyaminov & Krasnogorskaya 1994), the Parkes Radio Sources Catalog (PKS; Wright & Otrupcek 1990), the Palomar-Green Catalog of Ultraviolet-Excess Stellar Objects (PG; Green et al. 2009), the Uppsala General Catalog of Galaxies (UGC; Nilson 1995), the catalog of southern peculiar galaxies and associations (AM; Arp & Madore 1996), the atlas of peculiar galaxies (Arp; Arp 1966), and the Zwicky catalog of galaxies (Zwicky et al. 1996).

This AGN Sample includes a total of 170 galaxies: 54 Seyfert 1 galaxies (S1), 26 Seyfert 1 galaxies with hidden broad-line regions (S1h; broad lines detected in their IR and/or polarized light spectra), 57 Seyfert 2 galaxies (S2), and 33 LINERs. The redshift distribution of the AGN sample, shown in Figure 1(a), covers the range of $-0.0008\lt z\lt 1.8264$, though most of the galaxies are located in the Local Universe with a median redshift of 0.0243 and a median absolute deviation of 0.027. The far-IR luminosity distribution, based on the flux of the continuum emission adjacent to the [C ii]${}_{158\mu {\rm{m}}}$ line, is also shown in Figure 1(b), and has a median value and a median absolute deviation of $\mathrm{log}({L}_{\mathrm{FIR}}/\mathrm{erg}\,{{\rm{s}}}^{-1})=40.49\pm 0.61$, covering the $35.54\lt \mathrm{log}({L}_{\mathrm{FIR}}/\mathrm{erg}\,{{\rm{s}}}^{-1})\lt 42.52$ range. The metallicities, compiled from the literature (see Table 3), span the $0.24\lt Z\lt 2.95\,{Z}_{\odot }$ range, with a median value and a median absolute deviation of $0.91\pm 0.21\,{Z}_{\odot }$.

Apart from the sample of 170 AGNs, there are 8 AGNs from the Véron-Cetty & Véron (2010) catalog that were observed by PACS for which no fine-structure emission lines were detected at ${\rm{S}}/{\rm{N}}\gt 3$. Consequently, these eight targets are not included in our main AGN sample, still we include for completeness upper limits for the PACS line fluxes and equivalent widths, PACS continuum measurements, and Spitzer line fluxes in Tables 46, and 8.

Table 4.  Far-IR Fine Structure Line Fluxes from Herschel/PACS

    Line fluxes (${10}^{-17}\,{\rm{W}}\,{{\rm{m}}}^{-2}$)  
# Name [O iii] [N iii] [O i] [O iii] [N ii] [O i] [C ii] Notes
    $51.81\,\,\mu {\rm{m}}$ $57.32\,\,\mu {\rm{m}}$ $63.18\,\,\mu {\rm{m}}$ $88.36\,\,\mu {\rm{m}}$ $121.90\,\,\mu {\rm{m}}$ $145.52\,\,\mu {\rm{m}}$ $157.74\,\,\mu {\rm{m}}$  
AGN
1 Mrk334 ... ... ... ... ... ... 76.66 ± 1.77 × 3
2 Mrk938 ... ... 78.12 ± 6.36 19.24 ± 3.58 ... 5.26 ± 0.92 82.15 ± 1.51 × 3
3 IRAS00182-7112 ... ... 7.10 ± 1.67 <2.04 ... ... ... c
Starburst Galaxies
1 NGC253 389.21 ± 183.61 599.54 ± 178.33 2947.6 ± 85.7 625.21 ± 53.54 802.72 ± 28.14 389.61 ± 30.99 3392.7 ± 22.9 × 3
2 M74 ... ... 20.58 ± 10.84 <14.16 5.85 ± 2.70 ... 43.12 ± 1.82 × 3
3 NGC891 ... ... 85.54 ± 11.08 ... 73.28 ± 2.19 ... ... × 3
No Lines Detected by PACS
1 3C33 ... ... ... ... ... <1.54 ... × 3
2 PG1114+445 ... ... <3.06 ... ... ... <1.3 × 3
3 ESO506-G27 ... ... ... ... ... <0.96 ... × 3

Note. Far-IR fine structure line fluxes for the samples of AGN and starburst galaxies from Herschel PACS observations.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

2.2. Starburst Sample

As a comparison sample, we compiled a sample of 20 starburst galaxies and reduced the PACS spectroscopic observations from the Herschel archive. From the starburst galaxy sample of Bernard-Salas et al. (2009), we have selected the objects classified as "pure starburst galaxies," discarding those objects with an AGN contribution. From the sample of Goulding & Alexander (2009), we have explicitly included only those objects optically classified as "H ii" region galaxies—by their position in the BPT diagram (Baldwin et al. 1981)—and without an AGN contribution in the mid-IR range, as provided by these authors. For the mid-IR spectroscopy, we used the results from Spitzer/IRS high-resolution spectra published in Bernard-Salas et al. (2009) and Goulding & Alexander (2009). The final sample of 20 starburst galaxies covers a redshift range of $0.0001\lt z\lt 0.0159$ with a median value of 0.0022 and a median absolute deviation of 0.0006 (Figure 1(a)). The far-IR luminosity distribution is based on the flux of the continuum emission adjacent to the [C ii]${}_{158\mu {\rm{m}}}$ line and covers the $38.90\lt \mathrm{log}({L}_{\mathrm{FIR}}/\mathrm{erg}\,{{\rm{s}}}^{-1})\lt 41.15$ range, with a median value and a median absolute deviation of $\mathrm{log}({L}_{\mathrm{FIR}}/\mathrm{erg}\,{{\rm{s}}}^{-1})=39.90\pm 0.40$ (Figure 1(b)). The starburst sample spans a metallicity range of $0.46\lt Z\lt 2.75\,{Z}_{\odot }$, with a median value and a median absolute deviation of $1.10\pm 0.29\,{Z}_{\odot }$(see Table 3).

2.3. Dwarf Galaxy Sample

We also compared our samples of AGN and starburst galaxies to the mid- and far-IR fine-structure line data set for the DGS, presented in Cormier et al. (2015). The DGS includes blue compact dwarf galaxies with metallicities ranging from $1/50\,{Z}_{\odot }$ to $1/3\,{Z}_{\odot }$, selected from the Hamburg/SAO Survey and the First and Second Byurakan Surveys (Izotov et al. 1991; Ugryumov et al. 2003; Madden et al. 2013). Here we include the 43 dwarf galaxies from the DGS classified as compact (Cormier et al. 2015), i.e., excluding nearby galaxies (NGC 4214 is the closest one at $D=2.93\,\mathrm{Mpc}$). Dwarf galaxies probe a more extreme star formation environment at low metallicities when compared to the starburst galaxies. Their low metallicity might mimic the conditions of chemically unevolved galaxies at high redshift, and thus it is important to include these objects in our diagnostics in order to have a wider view of the star formation process. Far-IR continuum fluxes for the dwarf galaxies are taken from Rémy-Ruyer et al. (2015). The sample of dwarf galaxies covers a redshift range of $-0.0003\lt z\lt 0.0454$, with a median value of 0.0048 and a median absolute deviation of 0.0038 (Figure 1(a)). The far-IR luminosity distribution is based on the Herschel/PACS photometry at $160\,\,\mu {\rm{m}}$, published in Rémy-Ruyer et al. (2015), and covers the $36.13\lt \mathrm{log}({L}_{\mathrm{FIR}}/\mathrm{erg}\,{{\rm{s}}}^{-1})\lt 40.38$ range, with a median value and a median absolute deviation of $\mathrm{log}({L}_{\mathrm{FIR}}/\mathrm{erg}\,{{\rm{s}}}^{-1})=38.49\pm 0.91$ (Figure 1(b)). The dwarf galaxies included in this study have stellar masses in the 3 × 106–3.4 × 1010 M range, thus overlapping with the masses of nearby starburst galaxies (Madden et al. 2013; Rémy-Ruyer et al. 2015).

2.4. ISO/LWS Data

For comparison, 81 galaxies with far-IR line fluxes observed with ISO/LWS, taken from Brauher et al. (2008), were considered only for those active galaxies present in the original selection (Véron-Cetty & Véron 2010) and for local starburst galaxies, but were not observed by Herschel/PACS. The mid-IR line fluxes were compiled by us from published Spitzer/IRS spectra. These objects are shown only for comparison and are not included in our main sample, due to the larger aperture of ISO compared with that of PACS, therefore the galaxies observed by ISO/LWS are represented by open symbols in the figures shown in this work. In the Appendix, we investigate the possible effect of the large ISO apertures on the fluxes of the far-IR fine-structure lines. Figure 13 shows the comparison of ISO/LWS versus Herschel/PACS [C ii]${}_{158\mu {\rm{m}}}$ line fluxes for the subsample 45 AGN and starburst galaxies that were observed by both facilities. PACS fluxes are slightly higher with a median value and a median absolute deviation of ${F}_{[{\rm{C}}\,{\rm{II}}]158}^{\mathrm{LWS}}/{F}_{[{\rm{C}}\,{\rm{II}}]158}^{\mathrm{PACS}}=0.95\pm 0.37$, thus the inclusion of ISO data in our diagrams is reasonable since significant contamination from extended emission is not expected for our sample of galaxies.

3. OBSERVATIONS

The combination of Herschel/PACS and Spitzer/IRS allows us to cover the fine-structure emission lines from the mid- to the far-IR (10–200 μm in the rest-frame) for all the galaxies in the sample. This database was completed with the Herschel/SPIRE published values of the [N ii]${}_{205\mu {\rm{m}}}$, and [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$ line fluxes (mainly from Kamenetzky et al. 2015). In this section, we describe the reduction of the Herschel/PACS observations and the characteristics of the different data sets collected from the literature.

3.1. Herschel/PACS Spectroscopy

The far-IR spectra of the galaxies in the AGN and starburst samples were acquired with Herschel/PACS, which includes an integral field unit spectrograph observing in the ∼50–200 μm range, with 5 × 5 squared spaxel elements covering a field of view (FOV) of about 47'' × 47''  and a spectral resolving power in the range of R = 1000–4000 (Δv = 75–300 km s−1), depending on wavelength. The observations were retrieved from the Herschel Science Archive. The data were reduced using our own pipeline based on the standard recipes included in the hipe 6 v13.0.0 environment. The data reduction procedure includes outlier flagging, spectral flat-fielding, re-grid of the wavelength sampling, and flux calibration. Most of the targets were acquired using the chop-node mode and were reduced with the background normalization method: the off-source spectra are used to correct the spectral response of the on-source data and to perform the background subtraction. This is the standard procedure in the hipe pipeline for PACS since v13.0.0. Observations acquired with the unchopped mode include a dedicated off-source pointing, which was used to perform the background subtraction.

For each target, a final standard rebinned data cube was produced for the available fine-structure lines listed in Table 1. The [C ii]${}_{158\mu {\rm{m}}}$ transition shows the best coverage with 159 AGN and 17 starburst galaxies observed and detected, while the [O iii]${}_{52\mu {\rm{m}}}$ transition was detected only for 19 AGNs and 1 of the starburst galaxies. The final spectra are extracted from the innermost 3 × 3 spaxels ($28\buildrel{\prime\prime}\over{.} 2\times 28\buildrel{\prime\prime}\over{.} 2$), to avoid possible flux losses due to the known pointing jitter and offsets. In all cases, a point-source loss correction, implemented within hipe, is applied to the extracted spectra. For a total of 26 sources (all in the AGN Sample), one or more lines were not detected in the 3 × 3 spectrum, but only in the central spaxel. This is typically the case of faint sources in which the spaxels surrounding the central one are dominated by noise, and thus only the central spaxel has a substantial contribution to the line emission. For these targets, the final spectra, and derived fluxes in Table 4 are extracted only from the central spaxel (and indicated by a "c" table in the column "Notes"). A complete flux extraction has been performed for all galaxies using three different methods: (1) the central spaxel only, (2) the central 3 × 3 array, and (3) the whole detector area of 5 × 5 spaxels.The complete database is reported in Table 11. We note that the 5 × 5 fluxes are not corrected for point-source losses (this is not provided by hipe v13.0.0), which are expected to be of the order of $\sim 10 \% $ for a point-like source according to the PACS manual.

The analysis was performed using our own routines developed in Python7 , and routines from the Astropy8 package (Astropy Collaboration et al. 2013). In order to increase the S/N of the extracted spectra, we used a Wiener filter to remove those spurious spectral features characterized by a width smaller than the instrumental spectral resolution and a peak below 3 × rms. The latter are produced during the acquisition and/or reduction process and must be removed for a realistic estimate of the flux errors. Line fluxes measured before and after the filtering are in agreement within the estimated uncertainty, while the S/Ns improve by a factor of approximately three on average.

Following the approach in S15, the line fluxes are obtained by numerical integration across the line profiles (the blue-shaded area in Figures 13). The continuum level is estimated from a 1D polynomial fit performed for the spectral channels located in the blue and the red wings adjacent to the line (black-solid line). A single Gaussian was fitted to the profile of each line in order to obtain the central wavelength, standard deviation, and Gaussian flux (red-solid line). The original unfiltered spectrum and its associated uncertainty are shown in gray in Figures 13. Additionally, five spectral maps for each of the lines covered are also provided in the style of Figure 4, including the Wiener-filtered spectrum for the line (in blue) and the adjacent continuum (in black), and the unfiltered spectrum with its associated uncertainty (in gray). In order to facilitate the visualization and comparison of different line profiles and intensities in these figures, the spectra are shown after continuum subtraction (continuum flux density is quoted in the lower part of the figure as $\sum {F}_{C}/{\rm{\Delta }}\lambda $) and normalized by a certain factor (quoted in the upper left part of the figure).

