The Mrk 231 molecular outflow as seen in OH
E. González-Alfonso, J. Fischer, J. Graciá-Carpio, N. Falstad, E. Sturm, M. Meléndez, H. W. W. Spoon, A. Verma, R. I. Davies, D. Lutz, S. Aalto, E. Polisensky, A. Poglitsch, S. Veilleux, A. Contursi
aa r X i v : . [ a s t r o - ph . GA ] O c t Astronomy&Astrophysicsmanuscript no. ohmrk231˙def c (cid:13)
ESO 2018November 6, 2018
The Mrk 231 molecular outflow as seen in OH ⋆ E. Gonz´alez-Alfonso , J. Fischer , J. Graci´a-Carpio , N. Falstad , E. Sturm , M. Mel´endez , H. W. W. Spoon , A.Verma , R. I. Davies , D. Lutz , S. Aalto , E. Polisensky , A. Poglitsch , S. Veilleux , A. Contursi Universidad de Alcal´a, Departamento de F´ısica y Matem´aticas, Campus Universitario, E-28871 Alcal´a de Henares, Madrid, Spain Naval Research Laboratory, Remote Sensing Division, 4555 Overlook Ave SW, Washington, DC 20375, USA Max-Planck-Institute for Extraterrestrial Physics (MPE), Giessenbachstraße 1, 85748 Garching, Germany Department of Earth and Space Sciences, Chalmers University of Technology, Onsala Space Observatory, Onsala, Sweden Department of Astronomy, University of Maryland, College Park, MD 20742, USA Cornell University, Astronomy Department, Ithaca, NY 14853, USA University of Oxford, Oxford Astrophysics, Denys Wilkinson Building, Keble Road, Oxford, OX1 3RH, UKPreprint online version: November 6, 2018
ABSTRACT
We report on the Herschel / PACS observations of OH in Mrk 231, with detections in nine doublets observed within the PACS range,and present radiative-transfer models for the outflowing OH. Clear signatures of outflowing gas are found in up to six OH doubletswith di ff erent excitation requirements. At least two outflowing components are identified, one with OH radiatively excited, and theother with low excitation, presumably spatially extended and roughly spherical. Particularly prominent, the blue wing of the absorptiondetected in the in-ladder Π / J = / − / µ m, with E lower =
290 K, indicates that the excited outflowing gasis generated in a compact and warm (circum)nuclear region. Because the excited, outflowing OH gas in Mrk 231 is associated withthe warm, far-infrared continuum source, it is most likely more compact (diameter of ∼ −
300 pc) than that probed by CO andHCN. Nevertheless, its mass-outflow rate per unit of solid angle as inferred from OH is similar to that previously derived from CO, & × (2 . × − / X OH ) M ⊙ yr − sr − , where X OH is the OH abundance relative to H nuclei. In spherical symmetry, this wouldcorrespond to & × (2 . × − / X OH ) M ⊙ yr − , though significant collimation is inferred from the line profiles. The momentum fluxof the excited component attains ∼ L AGN / c , with an OH column density of (1 . − × cm − and a mechanical luminosity of ∼ L ⊙ . In addition, the detection of very excited, radiatively pumped OH peaking at central velocities indicates the presence of anuclear reservoir of gas rich in OH, plausibly the 130-pc scale circumnuclear torus previously detected in OH megamaser emission,that may be feeding the outflow. An exceptional OH enhancement, with OH / OH .
30 at both central and blueshifted velocities,is most likely the result of interstellar-medium processing by recent starburst and supernova activity within the circumnuclear torusor thick disk.
Key words.
