HerschelHerschel is an ESA space observatory with science instruments provided by European-led Principal Investigator consortia and with important participation from NASA./PACS spectroscopy of NGC 4418 and Arp 220: H{}_{2}O, H{}_{2}^{18}O, OH, {}^{18}OH, O i, HCN and NH{}_{3}

Herschelthanks: Herschel is an ESA space observatory with science instruments provided by European-led Principal Investigator consortia and with important participation from NASA./PACS spectroscopy of NGC 4418 and Arp 220:
HO, HO, Oh, Oh, O i, HCN and NH

E. González-Alfonso Universidad de Alcalá de Henares, Departamento de Física, Campus Universitario, E-28871 Alcalá de Henares, Madrid, Spain eduardo.gonzalez@uah.es    J. Fischer Naval Research Laboratory, Remote Sensing Division, 4555 Overlook Ave SW, Washington, DC 20375, USA    J. Graciá-Carpio Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    E. Sturm Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    S. Hailey-Dunsheath Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    D. Lutz Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    A. Poglitsch Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    A. Contursi Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    H. Feuchtgruber Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    S. Veilleux Department of Astronomy, University of Maryland, College Park, MD 20742, USA Astroparticle Physics Laboratory, NASA Goddard Space Flight Center, Code 661, Greenbelt, MD 20771 USA    H. W. W. Spoon Cornell University, Astronomy Department, Ithaca, NY 14853, USA    A. Verma University of Oxford, Oxford Astrophysics, Denys Wilkinson Building, Keble Road, Oxford, OX1 3RH, UK    N. Christopher University of Oxford, Oxford Astrophysics, Denys Wilkinson Building, Keble Road, Oxford, OX1 3RH, UK    R. Davies Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    A. Sternberg Sackler School of Physics & Astronomy, Tel Aviv University, Ramat Aviv 69978, Israel    R. Genzel Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany    L. Tacconi Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany
Key Words.:
Line: formation – Galaxies: ISM – ISM: jets and outflows – Infrared: galaxies – Submillimeter: galaxies

Full range Herschel/PACS spectroscopy of the (ultra)luminous infrared galaxies NGC 4418 and Arp 220, observed as part of the SHINING key programme, reveals high excitation in HO, OH, HCN, and NH. In NGC 4418, absorption lines were detected with 800 K (HO), 600 K (OH), 1075 K (HCN), and 600 K (NH), while in Arp 220 the excitation is somewhat lower. While outflow signatures in moderate excitation lines are seen in Arp 220 as have been seen in previous studies, in NGC 4418 the lines tracing its outer regions are redshifted relative to the nucleus, suggesting an inflow with 12 M yr. Both galaxies have compact and warm (100 K) nuclear continuum components, together with a more extended and colder component that is much more prominent and massive in Arp 220. A chemical dichotomy is found in both sources: on the one hand, the nuclear regions have high HO abundances, , and high HCN/HO and HCN/NH column density ratios of 0.10.4 and 25, respectively, indicating a chemistry typical of evolved hot cores where grain mantle evaporation has occurred. On the other hand, the high OH abundance, with OH/HO ratios of 0.5, indicates the effects of X-rays and/or cosmic rays. The nuclear media have high surface brightnesses ( L/kpc) and are estimated to be very thick ( cm). While NGC 4418 shows weak absorption in HO and OH, with a O-to-O ratio of 250500, the relatively strong absorption of the rare isotopologues in Arp 220 indicates O enhancement, with O-to-O of 70130. Further away from the nuclear regions, the HO abundance decreases to and the OH/HO ratio is reversed relative to the nuclear region to 2.510. Despite the different scales and morphologies of NGC 4418, Arp 220, and Mrk 231, preliminary evidence is found for an evolutionary sequence from infall, hot-core like chemistry, and solar oxygen isotope ratio to high velocity outflow, disruption of the hot core chemistry and cumulative high mass stellar processing of O.

1 Introduction

From the first CO observations of IRAS galaxies (Young et al., 1984), much effort has been devoted to understanding spectroscopic observations of molecules in (ultra)luminous infrared galaxies (hereafter, (U)LIRGs), and how they are related to the conditions traced by the continuum and line observations of atoms and ions. The launch of the Infrared Space Observatory (ISO) and, more recently, the Herschel Space Observatory, have opened a new window to study extragalactic sources in the far-infrared (hereafter, far-IR) domain, where the bulk of the (U)LIRG luminosity is emitted. The relevance of far-IR molecular spectroscopy of (U)LIRGs is well illustrated with the initial ISO result of Fischer et al. (1999), that the fine-structure lines of atoms and ions decrease in strength, as absorption in molecular lines, specifically HO and OH, increase in depth.

Water (HO) and hydroxyl (OH) are indeed key molecular species that trace relevant physical and chemical properties of the interstellar medium in luminous infrared extragalactic sources. Powerful OH mega-masers are common in ULIRGs, and HO masers are found in AGN accretion disks and in post-shocked gas associated with nuclear jets (see Lo, 2005, for a review). With many transitions lying at far-IR wavelengths, where the bulk of the luminosity is emitted, HO and OH are mainly excited through absorption of far-IR photons (González-Alfonso et al., 2004, hereafter G-A04), and thus their excitation helps probe the general properties of the underlying far-IR continuum sources (González-Alfonso et al., 2008). Chemically, HO is expected to trace an undepleted chemistry where grain mantles are evaporated; both HO and OH are also boosted in shocks and, especially OH, in X-ray and Cosmic Ray Dominated Regions (XDRs and CRDRs). Furthermore, OH has turned out to be a unique tracer of massive molecular outflows (Fischer et al., 2010; Sturm et al., 2011). Recently, HO has been detected in high- sources (Impellizzeri et al., 2008; Omont et al., 2011; Lis et al., 2011; van der Werf et al., 2011; Bradford et al., 2011), and thus understanding its role in the local Universe will be crucial for future routine observations of HO with ALMA in the far Universe. HCN is another key molecule widely studied in local (U)LIRGs at (sub)millimeter wavelengths (e.g., Aalto et al., 1995). NH, first detected in extragalactic sources by Martin & Ho (1979), is another important tracer; in Arp 220, it was detected at far-IR wavelengths with ISO (G-A04) and at centimeter wavelengths by Takano et al. (2005) and recently by Ott et al. (2011). With its high sensitivity and spectral resolution at far-IR wavelengths, the Herschel/PACS instrument (Pilbratt et al., 2010; Poglitsch et al., 2010) provides new opportunities for using the far-IR transitions of these molecules to probe the kinematic, chemical, and radiative activity in IR luminous galaxies.

We report in this work Herschel/PACS spectroscopic observations of HO, OH, their O isotopologues, NH, and the surprising detection of highly excited HCN, in NGC 4418 and Arp 220. NGC 4418 is a peculiar, single nucleus galaxy, with a moderate luminosity ( L) but with other properties similar to warm ULIRGs: a high ratio and an extreme [C ii] deficit (e.g. Graciá-Carpio et al., 2011), a high continuum surface brightness, an extremely compact luminosity source (Evans et al., 2003), and warm infrared colours (, ). We adopt a distance to the source of 29 Mpc. Arp 220 is the prototypical and nearest ULIRG, showing a double nucleus and a luminosity of L. We adopt a distance to the source of 72 Mpc. The observations are described in §2; models are shown in §3; model results are discussed in §4, and our main conclusions are summarized in §5.