Figure 1.

Figure 1. Redshift and luminosity distributions for the AGN sample. The colors correspond to the different spectral types: Seyfert 1 (blue), Seyfert 1 with hidden broad-line regions (pink), Seyfert 2 (red), LINER (green), starburst galaxies (yellow), and dwarf galaxies (purple). Left (a): redshift distribution. Five objects with $z\gt 0.5$ are out of the redshift range shown in the plot. Right (b): far-IR luminosity (${L}_{\mathrm{FIR}}$) distribution, based on the continuum flux adjacent to the [C ii]${}_{158\mu {\rm{m}}}$ line, observed for most of the galaxies in the AGN and starburst samples, and Herschel/PACS photometry at $160\,\,\mu {\rm{m}}$ for the dwarf galaxy sample from Rémy-Ruyer et al. (2015).

Standard image High-resolution image
Figure 2.

Figure 2. Ionization and density vs. temperature. The symbols correspond to the different spectral types: Seyfert 1 (blue triangles), Seyfert 1 with hidden broad-line regions (pink triangles), Seyfert 2 (red squares), LINER (green circles), starburst galaxies (yellow stars), and dwarf galaxies (purple stars). Open symbols are used for upper and lower limits derived from PACS data, and for ISO/LWS data. Left (a): the [O i]${}_{145/63}$ ratio vs. the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ ratio. The [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ is sensitive to the ionization, while the [O i]${}_{145/63}$ line ratio is sensitive to the temperature in the 100–400 K range. The line ratios corresponding to these temperature values, assuming a density of ${n}_{{{\rm{H}}}_{2}}\approx {10}^{4}\,{\mathrm{cm}}^{-3}$ in Liseau et al. (2006), are marked in the bottom part of the diagram. Right (b): the [O i]${}_{145/63}$ ratio vs. the [S iii]${}_{33.5/18.7}$ line ratio. The latter measures the electron density, whose values—assuming a gas in the pure collisional regime at $T={10}^{4}\,{\rm{K}}$ as in S15—are marked in the right part of the diagram. Arp 220 shows a strong absorption profile in [O i]${}_{63\mu {\rm{m}}}$ and thus has been excluded from the correlation analysis.

Standard image High-resolution image
Figure 3.

Figure 3. Ionization, density, and PDR temperature-sensitive line ratios (same notations as in Figure 2). Photoionization models of AGNs, starburst galaxies, and dwarf galaxies are shown as blue, yellow, and purple grids, respectively. The logarithmic values of the density (${n}_{{\rm{H}}}$) and ionization potential (U) of the photoionization models are indicated in the figures. Left (a): the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ is sensitive to the ionization, while the [S iii]${}_{33.5/18.7}$ line ratio is sensitive to the electron density. Right (b): the [C i]${}_{609/371}$ line ratio vs. the [O i]${}_{145/63}$ ratio, which is sensitive to the temperature in the 100–400 K range. Higher [C i]${}_{609/371}$ ratios correspond to lower temperatures within the 20–$100\,{\rm{K}}$ range, according to the models in Meijerink et al. (2007). Black diamonds indicate the predicted line ratios for an XDR, extracted from Table 2 in Ferland et al. (2013).

Standard image High-resolution image
Figure 4.

Figure 4. Density stratification. Derived values for the electron density vs. the ionization potential for the galaxies in the sample with two detected lines of the same species. A weighted least-square fit (black-solid line) shows a correlation between density and ionization. Blue- and green-dashed lines show the uncertainty associated with the slope and the intercept of the fit, respectively. The red dotted–dashed line is the result of the linear regression using the Kaplan–Meier residuals, and the red dotted lines correspond to its associated uncertainty.

Standard image High-resolution image

In the electronic version of this journal, we provide, for each of the far-IR fine-structure lines observed by PACS for the 170 AGN and 20 starburst galaxies in the sample the line fluxes (Table 4), the equivalent width values (Table 5), the continuum level (Table 6), the standard deviation of the best-fit Gaussian profile (Table 7), the corresponding figures for the central spaxel (Figure Set 1), 3 × 3 array (Figure Set 2), 5 × 5 array extracted spectra (Figure Set 3) and the 5 × 5 array line maps (Figure Set 4). The Herschel/PACS and Spitzer/IRS line fluxes for the sample of dwarf galaxies are taken from Cormier et al. (2015) and are not reported in this publication. Upper limits are also provided for eight AGN that were not detected by PACS.

Table 5.  Equivalent width for Far-IR Lines from Herschel/PACS

    Equivalent Width ($\,\mu {\rm{m}}$)  
# Name [O iii] [N iii] [O i] [O iii] [N ii] [O i] [C ii] Notes
    $51.81\,\,\mu {\rm{m}}$ $57.32\,\,\mu {\rm{m}}$ $63.18\,\,\mu {\rm{m}}$ $88.36\,\,\mu {\rm{m}}$ $121.90\,\,\mu {\rm{m}}$ $145.52\,\,\mu {\rm{m}}$ $157.74\,\,\mu {\rm{m}}$  
AGN
1 Mrk334 ... ... ... ... ... ... 2.252 ± 0.193 × 3
2 Mrk938 ... ... 0.057 ± 0.005 0.026 ± 0.005 ... 0.032 ± 0.006 0.656 ± 0.016 × 3
3 IRAS00182-7112 ... ... 0.064 ± 0.015 <0.049 ... ... ... c
Starburst Galaxies
1 NGC253 0.008 ± 0.004 0.007 ± 0.002 0.039 ± 0.001 0.013 ± 0.001 0.037 ± 0.001 0.031 ± 0.003 0.374 ± 0.003 × 3
1 M74 ... ... 0.017 ± 0.009 ... 0.014 ± 0.006 ... 0.430 ± 0.026 × 3
2 NGC891 ... ... 0.035 ± 0.005 ... 0.080 ± 0.002 ... ... × 3
No Lines Detected by PACS
1 3C33 ... ... ... ... ... <0.352 ... × 3
2 PG1114+445 ... ... ... ... ... ... ... × 3
3 ESO506-G27 ... ... ... ... ... <0.051 ... × 3

Note. Equivalent width values derived for the far-IR fine structure lines observed by Herschel/PACS.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 6.  Far-IR Continuum Fluxes from Herschel/PACS Spectra

    Continuum Flux ($\mathrm{Jy}$)  
# Name [O iii] [N iii] [O i] [O iii] [N ii] [O i] [C ii] Notes
    $51.81\,\,\mu {\rm{m}}$ $57.32\,\,\mu {\rm{m}}$ $63.18\,\,\mu {\rm{m}}$ $88.36\,\,\mu {\rm{m}}$ $121.90\,\,\mu {\rm{m}}$ $145.52\,\,\mu {\rm{m}}$ $157.74\,\,\mu {\rm{m}}$  
AGN
1 Mrk334 ... ... ... ... ... ... 2.83 ± 0.23 × 3
2 Mrk938 ... ... 18.11 ± 0.51 19.23 ± 0.31 ... 11.54 ± 0.16 10.39 ± 0.17 × 3
3 IRAS00182-7112 ... ... 1.49 ± 0.04 1.09 ± 0.03 ... ... ... c
Starburst Galaxies
1 NGC253 450.23 ± 14.16 936.35 ± 8.24 1013.0 ± 5.8 1206.1 ± 5.4 1071.4 ± 3.2 874.91 ± 4.07 753.27 ± 3.54 × 3
2 M74 ... ... 16.19 ± 0.91 ... 20.79 ± 0.25 ... 8.32 ± 0.36 × 3
3 NGC891 ... ... 32.97 ± 1.16 ... 45.68 ± 0.27 ... 351 ± 28a × 3
No Lines Detected by PACS
1 3C33 ... ... ... ... ... 0.31 ± 0.08 ... × 3
2 PG1114+445 ... ... <0.34 ... ... ... <0.25 × 3
3 ESO506-G27 ... ... ... ... ... 1.34 ± 0.05 ... × 3

Note. Continuum flux values derived for the far-IR fine structure lines observed by Herschel/PACS.

aHerschel/PACS continuum at $160\,\,\mu {\rm{m}}$ taken from Hughes et al. (2014).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 7.  Gaussian Sigma for far-IR Fine Structure Lines from Herschel/PACS

    Gaussian Sigma ($\mathrm{km}\,{{\rm{s}}}^{-1}$)  
# Name [O iii] [N iii] [O i] [O iii] [N ii] [O i] [C ii] Notes
    $51.81\,\,\mu {\rm{m}}$ $57.32\,\,\mu {\rm{m}}$ $63.18\,\,\mu {\rm{m}}$ $88.36\,\,\mu {\rm{m}}$ $121.90\,\,\mu {\rm{m}}$ $145.52\,\,\mu {\rm{m}}$ $157.74\,\,\mu {\rm{m}}$  
AGN
1 Mrk334 ... ... ... ... ... ... 151.8 ± 0.8 × 3
2 Mrk938 ... ... 165.9 ± 2.3 195.1 ± 8.0 ... 158.4 ± 6.3 185.8 ± 0.7 × 3
3 IRAS00182-7112 ... ... 632.6 ± 42.7 ... ... ... ... c
Starburst Galaxies
1 NGC253 142.0 ± 13.0 202.3 ± 10.8 145.4 ± 0.8 125.9 ± 1.5 169.0 ± 1.2 148.9 ± 2.2 136.8 ± 0.2 × 3
2 M74 ... ... 114.0 ± 14.1 ... 260.4 ± 134.8 ... 97.9 ± 1.4 × 3
3 NGC891 ... ... 101.8 ± 3.8 ... 159.4 ± 1.8 ... ... × 3

Note. Sigma values derived from the Gaussian fit to the profile of the far-IR fine structure lines observed by Herschel/PACS.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

3.2. Spitzer/IRS Spectroscopy

Table 8, in the electronic version of this journal, collects published mid-IR (10–35 μm) fine-structure line fluxes measured with Spitzer/IRS for our samples of AGN and starburst galaxies. HR observations are favored, completed with LR spectra in those cases without HR observations (14 galaxies have both IRS-LR spectra and Herschel/PACS). These values were complemented with unpublished IRS observations from the Spitzer archive and reduced by us for 20 galaxies, following the same procedure as in Pereira-Santaella et al. (2010a, 2010b). Spitzer/IRS HR line fluxes for the samples of starburst and dwarf galaxies were taken from Bernard-Salas et al. (2009) and Goulding & Alexander (2009), and Cormier et al. (2015), respectively.

Table 8.  Mid-IR Fine Structure Line Fluxes from Spitzer/IRS

    Line fluxes (10 ${}^{-17}\,{\rm{W}}\,{{\rm{m}}}^{-2}$) in SH ($4\buildrel{\prime\prime}\over{.} 7\times 11\buildrel{\prime\prime}\over{.} 3$) Line fluxes (10 ${}^{-17}\,{\rm{W}}\,{{\rm{m}}}^{-2}$) in LH ($11\buildrel{\prime\prime}\over{.} 1\times 22\buildrel{\prime\prime}\over{.} 3$) Mode Ref.a
# Name [S iv] [Ne ii] [Ne v] [Ne iii] [S iii] [Ne v] [O iv] [S iii] [Si ii]  
    $10.51\,\,\mu {\rm{m}}$ $12.81\,\,\mu {\rm{m}}$ $14.32\,\,\mu {\rm{m}}$ $15.56\,\,\mu {\rm{m}}$ $18.71\,\,\mu {\rm{m}}$ $24.32\,\,\mu {\rm{m}}$ $25.89\,\,\mu {\rm{m}}$ $33.48\,\,\mu {\rm{m}}$ $34.82\,\,\mu {\rm{m}}$  
AGN
1 Mrk334 11.0 ± 2.0 30.0 ± 8.0 13.0 ± 2.0 26.0 ± 3.0 23.0 ± 5.0 11.0 ± 2.0 15.0 ± 3.0 90.0 ± 8.0 ... LR 11
2 Mrk938 <1.46 52.10 ± 1.45 <2.19 6.37 ± 0.55 7.56 ± 0.64 <0.37 <0.66 <10.7 40.50 ± 4.21 HR+LR 41, 11
3 IRAS00182-7112 <0.08 6.3 ± 0.3 <0.13 2.1 ± 0.2 0.23 ± 0.03 <0.20 <0.35 ... ... HR 39
Starburst Galaxies
1 NGC253 <10.5 2832.3 ± 64.2 <20.5 204.6 ± 9.6 666.4 ± 14.9 <73.36 154.7 ± 26.9 1538.0 ± 30.1 2412.0 ± 48.0 HR 5
2 M74 0.22 2.31 ± 0.36 <0.15 <0.08 0.23 <0.46 <0.13 12.35 ± 0.94 4.78 ± 0.40 HR 20, 47
3 NGC891 0.88 8.57 ± 0.78 <0.04 0.84 ± 0.07 1.99 ± 0.17 <0.62 <1.03 10.74 ± 2.02 28.11 ± 0.92 HR 20, 47
No Lines Detected by PACS
1 3C33 1.2 ± 0.1 3.9 ± 0.2 2.0 ± 0.3 5.3 ± 0.2 2.5 ± 0.4 1.6 ± 0.2 8.1 ± 0.2 ... ... LR 24
2 PG1114+445 0.90 ± 0.18 ... ... <1.53 ... <1.70 2.15 ± 0.49 ... ... LR 47
3 ESO506-G27 <1.02 5.18 ± 1.25 ... 6.02 ± 0.89 1.99 ± 1.12 ... 3.80 ± 0.71 ... ... LR 47, 35

Note. Mid-IR fine structure line fluxes for the galaxies in our sample from Spitzer/IRS observations.

aReferences are coded as follows: (1) Alonso-Herrero et al. (2012), (2) Armus et al. (2004), (3) Armus et al. (2006), (4) Armus et al. (2007), (5) Bernard-Salas et al. (2009), (6) Bressan et al. (2006), (7) Dale et al. (2009), (8) Dasyra et al. (2008), (9) Dasyra et al. (2011), (10) Deo et al. (2006), (11) Deo et al. (2007), (12) Diamond-Stanic et al. (2009), (13) Dicken et al. (2012), (14) Dixon & Joseph (2011), (15) Donahue et al. (2011), (16) Dudik et al. (2007), (17) Dudik et al. (2009), (18) Farrah et al. (2007), (19) Gorjian et al. (2007), (20) Goulding & Alexander (2009), (21) Guillard et al. (2012), (22) Inami et al. (2013), (23) Keremedjiev et al. (2009), (24) Ogle et al. (2006), (25) Ogle et al. (2010), (26) Pereira-Santaella et al. (2010b), (27) Pereira-Santaella et al. (2010a), (28) Pérez-Beaupuits et al. (2011), (29) Perlman et al. (2007), (30) Petric et al. (2011), (31) Privon et al. (2012), (32) Rampazzo et al. (2013), (33) Roussel et al. (2007), (34) Sales et al. (2010), (35) Sargsyan et al. (2011), (36) Satyapal et al. (2007), (37) Schweitzer et al. (2006), (38) Spinoglio et al. (2015), (39) Spoon et al. (2009), (40) Tommasin et al. (2008), (41) Tommasin et al. (2010), (42) Tommasin upublished, (43) Veilleux et al. (2009), (44) Weaver et al. (2010), (45) Willett et al. (2010), (46) Willett et al. (2011), (47) This work. Polarimetry: (48) Alexander et al. (2000), (49) Moran et al. (2001), (50) Pernechele et al. (2003).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

3.3. Herschel/SPIRE Spectroscopy

In Table 9, we compiled the published Herschel/SPIRE fluxes of the [N ii]${}_{205\mu {\rm{m}}}$, [C i]${}_{371\mu {\rm{m}}}$, and [C i]${}_{609\mu {\rm{m}}}$ lines, for the samples of AGN, starburst, and dwarf galaxies. These values are mainly provided by Kamenetzky et al. (2015) and complemented with the data published in Israel et al. (2015), Pereira-Santaella et al. (2013), Pereira-Santaella et al. (2014), and Kamenetzky et al. (2014).

Table 9.  [N ii]${}_{205\mu {\rm{m}}}$ and [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$ Line Fluxes

          Line fluxes (${10}^{-17}\,{\rm{W}}\,{{\rm{m}}}^{-2}$)  
Name R.A. decl. Redshift Class [N ii] [C i] [C i] RefNCa
          $205.2\,\,\mu {\rm{m}}$ $371\,\,\mu {\rm{m}}$ $609\,\,\mu {\rm{m}}$  
Mrk938 00:11:06.5412 −12:06:27.6624 0.019617 S2 3.47 ± 0.24 1.16 ± 0.13 0.81 ± 0.26 6
Haro11 00:36:52.4544 −33:33:16.7652 0.020598 Dwarf 2.31 ± 0.46 <0.81 ... 6
NGC253 00:47:33.0727 −25:17:18.9960 0.000811 H ii 175.93 ± 11.21 107.82 ± 2.70 42.69 ± 2.13 6
IZw1 00:53:34.9236 +12:41:35.9232 0.0612 S1n 1.11 ± 0.19 <0.55 <1.17 6
IRAS01003-2238 01:02:49.9894 −22:21:57.2616 0.117835 S2 ... <0.40 ... 6
IIIZw35 01:44:30.5400 +17:06:08.8056 0.027436 S2 <1.07 0.44 ± 0.10 <0.63 6
Mrk1014 01:59:50.2483 +00:23:40.7436 0.16311 S1.5 0.46 ± 0.15 <0.63 ... 6
NGC891 02:22:32.8464 +42:20:52.6740 0.001761 H ii 78.46 ± 1.95 4.93 ± 0.40 3.60 ± 0.55 6
NGC1068 02:42:40.7071 −00:00:48.0204 0.003793 S1h 186.65 ± 6.83 29.65 ± 0.81 14.04 ± 0.69 6
NGC1097 02:46:18.9775 −30:16:28.9344 0.00424 LINb 92.11 ± 2.44 6.68 ± 0.26 4.07 ± 0.47 6

Note. Line fluxes for nitrogen [N ii]${}_{205\mu {\rm{m}}}$ and neutral carbon [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$ lines in our sample, collected from Herschel/SPIRE and ground-based observations published in the literature.

aReferences are coded as follows: (1) Gerin & Phillips (2000), (2) Israel & Baas (2002), (3) Israel (2009), (4) Israel et al. (2015), (5) Kamenetzky et al. (2014), (6) Kamenetzky et al. (2015), (7) Rigopoulou et al. (2013).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

4. MODELS

The photoionization code Cloudy 9 has been used in this study to model the physical conditions of the gas exciting the IR fine-structure lines. Calculations have been performed with its version C13.03, last described by Ferland et al. (2013), using the pyCloudy library (Morisset 2013). To reproduce the emission of the NLR associated with the nuclei of active galaxies, a grid of constant density models with a plane-parallel geometry and sampling hydrogen density (${n}_{{\rm{H}}}$) in the $\mathrm{log}({n}_{{\rm{H}}}/{\mathrm{cm}}^{-3})=1$ to 6 range, has been built using AGN spectra as primary radiation sources. The AGN ionizing continuum corresponds to a power law extending from the optical to the X-rays with a slope of $\alpha =-1.4$ (${S}_{\nu }\propto {\nu }^{\alpha }$), and ionization parameters (U10 ) with values of $\mathrm{log}U=-2.0,-2.5,-3.0,-3.5$. For these models, we choose a maximum column density of ${N}_{{\rm{H}}}={10}^{23}\,{\mathrm{cm}}^{-2}$ as the stopping criterion of the spatial integration, representative of the column density found in NLR clouds (Moore & Cohen 1994). LINER galaxies typically show a steeper optical-to-UV continuum when compared with brighter AGNs (Halpern et al. 1996; Ho 1999; see also Fernández-Ontiveros et al. 2012; Mason et al. 2012). Following this, a similar grid of models covering the same density and ionization parameter values but with a steeper ionizing continuum ($\alpha =-3.5$) has been computed in order to reproduce the line ratios expected for a weaker ionizing spectra, i.e., for a vanishing accretion disk (Ho 2008). We note, however, that the excitation mechanism in LINERs is still a matter of debate (Ho 2008 and references therein) and different mechanisms might contribute (e.g., Dopita et al. 2015).

The star formation ionizing spectrum was simulated using the Starburst99 code (Leitherer et al. 1999) for two cases: (1) a young burst of star formation with an age of $1\,\mathrm{Myr}$ and a metallicity of $Z=0.004$ ($1/5\,{Z}_{\odot }$), in order to produce a hard UV ionizing spectrum in a low-metallicity environment, and (2) a continuous burst of star formation with an age of $20\,\mathrm{Myr}$ and solar metallicity (following our previous simulations in S15). In the first case, we aim to reproduce the conditions in violent and short-lived star formation events, like those in dwarf galaxies, thus we refer to these models as "dwarf galaxy models" hereafter. The second case aims to reproduce the more quiescent and continuous star formation in the disks of galaxies, characterized by a softer ionizing continuum; accordingly, these will be called "starburst models" hereafter. The metallicity of the stellar population determines the strength of the ionizing continuum, thus we used sub-solar-metallicity models for dwarf galaxies11 and solar metallicities for starburst models. The ionizing continuum is also dependent on the age of the stellar population for instant starburst models, but it is fairly independent of the age for continuum starburst models after the first few million years (Leitherer et al. 1999). Thus the main results of this work will not change if a $\gtrsim 5\,\mathrm{Myr}$ continuum starburst model is considered to be a reference for the starburst galaxies.

We use models with plane-parallel geometry, constant pressure, initial densities in the $\mathrm{log}({n}_{{\rm{H}}}/{\mathrm{cm}}^{-3})=1$ to 6 range, and ionization parameters in the $\mathrm{log}U=-2.0$ to −4.5 range. In both cases we assumed two intervals for the Kroupa initial mass function (IMF; with exponents 1.3, 2.3 and mass boundaries of 0.1, 0.5, and $100\,{M}_{\odot }$), the 1994 Geneva tracks with standard mass-loss rates, and the Pauldrach/Hillier atmospheres, which take into account the effects of non-LTE and radiation driven winds. For the study of the dependence of line ratios with metallicity (Section 5.5), the previous models were extended to the following metallicity values: $Z=0.001$, 0.004, 0.008, 0.02, and 0.04. These values are based on the metallicities available for the Geneva stellar tracks with standard mass-loss rates, used in the Starburst99 models to generate the starburst ionizing spectra.

In dwarf galaxy and starburst models, we chose a different stopping criteria based on the temperature, instead of the column density. The spatial integration was stopped at the radius where two different temperatures were reached, producing two kinds of models in each case: (1) models stopped at $T=1000\,{\rm{K}}$ that only include the contribution of photoionized gas to the emission lines and (2) models extended to $T=50\,{\rm{K}}$ that take into account the contribution from the PDR. In particular, the comparison of 1000 and 50 K models allows us to identify which line ratios are affected by the contribution of PDR emission and, e.g., to investigate the origin of [C ii]${}_{158\mu {\rm{m}}}$, which is produced by both neutral gas and low-density, low-excitation ionized gas (see also S15). In both cases, we used models with constant pressure, with the initial densities mentioned earlier, since they accommodate the changes in the structure of the cloud better when the calculation is extended deeper into the cold neutral gas, down to $50\,{\rm{K}}$, thus the predicted line ratios for the PDR are more reliable.

For AGN models, we did not choose a temperature-based stopping criterion to differentiate the XDR contribution because (1) the simulations do not converge to temperatures lower than $\gtrsim 100\,{\rm{K}}$, due to the harder spectra of AGNs and the X-ray contribution heating the neutral gas, (2) further discussion in Sections 5.1 and 5.2 show that the XDR contribution to neutral gas tracers, e.g., [C i]${}_{609/371}$ and [C ii]158/[O i]63, is negligible, while the values found for these line ratios are in agreement with PDR emission.

All the models presented here to reproduce AGNs (Seyfert and LINERs), dwarf galaxies (with and without PDR), and starburst galaxies (with and without PDR), aim to predict ideal cases for these ionizing sources. Most of the objects in our sample are expected to show mixed contributions since our large aperture spectra include emission from the host galaxy and therefore star formation processes contribute to the emission of most lines. Thus, the observed line ratios are expected to lie among these models for most of the diagnostic diagrams.

5. RESULTS

The final database includes a variety of lines sensitive to different sources of excitation and physical conditions, from the high-ionization [Ne v]${}_{14.3,24.3\mu {\rm{m}}}$ lines—indicative of AGN activity, to star formation/H ii region tracers (e.g., [Ne ii]${}_{12.8\mu {\rm{m}}}$, [Ne iii]${}_{15.6\mu {\rm{m}}}$, [O iii]${}_{88\mu {\rm{m}}}$), and PDR sensitive lines (e.g., [O i]${}_{\mathrm{63,145}\mu {\rm{m}}}$, [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$, [C ii]${}_{158\mu {\rm{m}}}$, the latter also produced in H ii regions). In the following sections we will explore the dependence of line ratios on various physical quantities.

5.1. Density, Ionization, and Temperature

As shown in our previous work (S15), emission-line ratios obtained from pairs of mid- and far-IR lines from the same ionic species, e.g., [N ii]205/[N ii]122 (hereafter [N ii]${}_{205/122}$), [S iii]${}_{33.5/18.7}$, [O iii]${}_{88/52}$, and [Ne v]${}_{24.3/14.3}$, have the same ionization potential but different critical densities (see Table 1), and thus can be used to trace the densities of the ionized gas in the ${n}_{{\rm{H}}}\approx 10\,{\mathrm{cm}}^{-3}$ to ${10}^{5}\,{\mathrm{cm}}^{-3}$ range (e.g., Rubin et al. 1994). On the other hand, the cooling lines of the neutral gas, [O i]${}_{\mathrm{63,145}\mu {\rm{m}}}$, [C ii]${}_{158\mu {\rm{m}}}$, and [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$, provide density and temperature diagnostics when they are compared to PDR and XDR models (e.g., Tielens & Hollenbach 1985; Liseau et al. 2006; Meijerink et al. 2007). In particular, theoretical estimates based on the statistical equilibrium equations predict that the [O i]${}_{145/63}$ line ratio is, in the optically thin limit, a good temperature tracer for the neutral gas in the T ∼ 100–400 K range at ${n}_{{\rm{H}}}\lesssim {10}^{4}\,{\mathrm{cm}}^{-3}$ (Liseau et al. 2006). The [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$ lines, observed by Herschel/SPIRE, are formed in the transition layer between ionized carbon and molecular CO gas (C+/C/CO), typically above their critical densities (${n}_{{\rm{H}}}\gtrsim {10}^{4}\,{\mathrm{cm}}^{-3}$). The line ratio is sensitive to the XDR contribution, since X-rays are able to penetrate deep into the cloud and warm all the neutral carbon, thus lowering the [C i]${}_{609/371}$ ratio as the temperature increases (Meijerink et al. 2007; Ferland et al. 2013). In this section, we use our sample to review the possible correlation between density, ionization, and neutral gas temperature already presented in S15. Thus, Figures 2(a), (b), and 3(a) are extensions of the work initiated in S15 for a smaller sample of galaxies.