2 Observations and results

2.1 Observations

The full (, m), high resolution, PACS spectra of NGC 4418 and Arp 220, taken as part of the guaranteed-time key programme SHINING, were observed on July 27th and February 27th (2010), respectively. The spectra were taken in high spectral sampling density mode using first and second orders of the grating. The velocity resolution of PACS in first order ranges from km s at 105 m to km s at 190 m, and in second order from km s at 52 m to km s at 98 microns. For NGC 4418, we also present a scan around the [O i] 63 m line taken in third order with velocity resolution of km s. The data reduction was mostly done using the standard PACS reduction and calibration pipeline (ipipe) included in HIPE 5.0 975. The reduction of Arp 220 required, however, two additional steps in order to correct for pointing errors that resulted in significant variations of the continuum level of the central spaxel111PACS has spatial pixels (spaxels), each with resolution and covering a field of (Poglitsch et al., 2010). as well as in wavelength offsets generated by pointing errors along the dispersion direction of the slit (Poglitsch et al., 2010). These effects were produced by the movement of the source between the central spaxel and a neighbouring spaxel. In order to correct for them, the emission from the two spaxels with highest signal were added together in an intermediate step in the pipeline, which significantly improved the continuum baseline of the combined spectrum but increased the noise by a factor222 Since the neighbouring spaxel has a signal much lower than the central one, the subsequent final calibration leaves the factor nearly unchanged. ; the time evolution of the pointing shifts was reconstructed from the star tracker frames, and wavelength calibration was corrected according to Fig. 8 by Poglitsch et al. (2010). Pointing errors across the slit were found from 0 to , resulting in velocity shifts of km s. We estimate that the associated uncertainties are below 30 km s, though they may be higher for some individual lines.

Figure 1: Spectral energy distribution of NGC 4418 and Arp 220 from mid-IR to millimeter wavelengths. Both the Herschel/PACS and the Spitzer/IRS spectra are shown. Data points at (sub)millimeter wavelengths for NGC 4418 are from (Yang & Phillips, 2007, 350 m), (Roche & Chandler, 1993, 450, 800, and 1100 m), (Dunne et al., 2000; Lisenfeld et al., 2000, 850 m), and (Albrecht et al., 2007, 1300 m). For Arp 220, they are taken from (Eales et al., 1989, 450 m); (Rigopoulou et al., 1996, 350, 800, and 1100 m); and (Sakamoto et al., 1999, 2008, 1300 m). Models discussed in §3 are included. For NGC 4418, the yellow, blue, light-blue, and solid-green curves show the models for the hot, core, warm, and extended components, respectively, with parameters given in Table 1. An additional cold component, shown with the dashed-green curve, is included to better fit the SED at long wavelengths. For Arp 220, the yellow, blue, light-blue, and solid-green curves show the models for the hot, western nucleus, eastern nucleus, and extended components, respectively (Table 1).
Figure 2: Energy level diagrams of OH and HO (ortho and para). Red arrows indicate lines detected in both sources; blue arrows denote lines detected only in NGC 4418, green-dashed arrows mark transitions contaminated (blended) by lines of other species, and blue-dashed arrows denote marginal detections in NGC 4418. The small numbers in the OH diagram indicate rounded wavelengths in m.

For the final calibration, the spectrum was normalized to the telescope flux and recalibrated with a reference telescope spectrum obtained from observations of Neptune. The flux density of the spectrum in the central spaxel(s) was then rescaled to the value of the total continuum of the source obtained by the combination of the 25 spaxels of the PACS spectrometer. This method assumes that the distribution of the IR continuum and the lines are similar for the spatial resolution of the observations. Indeed the line to continuum ratio in the central and the combined 25 spaxels was found to be the same for all strong lines except for the ground-state OH 119 and 79 m in Arp 220, which display 9% and 3% deeper absorption in the central spaxels. The latter OH lines thus appear to show low-level reemission (probably through scattering) over kpc spatial scales. In general, the comparison indicates that both sources are at far-IR wavelengths point-sources compared with the PACS spatial resolution ( kpc for NGC 4418 and kpc for Arp 220).

The continua of both sources are shown in Fig 1, including the Spitzer IRS spectra (Spoon et al., 2007); the models are discussed in §3. Both galaxies show strong silicate absorption at 9.7 m, with PAH emission that is weak in Arp 220 and only weakly detected at 11.2 m in NGC 4418 (Spoon et al., 2007). Besides the obvious differences in luminosity, a striking contrast between both sources is the ratio, which is in NGC 4418 and approaches in Arp 220. The higher value in Arp 220 indicates that larger amounts of dust surround the nuclei in an extended region (ER in G-A04, denoted here as ), both extinguishing the nuclear emission at 25 m and contributing to the far-IR continuum. This emission can be interpreted as heating of dust by the radiation emanating from the nuclei (Soifer et al., 1999, G-A04). In NGC 4418, the warmer spectral energy distribution (SED) suggests that the amount of dust in  is lower, but still probably significant (see §3). On the other hand, the (sub)millimeter emission in Arp 220 is dominated by the nuclei as revealed by interferometric (sub)millimeter observations (Downes & Eckart, 2007; Sakamoto et al., 2008), indicating extreme opacities in, and high luminosities from the nuclei. No interferometric observations are available for NGC 4418.

Spectroscopic data used for both line identification and radiative transfer modeling were taken from the JPL (Pickett et al., 1998) and CDMS (Müller et al., 2001, 2005) catalogues. Figure 2 shows the energy level diagrams of OH and HO. The spectra around the HO lines are shown in Fig. 3 (blue and green bands, m) and Fig. 4 (red band, m), while high-lying HO lines tentatively detected in NGC 4418 (blue-dashed arrows in Fig 2) are shown in Fig. 5. The OH line profiles are presented in Fig. 6. Figs. 7 and 8 show excerpts of the PACS spectra around the HO and OH lines, respectively. The HCN and NH lines are displayed in Figs. 9 and 10, respectively. Figure 11 compares the line profiles of the [O i] 63 and 145 m transitions with those of some OH lines. Figure 12 compares the line shapes of two HCO transitions observed with high spatial resolution by Sakamoto et al. (2009) in Arp 220, with selected lines observed with Herschel/PACS. Results of the continuum and line modeling are summarized in Tables 1 and 2, and the line equivalent widths of the HO, [O i] and OH, HCN, NH, HO, and OH lines are listed in Tables 3 to 8.