In Figure 2(a), the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ ratio—sensitive to the hardness of the ionizing radiation—is compared to the neutral gas temperature-sensitive [O i]${}_{145/63}$ ratio. In the right panel, the [S iii]${}_{33.5/18.7}$ ratio—sensitive to the density in the 102${10}^{4}\,{\mathrm{cm}}^{-3}$ range—is compared to the same temperature tracer, [O i]${}_{145/63}$. As mentioned earlier, open symbols correspond to ISO/LWS observations (Brauher et al. 2008), complemented with Spitzer/IRS measurements in the literature, for those objects without Herschel/PACS spectroscopy. As expected, AGNs and starbursts are separated by the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ ratio, sensitive to the ionization parameter, but no correlation is found between both ratios, i.e., harder radiation fields are not associated with a warmer neutral ISM. On the other hand, density and temperature were tentatively correlated in S15. The extended sample does not show a significant correlation between temperature and density traced by [S iii]${}_{33.5/18.7}$, with log [S iii]33.5/18.7 = (0.45 ± 0.56) log [O i]145/63 + (0.79 ± 0.60) and a low Spearman's rank correlation coefficient of $R=0.32$.

A likely explanation for the lack of correlation between these line ratios is the different ISM phases traced, i.e., ionized gas in [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ and [S iii]${}_{33.5/18.7}$ versus neutral gas in [O i]${}_{145/63}$. Figures 2(a) and (b) show that a harder ionizing spectrum or a denser ionized gas does not imply a hotter temperature for the associated neutral gas, possibly suggesting a more complex ISM structure. Furthermore, we note that [O i]${}_{145/63}$ values $\gtrsim 0.1$ would imply optically thick emission (Tielens & Hollenbach 1985), but can be reconciled with optically thin emission if the effect of heavy element opacity is considered (Rubin 1983, 1985). Specifically, the [O i]${}_{63\mu {\rm{m}}}$ line can be affected by self-absorption even for small amounts of cold foreground gas (${N}_{{\rm{H}}}\sim 2\times {10}^{20}\,{\mathrm{cm}}^{-2};$ Liseau et al. 2006). This effect is particularly dramatic in the cases of Arp 220 (Figures 2(a), (b), and 3(b)), NGC 4945 and NGC 4418 (Figure 3(b)), and IRAS17208–0014 (Figure 2(b)). Their spectra for the central spaxel array—available as online material in their figures 1.77 (NGC 4418), 1.95 (NGC 4945), 1.135 (IRAS17208–0014), and 1.127 (Arp 220)—show a prominent absorption component in the [O i]${}_{63\mu {\rm{m}}}$ line (see also Figure 11 in González-Alfonso et al. 2012, for the cases of Arp 220 and NGC 4418), or even in the [O iii]${}_{88\mu {\rm{m}}}$ line (IRAS17208–0014). A self-absorption contribution to the [O i]${}_{63\mu {\rm{m}}}$ line would also play against the correlations explored in this section. Furthermore, shocks might also play a role in the [O i]${}_{63\mu {\rm{m}}}$ line emission (see Hollenbach & McKee 1989; Lutz et al. 2003).

Figure 3(a) shows the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ versus [S iii]${}_{33.5/18.7}$ line ratios, i.e., ionization versus density of the ionized gas. S1 galaxies, S1h galaxies, and dwarf galaxies, where the ionization is stronger, show tentatively lower [S iii]${}_{33.5/18.7}$ ratios with median values and median absolute deviations of 1.76 ± 1.02, 1.59 ± 0.50, and 1.53 ± 0.33, respectively, when compared to starburst galaxies with 2.86 ± 2.40. However, the difference is not statistically significant due to the high dispersion in the [S iii]${}_{33.5/18.7}$ ratio, thus no correlation is found between ionization and density. This lack of correlation can be explained by the combination of different ionization parameter values and densities in the Cloudy photoionization models shown in Figure 3(a). Most of the S1, S1h, and dwarf galaxies in our sample are in agreement with photoionization models in the $-3.0\lt \mathrm{log}U\lt -2.0$ and $1.0\lt \mathrm{log}(n/{\mathrm{cm}}^{-3})\lt 4.0$ ranges, starburst galaxies span the $-3.5\lt \mathrm{log}U\lt -2.5$ and $1.0\lt \mathrm{log}(n/{\mathrm{cm}}^{-3})\lt 3.0$ ranges, while S2 galaxies and LINERs are found in both the AGN and the starburst domains. Note that galaxies with [S iii]${}_{33.5/18.7}$ ratios $\gtrsim 2$ are above the low-density limit, thus [S iii]${}_{33.5/18.7}$ is no longer sensitive to density for those cases.

Figure 3(b) shows the [C i]${}_{609/371}$ versus the [O i]${}_{145/63}$ line ratios, both sensitive to the neutral gas temperature in the PDR, where [C i]${}_{609/371}$ decreases and [O i]${}_{145/63}$ increases with increasing temperature. Again, our sample does not show any correlation between these two ratios, probably due to the different density ranges probed by the [O i]${}_{145/63}$ (${n}_{{\rm{H}}}\lesssim {10}^{4}\,{\mathrm{cm}}^{-3}$) and the [C i]${}_{609/371}$ (${n}_{{\rm{H}}}\gtrsim {10}^{4}\,{\mathrm{cm}}^{-3}$) ratios, and the above mentioned self-absorption in the [O i]${}_{63\mu {\rm{m}}}$ line. The [C i]${}_{\mathrm{371,609}\mu {\rm{m}}}$ lines probe a denser and colder gas inside the cloud when compared to the [O i]${}_{\mathrm{63,145}\mu {\rm{m}}}$ lines. Overall, most of the galaxies show the typical values expected for PDR emission, including AGN galaxies, which show similar ratios when compared to starburst galaxies. A median [C i]${}_{609/371}$ ratio of 0.45 with a median absolute deviation of 0.11 are measured for the eight starburst galaxies detected, comparable to the median 0.53 ± 0.21 obtained for Seyfert and LINER galaxies. This implies that XDR emission, whose ratio is expected to be around [C i]${}_{609/371}\lesssim 0.19$ (see Table 2 in Ferland et al. 2013; also simulations by Meijerink et al. 2007), does not have an important contribution—within the Herschel/SPIRE aperture of ∼35''—to the neutral carbon emission for the AGN galaxies in our sample, which show [C i]${}_{609/371}$ ratios typical for PDR and similar to starburst galaxies.

5.2. Density Stratification

The observed properties of the emission lines are closely linked to the physical conditions of the region where they originated. Lines with a high-ionization potential such as [Ne v]${}_{14.3,24.3\mu {\rm{m}}}$ ($97.1\,\mathrm{eV}$) can only be produced in the innermost region of the NLR, as they are powered by the strong ionizing spectrum of the AGN. Thus they are expected to trace a denser gas when compared with [O iii]${}_{\mathrm{52,88}\mu {\rm{m}}}$ ($35.1\,\mathrm{eV}$) and [S iii]${}_{18.7,33.5\mu {\rm{m}}}$ ($23.3\,\mathrm{eV}$), produced in the outer NLR and H ii regions, and [N ii]${}_{\mathrm{122,205}\mu {\rm{m}}}$ ($14.5\,\mathrm{eV}$), produced also in low-density H ii regions. This density stratification was shown by S15 and is reviewed here using our larger sample.

Table 10 shows, for each object in the sample, the [Ne v]${}_{24.3/14.3}$, [O iii]${}_{88/52}$, [S iii]${}_{33.5/18.7}$, and [N ii]${}_{205/122}$ line ratios, where available, and their associated uncertainties, the average electron densities derived for each ratio—assuming purely collisionally excited gas at a temperature of ${10}^{4}\,{\rm{K}}$, as in S15—and the density error due to the line ratio uncertainty. Figure 4 shows the electron density versus the ionization potential for all the galaxies in the sample with at least a pair of lines detected of the same species. Our extended sample confirms the correlation found by S15: lines ionized by a harder radiation field originate in gas with higher densities (∼103–104 cm−3), while softer radiation is associated with lower densities (10–102 cm−3). This correlation is quantified by (1) a weighted least-square fit of the form $\mathrm{log}y=(1.38\pm 0.16)\mathrm{log}x+(0.27\pm 0.25)$, ${\chi }^{2}\ =24.6$, a correlation coefficient $R=0.71$ (black-solid line; blue-dashed lines indicate the uncertainty associated with the fit), (2) a linear regression using the Kaplan–Meier residuals12 of the form $\mathrm{log}y\,=(1.51\pm 0.13)\,\mathrm{log}x\mbox{--}0.10$ (red dotted–dashed line; red dotted lines indicate the associated uncertainty). The Kaplan–Meier residuals method, included in the asurv package13 , is further described in Isobe et al. (1986). A total of 155 pairs of lines were used for the weighted least-square fit, while 248 pairs of lines—93 of them as upper limits—were included in the Kaplan–Meier residuals fit. We note that the upper limits on the density are not caused by a detection limit in the line observations, but rather due to the limited range where the line ratio is sensitive to density (see Table 10, also Figure 2 in S15). The upper limits reported in Table 10 and Figure 4 derived from the [N ii]${}_{205/122}$, [S iii]${}_{33.5/18.7}$, and [Ne v]${}_{24.3/14.3}$ line ratios denote density values below the corresponding low-density limits of each of these three line ratios: $1\,{\mathrm{cm}}^{-3}$, $10\,{\mathrm{cm}}^{-3}$, and $100\,{\mathrm{cm}}^{-3}$ for [N ii]${}_{205/122}$, [S iii]${}_{33.5/18.7}$, and [Ne v]${}_{24.3/14.3}$, respectively. Only in the case of NGC 7172, for which the [N ii]${}_{122\mu {\rm{m}}}$ was extracted from the PACS central spaxel, the upper limit on the density based on the [N ii]${}_{205/122}$ line ratio seems to be caused by the different apertures between PACS (9farcs4) and SPIRE (∼17'').

Table 10.  Density Determinations

Name Class [N ii] $\mathrm{log}({n}_{{\rm{e}}})$ [S iii] $\mathrm{log}({n}_{{\rm{e}}})$ [O iii] $\mathrm{log}({n}_{{\rm{e}}})$ [Ne v] $\mathrm{log}({n}_{{\rm{e}}})$
    205.18/121.89 ${\mathrm{cm}}^{-3}$ 33.48/18.71 ${\mathrm{cm}}^{-3}$ 88.36/51.82 ${\mathrm{cm}}^{-3}$ 24.32/14.32 ${\mathrm{cm}}^{-3}$
Haro11 Dwarf 0.66 ± 0.18 1.49 (1.25–1.75) ... ... ... ... ... ...
Mrk334 S1.8 ... ... 3.91 ± 1.20 <1.0 ... ... 0.85 ± 0.28 3.30 (2.60–3.75)
IRAS00198-7926 S2 ... ... 3.17 ± 0.49 <1.0 ... ... 0.93 ± 0.05 3.15 (3.04–3.25)
ESO012-G21 S1.5 ... ... 1.99 ± 0.26 <1.0 ... ... 1.45 ± 0.15 <2.0
Mrk348 S1h ... ... 1.74 ± 0.30 <2.23 ... ... 0.85 ± 0.11 3.29 (3.08–3.47)
IRAS00521-7054 S1h ... ... ... ... ... ... 0.42 ± 0.07 4.00 (3.88–4.13)
ESO541-IG12 S2 ... ... ... ... ... ... 0.52 ± 0.21 3.81 (3.48–4.22)
3C33 S1h ... ... ... ... ... ... 0.80 ± 0.22 3.38 (2.95–3.72)
NGC454E S2 ... ... ... ... ... ... 1.49 ± 0.14 <2.0
IRAS01364-1042 LIN ... ... 13.19 ± 3.11 <1.0 ... ... ... ...