Figure 3: HO lines in the blue and green bands ( m) detected in NGC 4418 (upper spectra in each panel) and Arp 220 (lower spectra). The spectra, histograms with black solid lines, have been scaled by a factor of in panels k and p. Vertical dotted lines indicate the rest wavelengths of the lines relative to and for NGC 4418 and Arp 220, respectively. Model results are also shown. For NGC 4418, the blue and light-blue curves show the models for the  and  components, respectively; for Arp 220, the blue and light-blue lines show the estimated contributions from  and , and the green curve (in panel k) shows the contribution by . Red is total. Model parameters are given in Tables 1 and 2.
Figure 4: HO lines in the red band ( m) detected in NGC 4418 (upper spectra in each panel) and Arp 220 (lower spectra). The spectra have been scaled by a factor of in panels a,k and l, and by 3 in panel c. Model results are also shown. Color codes for the contributions of the different components are given in Fig. 3, with the yellow curve in panel q showing the additional contribution by the halo component () in front of the nuclear region of Arp 220.

2.2 Systemic redshifts

Line parameters were derived by subtracting baselines and fitting Gaussian curves to all spectral features. One relevant question in our study is the zero-velocity adopted for the high-excitation molecular gas in the nuclear region of the galaxies. In NGC 4418, the excited HO lines do not show any evidence of systematic shifts indicative of outflowing or inflowing gas. Fig 13 indicates that the only HO line clearly redshifted relative to all others is the ground-state at 179.5 m, which is expected to be contaminated by CH (2-1) (see below). From all other HO lines, the average redshift is ( km s), where the error denotes the standard deviation of the mean. The inferred redshift is fully consistent with observations of CO, CN, HNC, and other molecular species observed at millimeter wavelengths (e.g. Aalto et al., 2007a). The Sloan Digital Sky Survey (SDSS) spectrum of NGC 4418 also gives a consistent redshift for the optical absorption/emission lines of and , respectively (D. Rupke, private communication). Finally, this is also consistent with the redshift measured for the high-lying OH lines (see below), which also trace the nuclear component, and with that of the [C ii] emission line at m, , which we adopt as the reference value. In Arp 220, Sakamoto et al. (2009) detected P-Cygni profiles in excited submillimeter lines of HCO toward both nuclei, indicating dense outflowing gas in the nuclei (see also Fig. 12 and §2.7). The transition from absorption to emission occurs at a velocity of km s toward both nuclei, which we adopt as the systemic redshift (). The foreground absorption traced by CO (3-2) toward the western nucleus lies at a lower velocity, km s. The line velocity shifts () relative to the above systemic redshifts, the line widths, and the equivalent widths () are plotted in Fig. 13 for the HO lines and in Fig. 14 for the OH lines.

2.3 HO and OH

Based on the HO spectra and the molecular line positions shown in Figs. 3 and 4, most HO lines are free of contamination from lines of other species, with the following exceptions: as mentioned above, the H line in Fig. 4q at 179.5 m is blended with CH; the  line in Fig. 3f is blended with highly excited NH (see Fig. 10a); the  (Fig. 3i) and the  lines (Fig. 4q) may have some contribution from CH; finally, the spectral feature at 104 m associated in Fig. 4a with the  line is probably contaminated by OH. Other contaminated lines marked with dashed-green arrows in Fig. 2 are not shown. The OH lines in Fig. 6 have relatively weak contamination by NH in the red component of the 84 m doublet (panel e, see Fig. 10c), and by HO in the blue component of the 65 m doublet (panel f).

The detections summarized in Fig. 2 indicate extreme excitation in the nuclei of both galaxies. In Arp 220, lines with lower level energy up to K in both HO ( and , Fig. 3j, the latter contaminated by excited NH) and OH (, Fig. 6i) are detected. NGC 4418 shows even higher-lying lines, with detected HO lines as high as  and  (Fig. 3f,g; K). These are the highest-lying HO lines detected in an extragalactic source to date.

The relative strengths of the HO and OH lines in NGC 4418 and Arp 220 show a striking dependence on . The lowest-lying ( K) HO lines, especially the  transition at m (Fig. 4q), are stronger in Arp 220 than in NGC 4418, but higher-lying lines are stronger in NGC 4418 (see Fig. 13). The effect is even more clearly seen in OH (Fig. 14), where the ground-state lines (upper row in Fig. 6) are much stronger in Arp 220, the intermediate transitions at 84 m (Fig. 6e) have similar strengths, and higher excited OH lines (Fig. 6f,g,h,i) are stronger in NGC 4418. This is partially a consequence of the  component in Arp 220: since are calculated relative to the observed (nuclear+) continuum, the high-lying lines arising from the nuclei are reduced due to the contribution by  to the continuum, which is more prominent in Arp 220. In other words, the high-lying lines are more continuum-diluted in Arp 220 than in the relatively “naked” NGC 4418, which is consistent with the continuum models shown in §3. On the other hand,  in Arp 220 produces strong absorption in the low-lying lines, thus increasing their relative to that in NGC 4418.

Nevertheless, dilution in the continuum cannot explain the higher excitation indicated by the very high-lying lines of HO and OH in NGC 4418. This is best seen in Fig. 13gh, where the line fluxes are normalized relative to that of the excited  HO line at 66.4 m. The latter line (Fig. 3g) is well detected in both sources and, with K, is not expected to be significantly contaminated by absorption of extended low-excitation HO in Arp 220. Lines with above 500 K are shown to produce, relative to , more absorption in NGC 4418 than in Arp 220. Also, as mentioned above, the H and  transitions (Fig. 3f,g) are only detected in NGC 4418 (the  line in Fig. 3f is contaminated by NH, and the feature is not detected in Arp 220). Two more para-HO transitions detected only in NGC 4418 are the  and  lines (Fig. 3a,r); the lower levels of these lines are non-backbone and thus require high columns to be significantly populated. Other HO lines detected only in NGC 4418 are the  and  at m, but the Arp 220 spectrum at this wavelength shows a very broad feature, possibly due to pointing or instrumental effects (Fig. 3q).

Figure 5: Marginally detected HO lines in NGC 4418 (upper spectra in each panel); the corresponding lines from Arp 220 are also shown (lower spectra). Model predictions for these lines (for  in NGC 4418 and for  in Arp 220) are also shown (red curves).
Figure 6: OH lines in NGC 4418 (upper spectra) and Arp 220 (lower spectra). doubling is resolved in most transitions, except for the two narrowly separated 71 m components that are blended into a single spectral feature. Model results are also shown. For NGC 4418, the blue, light-blue, and green curves show the models for the , , and  components, respectively; for Arp 220, the blue, light-blue, green, and yellow curves show the models for , , , and , respectively. The gray curves in panel f show the expected contribution of the H line to the spectra around m . The red curves show the total from all components.

The spectrum of NGC 4418 clearly shows two intrinsically weak HO lines with , the  line at m and the  line at m (Fig. 3e,h). All other detected HO lines have higher transition probabilities and, with the exception of the very high-lying lines, are expected to be strongly saturated. While the ortho-H m line is also detected in Arp 220 within a wing-like feature that probably has some contribution from HO , the para-H m line is not detected in Arp 220. The high columns in NGC 4418 are also confirmed by comparing the absorption in the ortho/para pair  and  (Fig. 3q,r), which have similar lower level energies ( K), wavelengths, and transition probabilities ( s). For an ortho-to-para ratio of 3, one would expect the same ratio of 3 for in the optically thin regime, but this is as low as indicating that both lines are still well saturated.