Note. Density determinations from fine-structure emission lines from Herschel/PACS and SPIRE, and Spitzer/IRS observations.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 11.  Central Pixel, 3 × 3 Array, and 5 × 5 Array Line Fluxes

    Line fluxes (${10}^{-17}\,{\rm{W}}\,{{\rm{m}}}^{-2}$)  
# Name [O iii] [N iii] [O i] [O iii] [N ii] [O i] [C ii] Notes
    $51.81\,\,\mu {\rm{m}}$ $57.32\,\,\mu {\rm{m}}$ $63.18\,\,\mu {\rm{m}}$ $88.36\,\,\mu {\rm{m}}$ $121.90\,\,\mu {\rm{m}}$ $145.52\,\,\mu {\rm{m}}$ $157.74\,\,\mu {\rm{m}}$  
AGN
1 Mrk334 ... ... ... ... ... ... 66.56 ± 0.88 c
    ... ... ... ... ... ... 76.66 ± 1.77 × 3
    ... ... ... ... ... ... 68.98 ± 1.37 × 5
2 Mrk938 ... ... 65.41 ± 4.45 12.58 ± 2.38 ... 5.26 ± 0.38 65.95 ± 1.86 c
    ... ... 78.12 ± 6.36 19.24 ± 3.58 ... 5.26 ± 0.92 82.15 ± 1.51 × 3
    ... ... 71.14 ± 5.27 21.49 ± 5.71 ... 3.88 ± 1.13 78.59 ± 2.67 × 5
3 IRAS00182-7112 ... ... 7.10 ± 1.67 <2.04 ... ... ... c
    ... ... <9.73 <1.5 ... ... ... × 3
    ... ... 10.49 ± 6.69 <2.12 ... ... ... × 5
Starburst Galaxies
1 NGC253 <478.9 408.33 ± 121.33 1409.54 ± 70.02 332.54 ± 48.72 413.11 ± 41.42 181.22 ± 13.58 1277.46 ± 18.81 c
    389.21 ± 183.61 599.54 ± 178.33 2947.62 ± 85.69 625.21 ± 53.54 802.72 ± 28.14 389.61 ± 30.99 3392.69 ± 22.89 × 3
    438.08 ± 169.61 626.13 ± 150.87 3214.70 ± 94.83 678.20 ± 47.67 824.30 ± 24.25 393.91 ± 36.99 3985.12 ± 25.32 × 5
2 M74 ... ... 10.63 ± 2.16 <4.55 <3.38 ... 6.31 ± 0.93 c
    ... ... 20.58 ± 10.84 <14.16 5.85 ± 2.70 ... 43.12 ± 1.82 × 3
    ... ... <46.58 <10.16 12.60 ± 2.57 ... 80.18 ± 2.35 × 5
3 NGC891 ... ... 18.82 ± 6.38 ... 19.35 ± 0.80 ... ... c
    ... ... 85.54 ± 11.08 ... 73.28 ± 2.19 ... ... × 3
    ... ... 139.03 ± 10.01 ... 103.16 ± 5.08 ... ... × 5
No Lines Detected by PACS
1 3C33 ... ... ... ... ... <0.7 ... c
    ... ... ... ... ... <1.54 ... × 3
    ... ... ... ... ... <1.54 ... × 5
2 PG1114+445 ... ... <1.71 ... ... ... <0.9 c
    ... ... <3.06 ... ... ... <1.3 × 3
    ... ... <5.85 ... ... ... <2.18 × 5
3 ESO506-G27 ... ... ... ... ... <0.96 ... c
    ... ... ... ... ... <0.96 ... × 3
    ... ... ... ... ... <1.9 ... × 5

Note. Far-infrared fine structure line fluxes of the active, starburst, and dwarf galaxies observed by Herschel/PACS, extracted from the central pixel, the 3 × 3 array, and the full 5 × 5 array. Note that no point-source correction has been applied to the 5 × 5 fluxes.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

The slope of the linear regression obtained using these two different methods is consistent within the uncertainties, and also with the results in S15, being the error here a factor of approximately two smaller due to the improved statistics.

An additional line ratio, [Ne iii]${}_{36.0/15.6}$, was considered for this analysis. This ratio traces gas at very high densities in the $3\lesssim \mathrm{log}({n}_{{\rm{e}}}/{\mathrm{cm}}^{-3})\lesssim 6$ range, even higher than those probed by [Ne v]${}_{24.3/14.3}$, with a lower excitation ($\sim 41\,\mathrm{eV}$). Unfortunately, the [Ne iii]${}_{36.0\mu {\rm{m}}}$ is out of the Spitzer/IRS-SH spectral coverage for most of the galaxies in our sample. We compiled a few Spitzer measurements given by Bernard-Salas et al. (2009) and Rampazzo et al. (2013), plus ISO/SWS and LWS measurements14 from Sturm et al. (2002), Verma et al. (2003), and Satyapal et al. (2004). From a total of 29 galaxies (12 of them are upper limits in [Ne iii]${}_{36.0\mu {\rm{m}}}$), 28 have [Ne iii]${}_{36.0/15.6}$ ratios (16 galaxies) or upper limits (12 galaxies) below the low-density limit, i.e., $\lesssim {10}^{3}\,{\mathrm{cm}}^{-3}$. This is consistent with the scenario proposed in S15 and confirmed here, i.e., the gas with the highest densities is traced by the lines with the highest ionization potential. The measured [Ne iii]${}_{36.0/15.6}$ ratios are associated with a lower density gas when compared to the [Ne v]${}_{24.3/14.3}$ ratio.

5.3. Diffuse Low-excitation Photoionized Gas and Dense Neutral Gas

The UV radiation, emitted by the H ii regions in galactic star-forming regions or by the accretion disk in AGNs, can heat the nearby gas clouds via photoelectric emission by dust grains (e.g., Tielens & Hollenbach 1985). This originates the PDR, a transition region in the cloud between ionized and molecular gas (${N}_{{\rm{H}}}\lesssim {10}^{22}\,{\mathrm{cm}}^{-2}$), in which [C ii]${}_{158\mu {\rm{m}}}$ and [O i]${}_{63\mu {\rm{m}}}$ are the most important cooling lines, mainly excited by collisions with H2 molecules and H i (${n}_{{\rm{H}}}\gtrsim {10}^{3}$${10}^{4}\,{\mathrm{cm}}^{-3};$ Tielens & Hollenbach 1985; Liseau et al. 2006). Although most of the [C ii]${}_{158\mu {\rm{m}}}$ emission comes from neutral gas, it might also have a non-negligible contribution ($\sim 15 \% $) from low-excitation ionized gas (e.g., S15, Cormier et al. 2015).

The Figures 5(a), (b) and 6(a), (b) show the [C ii]158/[O i]63 ratio versus the [O iii]88/[O i]63, the [O iii]88/[N ii]122, the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$, and the [O iii]88/[O iv]${}_{25.9}$ line ratios, respectively. The [C ii]158/[O i]63 ratio in PDR depends mostly on density, though optical depth effects in [O i]${}_{63\mu {\rm{m}}}$ and H ii region contributing to the [C ii]${}_{158\mu {\rm{m}}}$ line might limit the ability of this ratio as a density tracer (Abel et al. 2007). No statistically significant difference has been found for the [C ii]158/[O i]63 ratio between the different subclasses. As suggested by the [C i]${}_{609/371}$ line ratio in Section 5.1, the XDR does not seem to contribute significantly, since the [C ii]158/[O i]63 line ratio is similar for both AGN and starburst galaxies.

Figure 5.

Figure 5. Line ratio diagrams (same notations as in Figure 2). Left (a): the [C ii]158/[O i]63 line ratio vs. the [O iii]88/[O i]63 ratio. Right (b): the [C ii]158/[O i]63 line ratio vs. the ionization-sensitive [O iii]88/[N ii]122 ratio.

Standard image High-resolution image

The [O iii]88/[O i]63 ratio in Figure 5(a) is almost insensitive to ionization because dwarf galaxies are the only class that is clearly distinguished with values of $\gtrsim 1$. Our models confirm this behavior, with a few S1 and S1h located in the AGN grid, while the rest of the Seyfert galaxies, LINERs, and starburst galaxies are found within the grid of starburst models. The [O iii]88/[N ii]122 ratio in Figure 5(b) is mildly sensitive to ionization, it is not able to separate Seyfert's from starburst galaxies, while dwarf galaxies are clearly distinguished due to their strong [O iii]${}_{88\mu {\rm{m}}}$ emission. Cloudy models in Figure 5(b) also show this dependency on ionization, with a superposition for AGN and dwarf galaxy models at low densities ($\lesssim 100\,{\mathrm{cm}}^{-3}$) and high-ionization parameter values ($\mathrm{log}U\gtrsim -3$). A high-excitation photoionized gas tracer, such as the [O iii]88/[O iv]${}_{25.9}$ line ratio in Figure 6(b), is able to separate starburst galaxies, Seyfert nuclei, and dwarf galaxies. Cloudy models in Figure 6(a) are in agreement with the observed [C ii]158/[O i]63 and [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ line ratios, and explain the higher [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ratios shown by dwarf galaxies. We note in Figure 6(b) that neither starburst models (in yellow; most of the grid falls outside the figure due to the high [O iii]88/[O iv]${}_{25.9}$ predicted ratios) nor dwarf galaxy models (in purple) are able to reproduce the observed [O iii]88/[O iv]${}_{25.9}$ line ratios, which are overestimated by more than an order of magnitude. The model predictions for oxygen and neon high-ionization ratios will be further discussed in Section 5.4.

Figure 6.

Figure 6. Line ratio diagrams (same notations as in Figure 2). Left (a): the [C ii]158/[O i]63 line ratio vs. the ionization-sensitive [Ne iii]${}_{15.5}$/[Ne ii]${}_{12.8}$ ratio. Right (b): the [C ii]158/[O i]63 line ratio vs. the ionization-sensitive [O iii]88/[O iv]${}_{25.9}$ ratio.

Standard image High-resolution image

Figures 7(a) and (b) show the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ratio versus the [C ii]158/[N ii]122 and [C ii]158/[N ii]205 ratios, respectively. The latter ratios are sensitive to the fraction of PDR to low-excitation photoionized gas emission, since [C ii]${}_{158\mu {\rm{m}}}$ originated mostly in PDR while [N ii]${}_{\mathrm{122,205}\mu {\rm{m}}}$ are emitted by ionized gas. The PDR emission is inherently included in the Cloudy models whose radial integration have been extended to the low temperature of $T=50\,{\rm{K}}$ (yellow grid in Figure 7), while models that stopped at $T=1000\,{\rm{K}}$ denote the ratios predicted for pure ionized gas (dark orange grid). Similarly, we compare dwarf galaxy models stopped at $T=1000\,{\rm{K}}$ (dark violet grid) and those extended to $T=50\,{\rm{K}}$, thus including the PDR emission (light violet grid). The difference in the [C ii]158/[N ii]${}_{\mathrm{122,205}}$ ratios between pure photoionized models and those including the PDR is even more extreme in the case of dwarf galaxies. When compared to the models, the observed ratios suggest that most of the [C ii]${}_{158\mu {\rm{m}}}$ emission originated in the PDR, in agreement with the results of Cormier et al. (2015) for the same sample of dwarf galaxies included here. These authors estimate that only $\approx 15 \% $ of [C ii]${}_{158\mu {\rm{m}}}$ originated in H ii regions. The observed line ratios in Figure 7(a) also reveal a possible dependence, for starburst galaxies, of the [C ii]158/[N ii]122 ratio on the ionization, traced by the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ratio. A harder ionizing spectrum might produce a larger amount of ${{\rm{N}}}^{2+}$ in the difuse ionized gas, thus decreasing the relative contribution of the [N ii]${}_{122\mu {\rm{m}}}$ emission. Consequently, the high [C ii]158/[N ii]122 ratios found in dwarf galaxies are explained by the high excitation in these objects (Cormier et al. 2015). This is also the case of NGC 1222, a near-solar-metallicity starburst that possibly experienced a merger with a low-metallicity companion (Beck et al. 2007). In this regard, the [C ii]158/([N ii]122+[N iii]57) ratio might be more appropriate to evaluate the relative PDR to ionized gas contribution. Unfortunately, the [N iii]${}_{57\mu {\rm{m}}}$ line has been observed only for six starburst galaxies (four detections), thus we do not have enough statistics to test the proposed PDR-to-ionized gas ratio.

Figure 7.

Figure 7. Line ratio diagrams (same notations as in Figure 2). The grid of models in yellow shows the usual starburst+PDR models, with the integration going down to the temperature of $50\,{\rm{K}}$, while the grid in orange shows the pure photoionized models (${T}_{\mathrm{stop}}=1000\,{\rm{K}}$). Similarly, the dark purple grid corresponds to dwarf galaxy models stopped at ${T}_{\mathrm{stop}}=1000\,{\rm{K}}$, while the light purple grid shows the dwarf+PDR models, stopped at $50\,{\rm{K}}$. Left (a): the [Ne iii]${}_{15.5}$/[Ne ii]${}_{12.8}$ line ratio vs. the [C ii]158/[N ii]122 ratio. Right (b): the [Ne iii]${}_{15.5}$/[Ne ii]${}_{12.8}$ line ratio vs. the [C ii]158/[N ii]205 ratio.

Standard image High-resolution image

Figure 8 shows the [N ii]122/[C i]371 line ratio versus the ionization-sensitive ratios [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ and [O iv]${}_{25.9}$/[O iii]88. The [N ii]122/[C i]371 ratio accounts for the relative contribution of low-excited photoionized gas to neutral gas in the PDR. Overall, starburst galaxies seem to be biased toward higher values of the [N ii]122/[C i]371 ratio, always being found $\gt 7$, with a median value of 16.6 and a median absolute deviation of 7.4. LINERs show tentatively lower values with an average of [N ii]122/[C i]${}_{371}\sim 5.1$ and a deviation of 3.0. A possible explanation would be the presence of an XDR in LINERs, which would contribute with a brighter [C i]${}_{371\mu {\rm{m}}}$ line emission. However, the [C i]609/[C i]371 ratio in Figure 3(b) showed typical PDR values, suggesting that the XDR contribution to the atomic carbon lines is negligible in our sample at the angular resolution of the SPIRE data. A decreasing [N ii] contribution in favor of [N iii] emission at higher excitation would explain the lower [N ii]122/[C i]371 ratios found in LINERs, if a harder ionizing spectrum is assumed for LINERs when compared to starburst galaxies. However, Figure 7 does not show a dependence of the [C ii]158/[N ii]122 ratio with the ionization parameter for LINERs, as was the case for starburst galaxies. An alternative explanation for the different [N ii]122/[C i]371 ratios is that starburst galaxies have a larger filling factor of diffuse low-excitation photoionized gas within the SPIRE aperture when compared to LINER galaxies. In this regard, we note that most of the LINERs in Figure 8 are found in early-type spirals, while the LINER with the highest ratio, NGC 1097 (25 ± 1), has a bright circumnuclear starburst ring unresolved within the SPIRE aperture of ∼35''. A larger sample of LINERs and starburst galaxies would be needed in order to prove this behavior.