Even higher-excitation lines of HO, with K, may be present in the spectrum of NGC 4418 (blue-dashed arrows in Fig 2). Among these features, shown in Fig 5, the most likely detection is the  line in Fig 5c, all other features being marginal. In particular, the m feature in (f) has a non-Gaussian shape. Nevertheless, all features together appear to indicate more extreme excitation in the nucleus of NGC 4418. This is plausible, as HCN rotational lines with K are detected in NGC 4418 (see below).

As discussed earlier, dilution in the continuum also lowers the of the high-lying OH lines in Arp 220 relative to the values in NGC 4418; however, the most excited OH transitions, the and lines at 56 and 53 m, respectively, are intrinsically stronger in NGC 4418 according to our decomposition of the continuum. Another effect of  is also worth noting: the only OH transition observed in emission above the continuum, the doublet at 163 m, is excited through absorption of continuum photons in the 35 and 53.3 m doublets followed by a cascade down to the ground OH state (see App. II by Genzel et al., 1985), and is therefore expected to arise in the  component (G-A04). The strong OH 163 m emission in Arp 220 reflects the massive envelope around its nuclei. In NGC 4418, the doublet is weaker but still in emission, also hinting at the existence of a  component around its nucleus.

2.4 HO and Oh

The spectra around relevant HO and OH transitions in NGC 4418 and Arp 220 are displayed in Figs. 7 and 8. Close spectral features due to HO, NH, C, and CH are also indicated. HO is detected in both sources, though in relatively low-lying lines. The highest-lying detected line is the  transition with K in Fig. 7a, as detection of the  transition in panel e is rather marginal. The most striking feature in Fig. 7 is the relative amount of absorption observed in NGC 4418 and Arp 220. While the absorption in the main isotopologues is stronger in NGC 4418 than in Arp 220 (Figs. 3 and 4), the opposite behavior is generally found for the absorption in the rare isotopologues. This is best seen in the , ,  (showing redshifted emission in Arp 220), , and  lines. There are, however, two spectral features nearly coincident with the  and  HO lines at and m (panels d and h, respectively), with absorption in NGC 4418 apparently stronger than in Arp 220, but the latter is subject to an uncertain baseline.

Figure 7: Spectra around the wavelengths of relevant HO lines in NGC 4418 (upper profiles) and Arp 220 (lower profiles). Model predictions for NGC 4418 () with , and for Arp 220 () with are shown in red.
Figure 8: Spectra around the wavelengths of relevant OH lines in NGC 4418 (upper profiles) and Arp 220 (lower profiles). Model predictions for NGC 4418 () with , and for Arp 220 with are also shown. For the OH 120 m doublet, the contributions by the different components in Arp 220 (see Fig. 6) are indicated; red denotes the total of all components.

The OH lines in Fig. 8, specifically the OH 85 and 120 m doublets, confirm the O enhancement in Arp 220. The blueshifted component of the OH 120 m doublet is strongly contaminated by CH in both sources, but the uncontaminated redshifted component, undetected in NGC 4418, is strong in Arp 220. It is also worth noting that the two OH 85 m components show rather different absorption depths in Arp 220. This asymmetry is not expected to arise from radiative pumping or opacity effects, because the components of any transition have the same radiative transition probabilities, and the corresponding lower levels tend to be equally populated. Indeed, the absorption in the corresponding components of the main isotopologue at 84 m are very similar (Fig. 6e). The OH 85 m asymmetry in Arp 220 is further discussed in §3.2.3. The OH component at 65.54 m in Fig. 8a is partially blended with two lines of NH and, together with the expected contribution by HO at 65.61 m, form a broad feature with uncertain baseline. The red component at 65.69 m appears to be stronger in Arp 220 than in NGC 4418.

2.5 Hcn

HCN is a key molecule widely studied in both Galactic and extragalactic sources. In NGC 4418 and Arp 220, the low-lying lines at millimeter wavelengths have been observed by Aalto et al. (2007a, b) and Wiedner et al. (2004). The vibration-rotation band at 14 m was detected in both sources by Lahuis et al. (2007), and rotational emission from the upper vibrational state in NGC 4418 has been detected by Sakamoto et al. (2010). The l-type HCN lines from at centimeter wavelengths were detected in absorption toward Arp 220 (Salter et al., 2008). Among the 15 extragalactic sources detected in the HCN band, the Lahuis et al. (2007) analysis indicated that NGC 4418 has the second highest HCN column. Recently, Rangwala et al. (2011) have reported the detection of several emission/absorption HCN lines with Herschel/SPIRE in Arp 220.

Figure 9: Spectra around the frequencies of HCN lines in NGC 4418 (upper profiles in each panel) and Arp 220 (lower profiles). Model predictions for NGC 4418 () and Arp 220 () are included.
Figure 10: Spectra around the frequencies of relevant NH lines in NGC 4418 (upper profiles in each panel) and Arp 220 (lower profiles). Model predictions for NGC 4418 () and Arp 220 () are included.

We report in Fig. 9 the PACS detection of HCN absorption in pure rotational transitions at far-IR wavelengths in both NGC 4418 and Arp 220. In NGC 4418, all transitions from ( K) to ( K) are clearly identified. Furthermore, there are hints of absorption up to the transition ( K). In Arp 220, the line is clearly contaminated by HO and NH, and the one by C and possibly also by NH. The line is weak, suggesting that this transition is tracing the tail of the Spectral Line Energy Distribution (SLED). However, a clear feature is found at the wavelength of the line, and the line could show some marginal absorption as well. Nevertheless, we consider the identification of the line as questionable given the weakness of the nearby rotational HCN lines and the possible contamination by other species.

Fluxes derived from Gaussian fits are shown in Fig. 15 as a function of , where values derived from Herschel/Spire observations by Rangwala et al. (2011) are included for Arp 220. The HCN excitation in NGC 4418 is extreme, with the SLED apparently peaking at ( K). The striking characteristic of the far-IR HCN lines is that they are detected in absorption against the far-IR continuum, indicating that the dust temperature and far-IR continuum opacities behind the observed HCN are high; otherwise the lines would be detected in emission. We argue in §3 that the HO, OH, and HCN lines in NGC 4418 are tracing a high luminosity, compact nuclear source, denoted as the nuclear core (). In Arp 220, the HCN SLED as seen by Herschel/Spire (Rangwala et al., 2011) peaks at ( K) indicating, like HO, more moderate excitation, but a second, higher-excitation component could be present if the m spectral feature is due to HCN . As shown in §3.2.4 and 3.4.1, high HCN columns are required to explain the data.

Figure 11: Line profiles of the [O i] 63 and 145 m (a & b) transitions compared with those of some OH lines (c-f). Upper histograms, NGC 4418; lower histograms, Arp 220. Model predictions are also included. For NGC 4418, the blue curves show the combined contribution of the  and  components, and the green curves show the model for . For Arp 220, see caption of Fig. 6. Red denotes the total of all components.
Figure 12: Comparison between the line shapes of the HCO (green histograms) and (blue histograms) transitions toward (a) the western and (b) eastern nucleus of Arp 220 (from Sakamoto et al., 2009), with selected lines detected with Herschel/PACS (c-i). The dashed vertical line indicates our adopted systemic velocity (§2.2). The velocity scale for the OH 79 (163) m doublet is relative to the red (blue) component, and the systemic velocity for the other component is also indicated. The velocity of the HO  line in (c) is also indicated.

2.6 Nh

Ammonia is another N-bearing species widely observed in Galactic sources mostly through the pure-inversion transitions from metastable () levels at centimeter wavelengths (see Ho & Townes, 1983, for a review). The rotation-inversion transitions lie at far-IR wavelengths, and many of them have been detected with ISO toward Sgr B2 (Ceccarelli et al., 2002; Polehampton et al., 2007) indicating a hot, low dense molecular layer interpreted in terms of shock conditions (Ceccarelli et al., 2002). Detections of far-IR NH lines in extragalactic sources were presented, together with a preliminary analysis, by G-A04 in Arp 220. In Fig. 10 we show the Herschel/PACS detection of high-excitation, far-IR NH lines in both NGC 4418 and Arp 220.

NH has two species, ortho-NH () and para-NH (), with an expected ortho-to-para abundance ratio of unity (Umemoto et al., 1999). All rotational levels, except those of the ladder, are split into 2 inversion doubling sublevels ( for asymmetric and for symmetric), which for a given have decreasing level energy with increasing . Radiative transitions are only allowed within ladders, thus only collisions can populate the ladders and the absorption from different ladders is then sensitive to the gas temperature. Within a given ladder, both collisions and absorption of far-IR photons can pump the non-metastable levels. Since the () transitions of different ladders have similar wavelengths, the lines are crowded in wavelength and overlapping in some cases, with severe blending between the and ladders. The PACS domain covers the range from the ( m, Fig. 10g) and the ( m, Fig. 10f) lines, up to at m.

Figure 10 also indicates (partial) blending with lines of other species, as HO (panels a and b), CH (panel b), and HO (panel f). Most lines are strongly blended with the OH 84 m doublet. From the comparison with the modeling described in §3, the and especially the lines are so blended that a pseudo-continuum is expected to be formed, with the consequent uncertainty in the baseline subtraction. As also found for HO, OH, and HCN, NGC 4418 shows in NH clear indications of higher excitation than Arp 220, and the absorption in non-metastable levels relative to metastable ones is also stronger in NGC 4418 (e.g., the group in panel d). The ortho-NH is probably detected in NGC 4418, as the blended para-H line (see also Fig. 3f) is probably less strong than the close ortho-H line, and also because the transition is detected. There are also some hints of absorption in the line. In summary, NH lines are detected up to in NGC 4418 ( K), and up to in Arp 220 ( K).

2.7 Kinematics

In Fig. 12, we compare the HCO (green histograms) and (blue histograms) line profiles toward the western (a) and eastern (b) nucleus of Arp 220 (from Sakamoto et al., 2009), with the line profiles of selected lines detected with Herschel/PACS (c-i). The HCO lines toward both nuclei, observed with high angular () and spectral (30 km s) resolution, exhibit P-Cygni profiles indicative of outflowing gas (Sakamoto et al., 2009). Specifically, the redshifted spectral features observed in emission have a large velocity extent of km s from the systemic velocity. This prominent redshifted HCO emission has its counterpart in several lines detected with Herschel/PACS: the H line at m (Fig. 12c), the [O i] 63 m transition (d), and the OH 79 m doublet (e). The latter also shows deeper absorption in the blue component than in the red one, suggesting the occurrence of redshifted emission in the blue component as well. The line emission features are expected to be formed in gas located at the back/lateral sides (i.e. not in front of the nuclei where absorption of the continuum dominates the profile), and will therefore be redshifted, as observed, if the gas is outflowing. The velocity of the emission features ( km s) is significantly higher than the redshifted velocities measured for the rotating disks around both nuclei (up to km s, Sakamoto et al., 2008), indicating that little of this emission is associated with the rotation motions. Thus we conclude that moderate excitation lines of HO, OH, and O i trace the outflow at redshifted velocities detected in HCO. The asymmetric shape of the OH 163 m emission doublet, which peaks at redshifted velocities, is consistent with this scenario, as is the detection of redshifted emission features in the Herschel/SPIRE spectrum of Arp 220 in several species including H(Rangwala et al., 2011).

The blueshifted velocity extent of the OH 163 and [O i] 145 m lines, observed in emission, reaches km s from the systemic velocity (Fig. 12h-i), and could also trace outflowing gas at the most extreme velocities. The full velocity extent in these lines ( km s) is similar to that of CO (3-2) (Sakamoto et al., 2009). The full velocity extent of the high-lying HO lines is significantly lower, km s. This is specifically illustrated in Fig. 12f-g for the  and  lines, which lie in the range m where PACS has relatively high velocity resolution ( km s). In contrast with the OH 163 and the [O i] 145 m lines observed in emission, which probably trace spatially extended gas, the formation of the high-lying absorption HO lines is restricted to regions with high far-IR radiation densities that are optically thick at far-IR wavelengths, obscuring the emitting gas behind the nuclei. The absorption in these HO lines peaks at around central velocities, indicating that the lines mainly trace gas rotating on the surface of the nuclei. However, due to uncertainties in the velocity correction due to pointing shifts (see §2), and the similarity in the blueshifted velocity extent of the HCO and HO lines, we cannot rule out a significant contribution to the absorption by outflowing gas with velocities up to km s. Nor can we rule out spatially extended gas inflow, similar to what is inferred and discussed below for NGC 4418, since the peak absorption in the OH 79 m doublet is redshifted by km s.

In NGC 4418, the adopted redshift () is derived from the excited HO and OH lines, coinciding within km s with both the centroid of the [C ii] 158 m line and with the redshift inferred from the stellar absorption at optical wavelengths (§2.2). No indication of outflowing gas in the nuclear region is found. The excited HO and OH lines in NGC 4418 are indeed significantly narrower than in Arp 220 (Figs. 13 and 14), further indicating the relatively quiescent state of the NGC 4418 nucleus. On the other hand, the line shapes of the lowest-lying HO, OH, and [O i] lines show an intringuing behavior, opposite to the outflow signatures in Arp 220: the absorption features in the ground OH and [O i] 63 m lines are systematically redshifted by km s, and the [O i] 63 m line shows a blueshifted spectral component in emission, peaking at km s, i.e. an inverse P-Cygni profile (Fig. 11a). The lowest-lying HO line at 179.5 m is also redshifted, but contamination by CH at m makes the case more uncertain. Additional clues about this component come from the line shapes of the OH 84 and 163 m doublets: the blue component of the 84 m doublet ( K) shows an asymmetric shape with some redshifted excess (Fig. 6e and 11e; the red component is probably contaminated by excited NH, see Fig. 10c), suggesting that some redshifted gas at km s is significantly excited. The kinematics implied by the OH 163 m emission doublet are unclear due to the low signal-to-noise ratio (SNR), as one of the components is apparently blueshifted, whereas the other one peaks at central velocities (Figs. 6j, 11f, and 14). In summary, the ground-state lines of OH and [O i], together with the OH 84 and 163 m doublets, indicate the presence of a component different from the nuclear one in both excitation and kinematics, most probably extended in comparison with the nucleus but much less massive than the  component of Arp 220, and the velocity shifts suggest that this component may be inflowing onto the quiescent, nuclear region. We further investigate this scenario in §3.6.