Figure 8.

Figure 8. Line ratio diagrams (same notations as in Figure 2). Left (a): the [N ii]122/[Ne ii]${}_{12.8}$ line ratio vs. the [N ii]122/[C i]371 ratio. Right (b): the [O iv]${}_{25.9}$/[O iii]88 line ratio vs. the [N ii]122/[C i]371 ratio.

Standard image High-resolution image

5.4. Discriminating AGNs and Starbursts

5.4.1. Oxygen and Neon Line Ratios

High to low-excitation IR line ratios such as [Ne v]${}_{14.3}$/[Ne ii]${}_{12.8}$ and [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ have been commonly used to measure the relative contributions of AGNs and star formation in active galaxies (Sturm et al. 2002; Armus et al. 2007; Meléndez et al. 2008; Tommasin et al. 2008). Since photoionization by young stars cannot have a significant contribution to transitions with a high-ionization potential ($\gtrsim 50\,\mathrm{eV}$), as the AGN does, these line ratios are sensitive to the relative strength of the AGN compared to the starburst. So far, these diagnostics have been tested by comparing samples of active nuclei with starburst galaxies (e.g., Treyer et al. 2010). In this work, we compare AGN and starburst galaxies but also low-metallicity dwarf galaxies, and therefore we propose a new diagnostic including the long-wavelength [O iii]${}_{88\mu {\rm{m}}}$ line measured by Herschel/PACS.

In S15, we showed that the [O iv]${}_{25.9}$/[O iii]88 line ratio is a good tracer of the relative AGN to star formation contribution. These oxygen lines have critical densities in the 102${10}^{4}\,{\mathrm{cm}}^{-3}$ range and their ratio is sensitive to the hardness of the ionizing radiation in the 54.94–$35.12\,\mathrm{eV}$ range (4–$2.6\,\mathrm{Ry}$). Thus, the [O iii]${}_{88\mu {\rm{m}}}$ line can be excited by both young stars and AGN activity, while the [O iv]${}_{25.9\mu {\rm{m}}}$ transition becomes much more intense under the presence of an AGN (Sturm et al. 2002; Diamond-Stanic et al. 2009; Spinoglio et al. 2012a). The [O iv]${}_{25.9\mu {\rm{m}}}$ line is brighter than the [Ne v]${}_{14.3,24.3\mu {\rm{m}}}$ lines, and can still be detected in star formation environments. This makes this line ideal to identify the location of the different populations, i.e., AGNs and starbursts as well as dwarf galaxies, in the diagnostic diagrams.

In Figures 9(a) and (b) we show the usual [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ ratio (e.g., Genzel et al. 1998; Sturm et al. 2002; Meléndez et al. 2008) and the [O iv]${}_{25.9}$/([Ne ii]${}_{12.8}$+[Ne iii]${}_{15.6}$) ratio, respectively, as a function of the [O iv]${}_{25.9\mu {\rm{m}}}$ luminosity (${L}_{[{\rm{O}}{\rm{IV}}]25.9}$). If only AGN and starburst galaxies were taken into account in Figure 9(a), a gradient would be present from lower values of [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}\sim 2\times {10}^{-2}$ and relatively low [O iv]${}_{25.9\mu {\rm{m}}}$ luminosities of $\sim {10}^{38}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, where star formation dominates the ratio in H ii region galaxies, to [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}\sim 10$ and high ${L}_{[{\rm{O}}{\rm{IV}}]25.9}\sim {10}^{43}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$, where AGN activity is the dominant source of radiation (blue and pink triangles, and red squares). LINERs (green dots) are found in the intermediate region between star formation and Seyfert's, while five S2 galaxies (red squares) appear to be dominated by star formation. However, this correlation no longer applies if we consider the sample of dwarf galaxies, which are able to reach the AGN-dominated regime in this diagram, showing [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ line ratios similar to those of Seyfert's, but at low [O iv]${}_{25.9\mu {\rm{m}}}$ luminosities compared to AGNs. The least squares regression still shows a trend with higher line ratios found at higher luminosities,

Equation (1)

but the large dispersion introduced by dwarf galaxies, which results in a relatively low Spearman's rank correlation coefficient of $R=0.56$, limits the validity of the [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ line ratio as an AGN/starburst tracer.

Figure 9.

Figure 9. Line ratios sensitive to the relative AGN/starburst contribution (same notations as in Figure 2). Left (a): the [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ line ratio vs. the [O iv]${}_{25.9\mu {\rm{m}}}$ luminosity. Right (b): the [O iv]${}_{25.9}$/([Ne ii]${}_{12.8}$+[Ne iii]${}_{15.6}$) line ratio vs. the [O iv]${}_{25.9\mu {\rm{m}}}$ luminosity.

Standard image High-resolution image

The lower metallicity in dwarf galaxies allows the presence of a hotter main sequence and thus a harder radiation field, increasing the filling factor of low-density ionized gas (Schaller et al. 1992; Cormier et al. 2015). In particular, the ionizing spectra of dwarf galaxies is much harder just after the Lyman break, which also increases the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ line ratio (and [S iv]${}_{10.5}$/[S iii]${}_{18.7}$, see below) when compared to a solar-metallicity starburst (O'Halloran et al. 2006; Hao et al. 2009, see Figure 7(a)). Our Figure 9(a) demonstrates that the [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ line ratio does not probe the relative AGN-to-starburst contribution at low metallicities, since a large fraction of the star formation component is probably emitted in the form of [Ne iii]${}_{15.6\mu {\rm{m}}}$ (see Ho & Keto 2007). This casts doubts on our ability to separate AGN and starburst components, e.g., in composite objects with extreme star formation bursts or mergers involving a chemically unevolved dwarf companion, and prompts us to find a more suitable diagnostic ratio, which also should be valid for these scenarios that are expected to be even more common at higher redshift (e.g., Alavi et al. 2014; Atek et al. 2014, 2015).

Following the previous argument, we show the [O iv]${}_{25.9}$/([Ne ii]${}_{12.8}\,+\,$[Ne iii]${}_{15.6}$) ratio in Figure 9(b), since the sum of the two neon lines would be a more reliable tracer of the star formation contribution also for dwarf galaxies. In this case, both dwarf and star-forming galaxies have, on average, lower [O iv]${}_{25.9}$/([Ne ii]${}_{12.8}\,+\,$[Ne iii]${}_{15.6}$) ratios of $\sim 4\times {10}^{-2}$, when compared to Seyfert galaxies (∼1), though some overlap between dwarfs/starburst galaxies and Seyfert galaxies still remains in the range of [O iv]${}_{25.9}$/([Ne ii]${}_{12.8}\,+\,$[Ne iii]${}_{15.6}$) ∼0.1 to 0.5. The least-square regression confirms the trend of increasing rations with increasing ${L}_{[{\rm{O}}{\rm{IV}}]25.9}$,

Equation (2)

and the correlation coefficient improves ($R=0.66$), with regard to the [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ ratio.

In Figure 10(a), we show the ability of the [O iv]${}_{25.9}$/[O iii]88 line ratio to separate the contribution of AGN from starburst and dwarf galaxies in our sample, as a function of the [O iv]${}_{25.9\mu {\rm{m}}}$ luminosity. The least squares regression is steeper when compared to the previous line ratios,

Equation (3)

with a correlation coefficient of $R=0.7$. This is the best diagnostic ratio to distinguish Seyfert activity from star formation, with a minimum overlap between AGN-dominated ratios ([O iv]${}_{25.9}$/[O iii]${}_{88}\gtrsim 0.3$) and those dominated by starburst or dwarf galaxies. We note that dwarf galaxies are indistinguishable from starburst galaxies in this diagram, since their ionizing spectra are not hard enough in the 54–$35\,\mathrm{eV}$ range to produce a higher fraction of [O iv]${}_{25.9}$/[O iii]88. A few AGNs are dominated by the starburst component, thus their [O iv]${}_{25.9}$/[O iii]88 ratios are in the same range as starburst galaxies. This will be discussed further in Section 5.4.2. LINERs appear to show a dual behavior, with the three brightest sources in the Seyfert domain and the rest of them located in the starburst domain, with a line ratio of $5\times {10}^{-3}\,\lesssim $ [O iv]${}_{25.9}$/[O iii]${}_{88}\lesssim 0.3$. This suggests that the [O iii]${}_{88\mu {\rm{m}}}$ line for the fainter LINERs in our sample might include an important contribution from their host galaxies.

Figure 10.

Figure 10. Line ratios sensitive to the relative AGN/starburst contribution (same notations as in Figure 2). Left (a): the [O iv]${}_{25.9}$/[O iii]88 line ratio vs. the [O iv]${}_{25.9\mu {\rm{m}}}$ luminosity. Right (b): the [O iv]${}_{25.9}$/[O iii]88 line ratio vs. the ratio of absorption-corrected 2–$10\,\mathrm{keV}$ X-ray flux to far-IR flux (derived from the continuum adjacent to the [C ii]${}_{158\mu {\rm{m}}}$ line). Compton-thick objects are shown as lower limits in the horizontal axis.

Standard image High-resolution image

Finally, Figure 10(b) shows the [O iv]${}_{25.9}$/[O iii]88 ratio versus the absorption-corrected X-ray 2–$10\,\mathrm{keV}$ flux15 to far-IR flux ratio, the latter measured as the continuum adjacent to the [C ii]${}_{158\mu {\rm{m}}}$ line (Table 6). X-ray to far-IR flux ratios for Compton thick objects appear as lower limits. This diagram confirms the behavior observed in Figure 10(a) by using an independent estimate of the accretion to the total energy output. S1 and S1h galaxies are dominated by the AGN, and thus show the highest values for both the X-ray to far-IR ratio and the [O iv]${}_{25.9}$/[O iii]88 line ratios. Starburst galaxies are very faint in the 2–$10\,\mathrm{keV}$ range but bright far-IR emitters, and so they show low [O iv]${}_{25.9}$/[O iii]88 ratios. The least squares regression between both ratios is

Equation (4)

with a correlation coefficient of $R=0.79$.

We note that most of the [O iii]${}_{88\mu {\rm{m}}}$ fluxes are extracted from the 3 × 3 pixel array ($28\buildrel{\prime\prime}\over{.} 2\times 28\buildrel{\prime\prime}\over{.} 2$), an area that is a factor of 2.5 larger than the Spitzer/IRS-LH slit area ($11\buildrel{\prime\prime}\over{.} 1\times 22\buildrel{\prime\prime}\over{.} 3$). A possible contamination by star formation in the host galaxy would decrease the measured ratio for the central AGN. To minimize this effect, we have built the same diagrams using the fluxes extracted from PACS central pixel ($9\buildrel{\prime\prime}\over{.} 4\times 9\buildrel{\prime\prime}\over{.} 4$), i.e., using an aperture ∼3 smaller than the Spitzer/IRS slit for the [O iv]${}_{25.9\mu {\rm{m}}}$ line. The variations in the line ratios are within the errors for most of the objects (see Figure 13 and discussion in Appendix), thus we discard a strong contamination by star formation in the [O iii]${}_{88\mu {\rm{m}}}$ line outside of the innermost $\sim 10^{\prime\prime} $ for the galaxies in our sample.

5.4.2. A New AGN/Star Formation Diagnostic

The global view of the AGN-starburst diagnostics, discussed in the previous section, is fully represented in the diagnostic diagrams of Figures 11(a) and (b), which show the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ (41–$22\,\mathrm{eV}$ or 3.0–$1.6\,\mathrm{Ry}$) and the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ (35–$23\,\mathrm{eV}$ or 2.6–$1.7\,\mathrm{Ry}$) ratios, respectively, versus the [O iv]${}_{25.9}$/[O iii]88 ratio (35–$55\,\mathrm{eV}$ or 2.6–$4.0\,\mathrm{Ry}$). The filled symbols have been colored based on their relative [Ne v]${}_{14.3\mu {\rm{m}}}$ line flux to far-IR flux ratio, as a proxy of the relative AGN-to-far-IR luminosity. In the vertical axis, the neon and sulphur ratios probe the steepness of the ionizing continuum between $\sim 21\,\mathrm{eV}$, just after the Lymann break ($13.6\,\mathrm{eV}$), and $41\,\mathrm{eV}$, before the He ii ionization edge ($54\,\mathrm{eV}$). The stellar spectra are very sensitive to the metallicity in this range (Schaller et al. 1992; Mokiem et al. 2004), thus dwarf galaxies exhibit large [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ and [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ ratios, even larger than Seyfert galaxies. Therefore, the vertical axis in Figures 11(a) and (b) is especially sensitive to the excitation originated by thermal processes. A large dropout in the spectra of dwarf galaxies occurs shortward of the He ii ionization edge, thus their [O iv]${}_{25.9}$/[O iii]88 ratios are far from those exhibited by Seyfert galaxies, and very similar to the values found in starburst galaxies. LINERs are able to produce both higher [O iv]${}_{25.9}$/[O iii]88 and [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ([S iv]${}_{10.5}$/[S iii]${}_{18.7}$) line ratios with regard to starburst galaxies. The color scale confirms that AGNs with starburst-like [O iv]${}_{25.9}$/[O iii]88 line ratios are indeed galaxies in which the starburst contribution dominates both the far-IR and [O iii]${}_{88\mu {\rm{m}}}$ emission. The [O iv]${}_{25.9}$/[O iii]88 can be considered as a proxy for the relative contribution of the non-thermal photoionization.