Figure 13: Velocity shifts (a-b), line widths (c-d), and equivalent widths (e-f) of the HO lines in NGC 4418 (left) and Arp 220 (right). Panels g-h show the line fluxes normalized to that of the  transition at 66.4 m (see §2.3 for details). Error bars are 1- uncertainties from Gaussian fits to the lines.
Figure 14: Velocity shifts, line widths, and equivalent widths of the [O i] and OH lines in NGC 4418 (black symbols) and Arp 220 (blue symbols). Error bars are 1- uncertainties from Gaussian fits to the lines.
Figure 15: HCN line fluxes in NGC 4418 (a) and Arp 220 (b). In Arp 220, the fluxes of lines with up to 570 K have been taken from SPIRE-FTS data (Rangwala et al., 2011). The curves and triangles show model predictions for the  component of NGC 4418 and the  component of Arp 220. The dotted curve in (b) shows results obtained when only collisional excitation is included in the model. The solid curve in (a) shows the best fit model obtained with , while the dashed and dotted curves correspond to with and without gas-dust mixing, respectively.

3 Models

3.1 Overview

As the data presented in previous sections have shown, the far-IR spectra of NGC 4418 and Arp 220 are dominated by molecular absorption, with emission in only some lines and profiles. These rich line spectra and their associated dust emission cannot be described by a single set of ISM parameters, but different lines have different excitation requirements and are thus formed in different regions of the galaxies. Our approach for both galaxies is to fit these different regions and conditions, even though they are not spatially resolved with Herschel, with the smallest possible number of parameterized components.

Figure 16: Schematic representation of the modeled sources, showing the approximate spatial scales of the different far-IR components (see Fig. 1 and Tables 1 and 2). The  components in both galaxies (that account for the mid-IR spectra) are most probably associated with the nuclear regions and are not included here. The  component in Arp 220 has no associated continuum, and is responsible for the absorption of the nuclear continuum ( and ) in the ground-state lines of HO, OH, and O i. The plot is an oversimplification of the actual models, where the different components are modeled separately to account approximately for non-spherical symmetry.

For NGC 4418, we find that we need (see Fig. 1a) a hot component (yellow curve, hereafter ) that accounts for the mid-IR continuum; a nuclear core (blue curve, ) that provides absorption in the high-lying lines of HO, OH, HCN, and NH; a warm component (light-blue, ), that provides absorption in moderately excited lines of HO and OH as well as a significant fraction of the far-IR continuum emission; and an extended component (solid green, ), which accounts for the low-lying redshifted lines of OH and O i.

For Arp 220, we need (Fig. 1b) a single333Even if the mid-IR arises from the two nuclei, the continuum is simulated with just a single component with an effective diameter as given in Table 1. hot component (yellow curve, ) that accounts for the mid-IR continuum emission; the western and eastern nuclear components ( and  in blue and light blue, respectively), where the high-lying molecular lines are formed; the extended component (, green), which provides a significant fraction of the far-IR continuum emission and which is likely associated with moderate and low-excitation lines of HO and OH; and an additional absorbing “halo” component (), with no associated intrinsic continuum but located in front of the nuclei, is required to fit the absorption in the ground-state lines of HO, OH, and O i.

We use single dust temperatures for every component listed above, except for the  component in both galaxies where the dust temperature profile is calculated from the balance of heating (by the inner components) and cooling (González-Alfonso & Cernicharo, 1999). In our models, each single dust temperature component is attenuated by a foreground, screen-like shell, which is parameterized by its dust opacity at 25 m, . These screen-like shells are responsible for the silicate absorption features imprinted on the various modeled components in Fig. 1. The dust temperature profile in  is calculated by assuming spherical symmetry and a single nuclear heating component located at the center of the modeled source.

Parameters of the continuum models, and the inferred molecular parameters, are listed in Tables 1 and 2, respectively. A sketch of the modeled sources, showing the approximate spatial scales of the different far-IR components (excluding the  component of both galaxies), is shown in Fig. 16. Each component, however, can be interpreted in terms of a single source, as implicitly assumed, or alternatively applied to each one of an ensemble of smaller clouds of radius that do not spatially overlap along the line of sight. The scaling between these two approaches is discussed in G-A04.

The models for Arp 220 are discussed in sections 3.2 (high-lying lines) and 3.3 (low-lying lines), while those for NGC 4418 are developed in sections 3.4 (high-lying lines), 3.5 (mid-excitation lines), and 3.6 (low-lying lines). The continuum of NGC 4418 is further discussed in 3.7.

3.2 The high-lying lines in Arp 220: the nuclear region ( and )

3.2.1 HO

Our models for Arp 220 are similar to those described in G-A04, and generated on the basis of the interplay between the continuum emission and the molecular line absorption. Since the molecular excitation in Arp 220 is high and collisional excitation alone cannot account for it, the excitation mechanism is expected to be dominated by absorption of dust-emitted photons in the nuclear region of the galaxy (G-A04, González-Alfonso et al., 2008, 2010). The molecular excitation is thus a function of the dust temperature, , and of , and in combination with the observed continuum provides clues about the general properties of the far-IR continuum source and its asociated chemistry. If the observed HO excitation cannot be reproduced by assuming that the dust in the nuclear component has a temperature of “only” K (derived from the observed 25 and 1300 m emission), foreground extinction in the far-IR is included to attenuate also the 25 m emission from the nuclei in such a way that the far-IR is reproduced with K. This is Scenario 2, S, in G-A04, based on the continuum models by Soifer et al. (1999). Evidence for such foreground extinction in the mid-IR comes from the strong silicate absorption at 9.7 and 18 m, in the millimeter by self-absorption in CO (Downes & Eckart, 2007), and in the far-IR from the [O i] 63 m line that is observed in absorption, in contrast to most extragalactic sources in which it is observed in emission.

In the present models, we have simulated the dust emission by using a mixture of silicate and amorphous carbon grains with optical constants from Draine (1985) and Preibisch et al. (1993); the mass-absorption coefficient is 550, 150, and 12.3 cm/g of dust at 25, 50, and 200 m, respectively, and the spectral index is in the far-IR. We have also attempted to disentangle the emission from the eastern and western nuclei in Arp 220. We have modeled the eastern nucleus () as a sphere of diameter (Sakamoto et al., 2008, hereafter Sa08), radial opacity (i.e. the opacity along a radial path) at 200 m , and K. The emission from  is attenuated by foreground dust with , yielding Jy at m –consistent with measurements by Soifer et al. (1999) and Downes & Eckart (2007). The corresponding SED of  is shown in Fig. 1b in light-blue, and accounts for a luminosity of L. For the western nucleus () we also use as a first approach the size as derived by Sa08 in the “disk” approximation, i.e. a sphere with diameter pc ( at 72 Mpc), but allow it to vary to match the observed continuum, and leave as free parameters the dust temperature and foreground extinction .

Figure 17: Modeled absorbing flux in Arp 220 of several a) HO, b) OH, c) HO, d) NH, and e) HCN lines as a function of the depth of the molecular shell measured from the surface of the continuum source, as parameterized by the dust opacity at 50 m (). Numbers in parenthesis indicate rounded wavelengths in m. The total column of molecules is proportional to , and we have adopted cm, and abundance ratios of , , , and . The dust temperature is 110 K.

Calculations for the lines were carried out in spherical symmetry using the code described in González-Alfonso & Cernicharo (1999). The models indicate that most high-lying absorption lines are formed in the outermost shell surrounding the far-IR continuum source. This is illustrated in Fig. 17, where the absorbing flux of several lines of HO, OH, HO, NH, and HCN, is plotted as a function of the depth of the shell where the molecules are located, which is parameterized in terms of the continuum opacity at 50 m () measured from the surface of the far-IR source. Since the molecular abundances relative to the density of dust are uniform in these models, the column densities are proportional to . Results indicate that the HO and OH absorption is produced in a thin shell with , as extinction, thermalization by dust emission, as well as line opacity effects make results insensitive to the presence of molecules deeper into the far-IR source. The optically thinner HO lines are mostly formed in the same shell, though the lines at long wavelengths (e.g. the  at 183 m) still have significant contribution (%) from the region. The case for NH and HCN is somewhat different, as some lines probe deeper regions. The high-lying lines of NH lie at m and are also formed in the outermost shell, but the low-lying lines at m have significant contribution from . The high-excitation HCN lines sample regions deeper than the other species () as their excitation is sensitive to the far-IR radiation density at m (see §3.2.4). Based on these results, the models for the nuclei shown below use a “screen” approach, i.e. the molecules are mixed with the dust but located within a thin shell surrounding the nuclei extended up to a depth of for HO, OH, and their O isotopologues, and up to for NH and HCN. The values of the column densities of all species below, however, are given for the outermost shell that dominates the absorption of HO and OH, and are lower limits to the true columns through the nuclear sources. In order to estimate the abundances relative to H nuclei, we have normalized these column densities per unit of to a given value of . For a mass-absorption coefficient of 150 cm/g of dust at 50 m and a gas-to-dust mass ratio of 100, corresponds to a column density of cm, where H refers to hydrogen nuclei in both atomic and molecular forms. This value of has been applied to all species observed in the far-IR, and define the far-IR “photosphere” where most molecular absorption is produced.

Line broadening is caused by microturbulence with km s and a velocity shift of 130 km s through the absorbing shell. The latter simulates either the presence of outflowing gas or, more generally, gas velocity gradients across the nuclear regions, allowing us to nearly match the observed linewidths. Rates for collisional excitation are taken from Faure et al. (2007); with the adopted cm and gas temperature K (see §3.2.4), collisional excitation has little effect on the calculated fluxes. The line models have three free parameters: the dust temperature, , the HO column, , and the covering factor of the continuum source, .

The bulk of the HO absorption is produced in a luminous, compact region with high HO columns, which we identify with the double nucleus of Arp 220. A number of previous observational studies with high angular resolution have revealed that the western nucleus is brighter than the eastern one in the near-IR (Armus et al., 1995), mid-IR (by a factor of at 25 m; Soifer et al., 1999), and (sub)millimeter wavelengths (Downes & Eckart, 2007; Sakamoto et al., 2008). Given that the excitation of HO requires a high brightness continuum source, the observed high-lying HO absorption is tentatively attributed to . Nevertheless, the column densities derived below are independent of whether  alone is responsible for the absorption in the high-lying lines, or  has a significant contribution as well, and the properties we derive for  and  can be more generally interpreted as shared by both nuclei.

In the nuclear region where the HO lines are formed, the models shown below indicate that most of these lines are strongly saturated; however, there are still some high-lying lines that are sensitive to both and . For these critical lines, Figure 18 shows the observed and modeled fluxes relative to that in the  line at 66.4 m (Fig. 3g). As mentioned in §2, the  is chosen as the normalization line because it is detected with a high SNR and, with K, is not expected to be significantly contaminated by absorption of extended low-excitation HO in Arp 220. On the other hand, the  transition is still low in energy as compared with the high-lying lines, so that the line flux ratios are sensitive to both and . Furthermore, the line flux ratios do not depend on either the size of the continuum source, or on . These ratios are plotted as a function of for different “screen-like” HO columns, ranging from to cm, with dashed lines indicating the measured values, and dotted lines the estimated upper and lower limits. We also include in the lowest panel the upper limit derived for the  line. For each , the foreground extinction is determined by imposing that the continuum flux density at 25 m matches the observed 7.5 Jy from the western nucleus (Soifer et al., 1999), but this has little effect on the line flux ratios as the far-IR foreground opacities are low and, in any case, the considered lines have similar wavelengths. The minimum is 90 K, which is the optically thick, lower limit value derived for the western nucleus from the 860 m submillimeter continuum by Sa08.

Figure 18: Modeled HO line ratios in Arp 220 as a function of the dust temperature in the western nucleus. The dashed lines indicate the observed values, and the dotted lines the estimated upper and lower limits. In the lowest panel, the dotted line is the upper limit for the flux ratio. Squares, triangles, circles, and crosses (only for K) show model results for “screen” HO columns of , , , and cm, respectively.

Results of Fig. 18 show that the line ratios scale linearly with . The highest sensitivity to is found for the  line, since its lower level is non-backbone (Fig. 2). Among the detected lines, the line ratios can be almost equally well reproduced with K and the highest cm, or with K and cm. However, the model K and cm yields too much absorption in the undetected  line, slightly favoring K. Concerning the continuum models, if we assume that the observed high-lying HO absorption is dominated by , and that the photosphere at 25 m has the same size as the (sub)millimeter source, a reasonable match to the m SED with K is not possible; for example, the model with K yields Jy at 25 m with no foreground extinction and the size derived from the submillimeter (Sa08), insufficient to account for the observed Jy at 25 m (Soifer et al., 1999). Furthermore, any reasonable fit to the continuum requires foreground absorption at m, and the absolute HO fluxes also require a continuum source larger than the submillimeter one for K. However, the photosphere at 25 m (i.e. the nuclear region with ) may be larger and colder than the submillimeter source if heated by the central core (Sa08), because the optically thinner submillimeter emission samples warm regions that are obscured at 25 m. For K, the required source diameter is pc, yielding L, while for K, pc and L.

In summary, a range of K and a corresponding “screen” column per unit of of cm are derived, with the colder ( K) sources yielding diameters above the (sub)millimeter observed values. The corresponding luminosities are in the relatively narrow range L. The corresponding estimated HO abundances are relative to H nuclei.