Figure 11.

Figure 11. Ionization-sensitive line ratios (same notations as in Figure 2). Photoionization models of AGN, LINER, starburst galaxies, and dwarf galaxies are shown as blue, green, yellow, and purple grids, respectively. The logarithmic values of the density (${n}_{{\rm{H}}}$) and ionization potential (U) of the photoionization models are indicated in the figures. Symbols are color-coded according to their ${{\rm{F}}}_{[\mathrm{Ne}{\rm{V}}]14.3}/{{\rm{F}}}_{\mathrm{FIR}}$ flux ratio, when available (see the color bar). Top (a): the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ line ratio vs. the [O iv]${}_{25.9}$/[O iii]88 ratio. Bottom (b): the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ line ratio vs. the [O iv]${}_{25.9}$/[O iii]88 ratio.

Standard image High-resolution image

In Figures 11(a) and (b), we compare the observed line ratios with our Cloudy photoionization models for AGNs (blue grid), LINERs (green grid), starburst galaxies (yellow grid), and dwarf galaxies (purple grid). From this comparison, we conclude the following.

  • 1.  
    Observed Seyfert galaxies with the highest [O iv]${}_{25.9}$/[O iii]88 and [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ratios (or [S iv]${}_{10.5}$/[S iii]${}_{18.7}$) are consistent with AGN models with densities in the $\mathrm{log}({n}_{{\rm{H}}}/{\mathrm{cm}}^{3})=2.0$ to 4.0 range and ionization parameters in the $\mathrm{log}U=-2.0$ to −3.0 range.
  • 2.  
    The transition from AGN-dominated galaxies to star-forming galaxies defines a tail in the two diagrams that can be explained by (1) decreasing ionization ratios (Figure 11(b)), (2) decreasing the power-law index, as shown by LINER models, or (3) increasing contribution to the [O iii]${}_{88\mu {\rm{m}}}$ line by star formation in the host galaxy, as show by the [Ne v]${}_{14.3\mu {\rm{m}}}$ to far-IR flux ratio.
  • 3.  
    Extreme dwarf galaxy models with very high and thus unlikely densities ($\gtrsim {10}^{4}\,{\mathrm{cm}}^{-3}$) are needed in order to reproduce the [O iv]${}_{25.9}$/[O iii]88 ratios measured in dwarf galaxies. These same ratios cannot be reproduced by the softer ionizing spectra of starburst models, which underestimate the [O iv]${}_{25.9}$/[O iii]88 ratio by at least an order of magnitude even at the highest values of density and ionization parameter. A solar-metallicity starburst model with $1\,\mathrm{Myr}$ would increase the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ line ratio by $\sim 40 \% $ relative to the $20\,\mathrm{Myr}$ starburst model, but would not significantly increase the [O iv]${}_{25.9}$/[O iii]88 line ratio, since the contribution of the younger population does not increase considerably $\gtrsim 54\,\mathrm{eV}$.
  • 4.  
    LINER ratios in Figures 11(a) and (b) can be explained by a much steeper power law when compared to Seyfert nuclei with $\alpha \gtrsim -3.5$ and $-2.5\lesssim \mathrm{log}U\lesssim -3.5$, in agreement with the absence of the big blue bump in their spectra (Ho 1996, 2008) and the predicted recession of the innermost part of the accretion disk in radiatively inefficient AGNs (Narayan & Yi 1995). Additionally, shocks might play an important role in the excitation of IR lines in LINERs, as shown by Sturm et al. (2006).

To reproduce the observed line ratios in dwarf galaxies for a gas with $\mathrm{log}({n}_{{\rm{H}}}/{\mathrm{cm}}^{3})=2.0$–3.0 would require a ionizing spectrum in the ∼55–$35\,\mathrm{eV}$ range harder than that provided by our Starburst99 model ($1\,\mathrm{Myr}$ and $Z=0.004$), which relies on current stellar population synthesis models. The lack of ionizing photons above $\approx 40\,\mathrm{eV}$, known as the "[Ne iii]" problem, has been discussed in the literature (e.g., Sellmaier et al. 1996; Simón-Díaz & Stasińska 2008; Zastrow et al. 2013), especially at low metallicities where high-excitation lines are particularly bright (Stasińska et al. 2015). However, Figures 11(a) and (b) prove that this disagreement is even more dramatic for starburst galaxies: the observations show that starburst galaxies are able to produce a much brighter [O iv]${}_{25.9\mu {\rm{m}}}$ emission when compared to the model predictions. This line has an IP slightly above the He ii ionization edge at $54\,\mathrm{eV}$, which cannot be produced by hot stars in the main sequence. [O iv]${}_{25.9\mu {\rm{m}}}$ was previously detected in starburst galaxies by Lutz et al. (1998), and its origin has been associated to either X-ray emission produced by shocks in the stellar atmospheres or X-ray binaries (Sellmaier et al. 1996; Izotov et al. 2012; Stasińska et al. 2015), or to strong winds produced during the Wolf–Rayet (WR) phase of massive stars with ∼3–$5\,\mathrm{Myr}$ (Schaerer & Stasińska 1999). The latter could be the case for dwarf galaxies, which usually show other WR spectral features like He ii $4686\,{\rm{\mathring{\rm A} }}$ emission, but it seems hardly the explanation for low-excitation starburst galaxies.

Starburst models with ages in the 3.5–$5\,\mathrm{Myr}$, at the peak of the WR feedback contribution, are able to reproduce the observed [O iv]${}_{25.9}$/[O iii]88 ratios, but at the cost of overpredicting [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ratios by a factor of ∼30, due to their harder ionizing spectra. Considering this, and the short duration of the WR phase in the lifetime of a star-forming region ($\lesssim 1.5\,\mathrm{Myr}$), we consider it unlikely that the origin of high [O iv]${}_{25.9}$/[O iii]88 line ratios is caused by WR stars, especially in our samples of starburst galaxies. The similar [O iv]${}_{25.9}$/[O iii]88 ratios found for dwarf and starburst galaxies in our sample suggest that the strength of the ionizing spectrum at $\sim 54\,\mathrm{eV}$ in star-forming regions scales with the [O iii]${}_{88\mu {\rm{m}}}$ emission because this line ratio does not vary with the metallicity of the stellar population.

High-ionization lines above $54\,\mathrm{eV}$ are expected to be more sensitive to the contribution of shocks in a starburst galaxy than lower ionization lines, since the stellar contribution is negligible above that value. Among the low ionization and neutral lines, shocks can contribute significantly to the [O i]${}_{63\mu {\rm{m}}}$ emission (Hollenbach & McKee 1989), as shown by Lutz et al. (2003) for the case of NGC 6240, an LIRG merger. Furthermore, the [O i]${}_{63\mu {\rm{m}}}$ line can be affected by self-absorption, as discussed in Section 5.1, and depends on the PDR temperature. All of these contributions make it difficult to evaluate the possible effect of shocks in our starburst sample, and also limit the utility of this line in the diagnostic diagrams. We expect that the shock contribution to the mid- and far-IR lines in this study, if present, would be significant only for line ratios involving the [O iv]${}_{25.9\mu {\rm{m}}}$, the [Ne v]${}_{14.3,24.3\mu {\rm{m}}}$, and/or the [O i]${}_{63\mu {\rm{m}}}$ lines.

5.5. Metallicity

As mentioned in Section 1, to use diagnostics based on mid- and far-IR lines is a very useful tool for the study of galaxies at high redshift. Of special interest are the metallicity-sensitive line ratios proposed by Nagao et al. (2011, 2012), which allow metallicity determinations in dust-embedded sources, e.g., ULIRGs (D. Rigopoulou 2016, private communication), where optical line ratios cannot be measured reliably due to the high extinction. In this section, we test two metallicity-sensitive diagnostics by comparing the IR fine-structure line ratios with metallicity estimates based on optical lines: the [O iii]88/[N iii]57 line ratio (Figure 12(a)), proposed by Nagao et al. (2011), and a new diagram based on the ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) ratio (Figure 12(b)), presented here for the first time.

Figure 12.

Figure 12. Metallicity-sensitive line ratios vs. metallicities derived from optical line ratios (same notations as in Figure 2). AGN models (in blue; $\mathrm{log}U=-2.0$, $\mathrm{log}({n}_{{\rm{H}}}/{\mathrm{cm}}^{-3})=3$) and dwarf galaxy models (in purple; $\mathrm{log}U=-3.5$, $\mathrm{log}({n}_{{\rm{H}}}/{\mathrm{cm}}^{-3})=3$) are also shown for different metallicities in the 1/20–$2\,{Z}_{\odot }$ range. Left (a): [O iii]88/[N iii]57 line ratio vs. optical metallicities. Right (b): ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) line ratio vs. optical metallicities. Note that the AGN and dwarf galaxy models shown here include the effect of Sulphur depletion (see Section 5.5).

Standard image High-resolution image

Most of the optical metallicities shown in Figures 12(a) and (b) were compiled from the literature and are referenced in Table 3. For a few objects without published metallicities but with published optical line fluxes ([N ii]${}_{\mathrm{6548,6584}{\rm{\mathring{\rm A} }}}$, [O ii]${}_{3727{\rm{\mathring{\rm A} }}}$, [O iii]${}_{\mathrm{4959,5007}{\rm{\mathring{\rm A} }}}$, Hα, Hβ), we derived the optical metallicities using the Pettini & Pagel (2004) and Pilyugin & Thuan (2005) calibrations. The blue lines show the metallicity dependence predicted by our Cloudy AGN models for $\mathrm{log}U=-2.0$ and $\mathrm{log}({n}_{{\rm{H}}})=3$ (blue-solid line). The purple-solid lines correspond to the dwarf galaxy models with $\mathrm{log}U=-3.5$ and $\mathrm{log}({n}_{{\rm{H}}})=3$.

Figure 12(a) shows a large dispersion with the observed [O iii]88/[N iii]57 ratios distributed around the model predictions. No clear trend is found for the individual populations, though the low number of star-forming galaxies with observations of both lines (three dwarfs plus seven starburst galaxies) limits the assessment of this diagnostic. Our AGN models show a relatively constant ratio, while dwarf galaxy models show a decreasing ratio at super-solar metallicities, in agreement with the models in Nagao et al. (2011). This is related to the softening of the ionizing spectrum at higher metallicities, i.e., the [O iii]88/[N iii]57 in our models is mainly driven by the metallicity of the ionizing stellar population, instead of the metallicity of the ionized gas. Continuous starburst models (not shown) show a behavior very similar to that of dwarf galaxy models, with a slightly flatter dependency on metallicity. The dispersion of the observed ratios could be explained by different density and ionization values, which should be determined independently in order to use the [O iii]88/[N iii]57 ratio as a diagnostic for metallicity.

The ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) ratio in Figure 12(b) shows a clear correlation ($R=0.69$) with optical metallicities, of the form

Equation (5)

If only the dwarf and starburst galaxies are considered, the correlation is maintained with a higher correlation coefficient of $R=0.89$. As shown in Section 5.4, both neon and sulphur ratios are very sensitive to the metallicity-dependent excitation of the stellar population ionizing the gas. Thus, the ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}\,+\,$[S iii]${}_{18.7}$) ratio should cancel most of this dependency since [Ne ii]${}_{12.8\mu {\rm{m}}}$ ([S iii]${}_{18.7\mu {\rm{m}}}$) will turn into [Ne iii]${}_{15.6\mu {\rm{m}}}$ ([S iv]${}_{10.5\mu {\rm{m}}}$) with increasing ionization. The dependency of the ([Ne iii]${}_{15.6}\,+\,$[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) ratio could be explained by the depletion of sulphur on dust grains (e.g., Verma et al. 2003), which causes an increasing sulphur deficiency with increasing metallicity. In this scenario, the total ([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) emission does not change significantly, since depletion maintains the sulphur abundance roughly constant with increasing metallicity (see Figure 4 in Verma et al. 2003), while the total ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$) increases with the neon abundance. In order to test this effect, we modified our Cloudy models by assuming a constant sulphur abundance above $Z=0.004$, while the neon abundance increases with the metallicity. Instant starburst models (purple solid line in Figure 12), equivalent to dwarf galaxy models at $Z=0.004$, show a similar trend as that shown by the observations, i.e., the ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) ratio increases with metallicity due to the sulphur depletion. Seyfert galaxies show line ratios in agreement with those of starburst galaxies, as well as AGN models (blue-solid line). This suggests that the ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) ratio is a robust extinction-free metallicity tracer for both AGN and starburst galaxies. This is very important for the study of the dusty obscured galaxy evolution at redshift $1\lt z\lt 4$, at the peak of the star formation and black hole accretion activity, that will be addressed by future space missions, e.g., SPICA (Swinyard et al. 2009).

6. SUMMARY AND CONCLUSIONS

In this work, we have presented the complete database of far-IR fine-structure lines observed by Herschel/PACS during its 3.5 years of operational life for 170 AGN-classified galaxies in the Véron-Cetty & Véron (2010) catalog. In order to complete the full mid- to far-IR spectra in the 10–$600\,\,\mu {\rm{m}}$ range, we collected published fine-structure line fluxes measured with Spitzer/IRS and Herschel/SPIRE. As a comparison sample, we compiled an equivalent database for starburst galaxies extracted from Bernard-Salas et al. (2009) and Goulding & Alexander (2009). Additionally, we included 43 dwarf galaxies from the DGS (Madden et al. 2013; Cormier et al. 2015) in order to probe a more extreme star formation environment at low metallicities.

The photoionization code Cloudy has been used to reproduce the physical conditions of the gas exciting the mid- to far-IR fine-structure lines. We produced four models using different ionizing spectra: a pure AGN model ($\alpha =-1.4;$ ${S}_{\nu }\propto {\nu }^{\alpha }$), a LINER model ($\alpha =-3.5$), a starburst galaxy model ($20\,\mathrm{Myr}$ continuum star formation with $Z={Z}_{\odot }$), and a dwarf galaxy model ($1\,\mathrm{Myr}$ instant starburst with $Z=1/5\,{Z}_{\odot }$). Calculations for the starburst and the dwarf galaxy models have been extended down to $T=50\,{\rm{K}}$ in order to include the PDR emission region.

The main results of this study are as follows.

  • 1.  
    The [O i]${}_{145/63}$ line ratio, used as a temperature tracer in the 100–$400\,{\rm{K}}$ range for the neutral gas in the PDR, does not show a clear correlation with the ionization nor the density of the ionized gas, traced by the [S iv]${}_{10.5}$/[S iii]${}_{18.7}$ and the [S iii]${}_{33.5/18.7}$ ratios, respectively. High [O i]${}_{145/63}$ ratios of $\gtrsim 0.1$ are found in objects with self-absorption in the [O i]${}_{63\mu {\rm{m}}}$ line profile.
  • 2.  
    The [C i]${}_{609/371}$ line ratio, sensitive to the temperature in the PDR (20–$100\,{\rm{K}}$), shows similar median values for AGN (0.53 ± 0.21) and starburst galaxies (0.45 ± 0.11). This is much higher than the ratios predicted from XDR simulations (∼0.15–0.19), suggesting that the neutral carbon lines in our AGN sample are likely dominated by relatively low-density PDR emission that originated in the ISM of the host galaxies.
  • 3.  
    The density stratification found in S15 is confirmed here with 155 pairs of lines from [N ii]${}_{205/122}$, [S iii]${}_{33.5/18.7}$, [O iii]${}_{88/52}$, and [Ne v]${}_{24.3/14.3}$ line ratios, plus 93 upper limits. Both the least squares fit and the Kaplan–Meier residuals fit are consistent with an increasing gas density—measured by the line ratios—with the ionization potential (IP) of the transition. This suggests that harder radiation fields are traced by the higher density gas found in the innermost part of the NLR.
  • 4.  
    The line ratios of S1 and S1h are always very similar, suggesting that both AGN types are indistinguishable from an IR perspective, thus S1h correspond to optically obscured S1 (e.g., Tommasin et al. 2010).
  • 5.  
    The [C ii]158/[N ii]${}_{\mathrm{122,205}}$ line ratios, sensitive to the relative contributions of the PDR and the low-excitation photoionized gas, suggest a major contribution ($\gtrsim 80 \% $) of PDR emission to the [C ii]${}_{158\mu {\rm{m}}}$ line. The observed ratios [C ii]158/[N ii]${}_{\mathrm{122,205}}\gtrsim 3$ are in agreement with the simulations stopped at ${T}_{\mathrm{stop}}=50\,{\rm{K}}$, which include PDR emission. Values lower than those observed are predicted for pure photoionized gas (${T}_{\mathrm{stop}}=1000\,{\rm{K}}$). An additional correlation of [C ii]158/[N ii]${}_{\mathrm{122,205}}$ with ionization potential (traced by [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$) might also be present, as [N ii] emission decreases in favor of [N iii].
  • 6.  
    The inclusion in the diagnostic diagrams of dwarf galaxies demonstrates that classical line ratios such as [O iv]${}_{25.9}$/[Ne ii]${}_{12.8}$ fail as AGN/starburst tracers at low metallicities. Instead, the [O iv]${}_{25.9}$/[O iii]88 ratio is an excellent tracer to discriminate between non-thermal excitation in AGN and thermal ionization produced by any kind of star formation activity. Since it probes a relatively hard range of the spectrum (54.94–$35.12\,\mathrm{eV}$, 4–$2.6\,\mathrm{Ry}$), this line ratio is ideal for composite objects with extreme star formation bursts and mergers involving a chemically unevolved companion, a scenario that seems to be more common with increasing redshift. Alternatively, the [O iv]${}_{25.9}$/([Ne iii]${}_{15.6}\,+\,$[Ne ii]${}_{12.8}$) ratio can also be used as a proxy for the relative AGN to starburst contribution.
  • 7.  
    A new AGN/star formation diagnostic diagram is proposed, based on the [O iv]${}_{25.9}$/[O iii]88 and [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ (or alternatively [S iv]${}_{10.5}$/[S iii]${}_{18.7}$). The [O iv]${}_{25.9}$/[O iii]88 ratio is sensitive to the relative contributions from AGN and star formation, while the [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ ratio ([S iv]${}_{10.5}$/[S iii]${}_{18.7}$) is sensitive to the ionization, mostly driven by the metallicity in the stellar population. The combination of these line ratios allows us to clearly separate the different populations in the diagram, i.e., AGNs, dwarf galaxies, and starburst galaxies, and provides an extinction-free diagnostic ideal for future mid- and far-IR spectroscopic surveys.
  • 8.  
    Photoionization by current stellar atmosphere models cannot reproduce the observed [O iv]${}_{25.9}$/[O iii]88 line ratios due to the lack of predicted photons above the He ii ionization edge ($\gtrsim 54\,\mathrm{eV}$, $4\,\mathrm{Ry}$). Furthermore, both dwarf and starburst galaxies show similar [O iv]${}_{25.9}$/[O iii]88 line ratios in the $5\times {10}^{-3}$$3\times {10}^{-1}$ range, with no apparent dependency on the metallicity of the stellar population. An improved treatment of stellar atmosphere models above $54\,\mathrm{eV}$ will be necessary in order to explain the observed line ratios, e.g., by including the contribution of shocks and non-thermal emission in the EUV.
  • 9.  
    LINERs show intermediate line ratios of [O iv]${}_{25.9}$/[O iii]88, [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$, and [S iv]${}_{10.5}$/ [S iii]${}_{18.7}$, between Seyfert and starburst galaxies. These line ratios can be explained by assuming a softer UV spectrum with a slope of $\alpha \approx -3.5$, in line with the receding accretion disk scenario expected at low-luminosities and/or low-accretion efficiencies.
  • 10.  
    The ([Ne iii]${}_{15.6}$+[Ne ii]${}_{12.8}$)/([S iv]${}_{10.5}$+[S iii]${}_{18.7}$) line ratio is sensitive to the gas metallicity for AGN, starburst, and dwarf galaxies, this is confirmed by the models including the effect of sulphur depletion. Thus, this ratio is proposed as a powerful extinction-free metallicity tracer, ideal for dust regions and obscured objects where optical determinations are not possible. Future facilities, such as JWST for the Local Universe and SPICA along Cosmic history, at the peak of star formation and black hole accretion activity ($1\lt z\lt 4$), will be able to trace the build up of heavy elements during galaxy evolution.

The authors thank the referee for valuable comments, which helped to improve the manuscript. The authors also thank M. Melendez for his suggestions on Cloudy models. PACS has been developed by a consortium of institutes led by MPE (Germany) and including UVIE (Austria); KU Leuven, CSL, IMEC (Belgium); CEA, LAM (France); MPIA (Germany); INAF-IFSI/OAA/OAP/OAT, LENS, SISSA (Italy); IAC (Spain). This development has been supported by the funding agencies BMVIT (Austria), ESA-PRODEX (Belgium), CEA/CNES (France), DLR (Germany), ASI/INAF (Italy), and CICYT/MCYT (Spain). This research made use of Astropy, a community-developed core Python package for Astronomy (Astropy Collaboration, 2013).

Facilities: Herschel (PACS and SPIRE) - European Space Agency's Herschel space observatory, Spitzer (IRS). -

Software: python, cloudy, pycloudy, astropy.

Note added in proof

The authors refer here to Figure 9 in Weaver et al. (2010) for a three-line diagnostic similar to our diagram in Figure 11 (using [Ne ii]${}_{12.8\mu {\rm{m}}}$, [Ne iii]${}_{15.6\mu {\rm{m}}}$, and [O iv]${}_{25.9\mu {\rm{m}}}$). The different types of galaxies are separated in both diagnostic diagrams. However, the [O iv]${}_{25.9}$/[O iii]${}_{88}$ line ratio is able to separate, by itself, the AGN-dominated ratios ([O iv]${}_{25.9}$/[O iii]${}_{88\gt 0.3}$) from those dominated by starbursts or dwarf galaxies.

APPENDIX: APERTURE EFFECTS

In this section, we test the possible aperture effects between Herschel/PACS and ISO/LWS line flux measurements reported in Brauher et al. (2008). To illustrate this, we show in Figure 13 the comparison of PACS versus LWS measurements for the [C ii]${}_{158\mu {\rm{m}}}$ line, which is the worst case scenario in terms of the contribution of the extended emission from the galaxy, as can be seen in the line maps of Figure Set 4. For a subsample of 45 AGN and starburst galaxies with both PACS and LWS [C ii]${}_{158\mu {\rm{m}}}$ measurements, we find a median value and a median absolute deviation of ${F}_{[{\rm{C}}\,{\rm{II}}]158}^{\mathrm{LWS}}/{F}_{[{\rm{C}}\,{\rm{II}}]158}^{\mathrm{PACS}}\,=0.95\pm 0.37$. Despite the smaller aperture involved, PACS fluxes are ∼5% higher than LWS fluxes, which suggests that most of the emission in our sample is compact within the 3 × 3 spaxel aperture ($28\buildrel{\prime\prime}\over{.} 2\times 28\buildrel{\prime\prime}\over{.} 2$).

Figure 13.

Figure 13. [C ii]${}_{158\mu {\rm{m}}}$ line flux for the subsample of AGN and starburst galaxies measured with both Herschel/PACS in this work and ISO/LWS in Brauher et al. (2008). The dotted line corresponds to the ratio ${F}_{[{\rm{C}}{\rm{ii}}]158}^{\mathrm{LWS}}/{F}_{[{\rm{C}}{\rm{ii}}]158}^{\mathrm{PACS}}=1.0$.

Standard image High-resolution image

On the other hand, in order to show the impact of using different apertures in line ratios involving Spitzer/IRS and Herschel/PACS lines, we reproduce in Figure 14 the same diagram shown in Figure 11, but using only central spaxel fluxes for the [O iii]${}_{88\mu {\rm{m}}}$ line. In Figure 11, the [O iv]${}_{25.9\mu {\rm{m}}}$ is extracted from a $11\buildrel{\prime\prime}\over{.} 1\times 22\buildrel{\prime\prime}\over{.} 3$ aperture, that is a factor of approximately three smaller when compared to the [O iii]${}_{88\mu {\rm{m}}}$ aperture ($28\buildrel{\prime\prime}\over{.} 2\times 28\buildrel{\prime\prime}\over{.} 2$), thus extended [O iii]${}_{88\mu {\rm{m}}}$ emission could lower the line [O iv]${}_{25.9}$/[O iii]88 line ratio. Figure 14 shows that, by using central spaxel fluxes in [O iii]${}_{88\mu {\rm{m}}}$, thus an aperture approximately three times smaller than in the [O iv]${}_{25.9\mu {\rm{m}}}$ line, the variations in the diagram with respect to Figure 11 are within the errors for the majority of the galaxies. Therefore, we conclude that aperture effects in this study are not relevant and do not affect our results.

Figure 14.

Figure 14. [Ne iii]${}_{15.6}$/[Ne ii]${}_{12.8}$ line ratio vs. the [O iv]${}_{25.9}$/[O iii]88 ratio (same notations as in Figure 11) using fluxes extracted from the Herschel/PACS central spaxel for the [O iii]${}_{88\mu {\rm{m}}}$ line. Symbols are color-coded according to their ${{\rm{F}}}_{[\mathrm{Ne}{\rm{V}}]14.3}/{{\rm{F}}}_{\mathrm{FIR}}$ flux ratio, when available (see color bar).

Standard image High-resolution image

Footnotes

  • hipe is a joint development by the Herschel Science Ground Segment Consortium, consisting of ESA, the NASA Herschel Science Center, and the HIFI, PACS, and SPIRE consortia.

  • 10 

    Dimensionless ionization parameter as defined in Ferland et al. (2013).

  • 11 

    The majority of dwarf galaxies in the sample show metallicities in the $0.001\lesssim Z\lesssim 0.008$ range (Madden et al. 2013).

  • 12 

    The linear regression was performed using the buckleyjames routine, available in the stsdas data analysis package: http://stsdas.stsci.edu/cgi-bin/gethelp.cgi?buckleyjames.hlp.

  • 13 

    Astronomy Survival Analysis Package, described in: http://stsdas.stsci.edu/cgi-bin/gethelp.cgi?survival.hlp.

  • 14 

    As discussed in the Appendix, aperture effects on ISO/LWS line fluxes are not expected to be important in our sample of galaxies.

  • 15 

    Collected from values published in the literature, see Table 3. Compton thick objects are labelled with "Y" in the CT column.

Please wait… references are loading.
10.3847/0067-0049/226/2/19