As the reference model shown in Figs. 3-5 for detailed comparison with data, we use the combination K and cm, with complete coverage of the source (). The corresponding continuum is shown with a blue curve in Fig. 1b, with the diameter of the source (106 pc) increased by % relative to the submillimeter source, and with a foreground opacity . The data also suggests the presence of a lower excitation component to attain a better fit to the , , and  lines at 90, 108, and 174 m, respectively. Tentatively associating this component with the eastern nucleus, we derive cm and . Its contribution to the HO SLED, shown with light-blue curves in Figs. 3 and 4, is expected to be significant only for low-lying HO lines. Continuum and line parameters of the models are listed in Tables 1 and 2.

3.2.2 Oh

The observed high-lying OH doublets in Arp 220, namely the and at 65 and 53 m, and the and at 71 and 56 m, are compared in Fig. 6 (lower spectra) with the models for the western and eastern nuclei (blue and light-blue curves). Collisional rates between OH and H were taken from Offer et al. (1994). As for HO, the absorption in these lines is dominated by a source with high OH column, which we tentatively identify with . In the model of Fig. 6, we have used the same parameters as for HO, and varied the OH column density to match the observed absorption in the high-lying lines. We derive a OH/HO abundance ratio of ; higher OH columns overpredict the absorption in the OH 53 m doublet (Fig. 6h). The fit to the high-lying OH lines is satisfactory except for the underprediction of the width of the 65 m doublet, which is probably contaminated by relatively weak lines of HO (gray curve in Fig. 6f), NH, and HO. Using the same source sizes as for HO, we find a similar ratio for K. An additional extended region is required to fit the low-lying OH lines, which is discussed in §3.3.

3.2.3 HO and Oh

Although the detected HO lines arise from low-lying levels, we find that these absorption features are mainly produced toward the nuclear region with the highest column densities. Fig. 7 shows the best fit for the HO lines using the same reference model as for the main isotopologue, i.e. K. The best fit HO column density is cm, implying an HO-to-HO ratio as low as , and the upper limit for this ratio is estimated to be . The HO-to-HO ratio slightly depends on the dust temperature, and a value of is found for the model with K, though the overall fit is in this case less satisfactory. Thus our data suggest an enhancement of O in Arp 220 relative to the solar value of . This is confirmed by the models for OH shown in Fig. 8, where the same OH-to-OH ratio of is used with K. The OH-to-OH ratio would also increase up to K for K. The OH 65 and 85 m lines are nearly reproduced, but not the asymmetry in the OH 85 m doublet (Fig. 8b). Similar intensity asymmetries of % have been also observed in most OH doublets by Goicoechea et al. (2011) toward the Orion bar, where the lines are observed in emission. As Goicoechea et al. (2011) indicate, this is probably due to asymmetries in the collisional rates between OH and para-H (see also Offer et al., 1994). In our absorption lines, the asymmetry would arise from the collisional excitation of OH in the ground transition, which favors the sublevel where the m component arises. Since collisions with ortho-H quench the asymmetry, the gas is not very warm but should be dense enough so that collisional excitation with para-H competes efficiently with the excitation induced by the radiation field. The asymmetry is not observed in the main isotopologue (Fig. 6e), indicating that either the higher column of OH efficiently mixes the populations of the sublevels, or that OH is tracing warmer gas with increased contribution by collisions with ortho-H. However, our model with cm and K does not reproduce it, and several tests with higher density and lower were also unsuccesful. Further, since the ground-state OH 120 m doublet is not reproduced with the warm component alone, additional absorption by the  and  components described below, with the same OH-to-OH ratio of , is included (Fig. 8c).

3.2.4 Hcn

Our models for the western nucleus of Arp 220 can also reproduce the fluxes of the HCN lines reported by Rangwala et al. (2011) and the and lines detected with PACS (Figs. 9 and 15). For HCN, both collisional excitation and absorption of far-IR photons in the highest-lying lines are required to reproduce the observations. At long wavelengths (i.e. relatively low lines), collisional excitation in a warm and dense region is the primary excitation mechanism. For moderate dust opacities, these lines would be detected in emission above the continuum. If collisions are able to excite the molecules up to the rotational level where the dust emission becomes optically thick and the far-IR radiation becomes strong, absorption of far-IR photons in high- lines continues to excite the molecule to higher-lying levels and the corresponding HCN lines are observed in absorption. Thus the transition from emission to absorption lines is a measure of the dust opacity and temperature of this component, and the high-lying HCN lines detected in absorption are tracing high far-IR radiation densities as well as a warm-dense environment with high HCN column densities.

A grid of models were generated by varying in the range cm and between 110 and 250 K. For each value of , both and were varied to fit the observed SLED. In all these models, K was adopted. As discussed in §3.2.1, the high excitation HCN lines can be formed deeper into the dusty, continuum source than lines from other species (Fig. 17), and we extend our HCN shell up to a depth of . Rates for HCN collisions with He were taken from Dumouchel et al. (2010), and scaled for the reduced mass of the HCN-H system to obtain the expected collisional rates with H. Our most plausible results were found for cm and K, for which cm, respectively. is twice these values for K. Decreasing below cm would involve HCN columns above cm, implying extreme HCN abundances above that we consider unlikely. The increase of above cm involves lower HCN columns, but the relative absorption in the line becomes somewhat underpredicted and the value of (or the size of the continuum source) is above that derived for HO. Results for K and cm, with the resulting cm, are shown in Figs. 9 and 15. The involved far-IR continuum source has the same size as for HO and OH. Similar values of cm are obtained for K, though the sizes of the continuum and HCN sources are in these cases higher and lower than for K, respectively, and similar to those found for HO and OH.

In our most plausible models, the HCN/HO ratio is 0.1-0.4 (Table 2), and . In Fig. 15, the calculated HCN SLED is compared with results for a model that neglects the radiative excitation by dust (dotted curve), showing that the dust has a flattening effect on the predicted SLED. Less than 10% of the total column is stored in levels associated with the observed lines; the highest populated levels are . Nevertheless, owing to the high continuum opacity and dust temperature in the nuclear region, as well as to self-absorption in the outermost shells of the HCN region, all HCN lines above are predicted in absorption. The emission in the line (Rangwala et al., 2011) is not reproduced, probably indicating that the line is formed in a colder and more extended region associated with relatively weak continuum. On the other hand, the model in Fig. 15 accounts for the observed absorption up to but fails to explain the absorption coincident with the transition at 161.3 m. If this feature is due to HCN, a region as warm as that proposed for NGC 4418 (§3.4.1) would be involved.

Our models also include the HCN excitation through the pumping of the vibrationally excited state, but we find the effect negligible for the ground-vibrational state lines. Evidently, this is a consequence of the low used in our modeling. Rangwala et al. (2011) have proposed that the pumping of the state dominates the HCN excitation, requiring a dust temperature of K. If not completely extincted, this hot component would show up at mid-IR wavelengths, and there is indeed a hot component () in our continuum modeling with K (yellow curve in Fig. 1b) attenuated by mag, which yields an unattenuated flux at 14 m of Jy -twice as observed. However,  has an effective diameter of only pc, while the observed HCN absorption must be produced over spatial scales of pc because, otherwise, the far-IR HCN absorption would be negligible. Thus, if the HCN molecules are exposed to the radiation of