Line: formation – Galaxies: ISM – ISM: jets and outflows – Infrared: galaxies – Galaxies: individual: Mrk 231
1. Introduction
Current models of galaxy evolution involve galactic-scale out-flows driven by starbursts and active galactic nuclei (AGN) askey ingredients. The outflows trace the negative feedback fromAGN and / or star formation on the molecular gas, eventuallyshutting o ff the feeding process and quenching the growth ofthe stellar population and / or of the supermassive black hole (e.g.di Matteo et al., 2005). AGN feedback could be responsible forthe observed black hole mass-velocity dispersion relationship(Murray et al., 2005) and create a population of red gas-poor el-lipticals. In the past, outflows have been observed in many star-bursts and quasi-stellar objects (QSOs), mostly in the ionizedand neutral atomic gas component (e.g. Veilleux et al., 2005, fora review).Molecular gas may dominate the mass-outflow rate of out-flows, providing important constraints on the timescale for dis-persing the (circum)nuclear gas in the host galaxy. Molecularoutflows have been reported in several galaxies at millime-ter wavelengths (e.g. Baan et al., 1989; Walter et al., 2002;Sakamoto et al., 2009). The discovery of a massive molec- ⋆ Herschel is an ESA space observatory with science instruments pro-vided by European-led Principal Investigator consortia and with impor-tant participation from NASA. ular outflow in Mrk 231, an ultraluminous infrared galaxy(ULIRG) harboring the closest quasar known, in
Herschel / PACS(Pilbratt et al., 2010; Poglitsch et al., 2010) spectroscopy is a keyfinding. The outflows are traced by P-Cygni OH line profiles inthe 79 and 119 µ m doublets, and in the high-lying 65 µ m dou-blet ( E lower ≈
300 K), with high-velocity shifts of > / s(Fischer et al., 2010; Sturm et al., 2011, hereafter F10 and S11,respectively). Analysis and model fits of these lines yielded apreliminary mass-outflow rate of ˙ M ∼ M ⊙ y − . The extremeoutflow was also detected at millimeter wavelengths in CO, giv-ing a similar ˙ M (Feruglio et al., 2010), and in HCN, HCO + , andHNC (Aalto et al., 2012). Recently, Cicone et al. (2012) havefound that the outflowing CO is not highly excited relative to thequiescent gas, and that the outflow size decreases with increasingcritical density of the transition. From neutral Na I D absorption,Rupke & Veilleux (2011) estimated a similar ˙ M ∼
400 M ⊙ y − on spatial scales of ∼ tially resolved, the observed excitation conditions provide in-formation about the spatial extent of the outflow, which enablesthe estimation of the outflow physical parameters (mass-outflowrate, mechanical power and energy). The high-velocity molec-ular outflows were found to be common in local ULIRGs, andpreliminary evidence suggested that higher AGN luminosities(and higher AGN contributions to L IR ) correlate with higher ter-minal velocities and shorter gas depletion timescales (S11).In this work, we present the velocity profiles and fluxes of allof the OH and OH doublets seen in the
Herschel / PACS spec-troscopic observations of Mrk 231, and an analysis of these pro-files and fluxes based on radiative-transfer modeling. In Sect. 2we discuss the details of the observations and give an overviewof the general characteristics of the profiles, together with qual-itative assessments on the excitation conditions, optical depths,far-IR extinction, geometry, and O / O abundance ratio in thecircumnuclear region of Mrk 231. In Sect. 3, we discuss theradiative-transfer models that are used to quantitatively analyzethe observations, and the motivation for, properties of, and de-rived parameters of the several components that we use to char-acterize the gas seen in OH. In Sect. 4, we summarize the pic-ture that emerges from the observations and the modeled com-ponents, their relationship to structures seen in other diagnostics,and the implications for the role of the AGN and circumnuclearstarburst. We adopt a distance to Mrk 231 of 192 Mpc ( H = − Mpc − , Ω Λ = .
73, and z = .
2. Observations and description of the spectra
Following the detection of outflows traced by the ground-stateOH doublets at 119 and 79 µ m and of the excited OH 65 µ mtransition, based on the guaranteed-time key program SHININGobservations (hereafter GT observations, PI: E. Sturm; F10,S11), completion of the full (52 . −
98, 104 . − µ m),high-resolution PACS spectrum of Mrk 231 was carried out onOctober 16 (2012) as part of the Open Time-2 Herschel phase(hereafter OT2 observations; PI: J. Fischer). The spectrum of theOH Π / J = / − / µ m was taken from anOT1 program (PI: R. Meijerink). The GT observations and re-duction process were described in F10 and S11. For the OT2and OT1 observations, the spectra were also taken in high spec-tral sampling density mode using first and second orders of thegrating. The velocity resolution of PACS in first order rangesfrom ≈
320 km s − at 105 µ m to ≈
180 km s − at 190 µ m, andin second order from ≈
210 km s − at 52 µ m to ≈
110 km s − at 98 microns. The data reduction was carried out mostly us-ing the standard PACS reduction and calibration pipeline (ipipe)included in HIPE 6.0 and HIPE 10.0. The two HIPE versionsyielded essentially identical continuum-normalized spectra, withthe continuum level from v10 stronger than that from v6 by upto ∼ E lower =
620 K, were detected and are indicated in the energy-level diagram of Fig. 1. These are the same transitions as weredetected in NGC 4418 and Arp 220 (Gonz´alez-Alfonso et al.,2012, hereafter G-A12), except for the cross-ladder Π / − Π / J = / − / λ rest = µ m (detected in NGC 4418)that is redshifted in Mrk 231 into the gap at ≈ µ m between Fig. 1.
Energy level diagram of OH showing the transitions de-tected in Mrk 231 with Herschel / PACS and Spizer / IRS with blueand green arrows, respectively. The l − doubling splitting of thelevels is not indicated. Colored numbers indicate rounded wave-lengths in µ m. The Π / J = / − / Π / − Π / J = / − / λ rest ≈
99 and 96 µ m(the latter detected in NGC 4418, G-A12) were not observed, asthey are redshifted into the PACS gap at 100 µ m. In the text, wedenote a doublet by giving its wavelength (e.g. OH119).the green and red bands. For simplicity, we denote a given dou-blet by using its rounded wavelength as indicated in Fig. 1 (e.g.OH119). The OH doublets, lying close in wavelength to theOH transitions, were also observed and unambiguously detectedat 120, 85, and 65 µ m. Spectroscopic parameters for the OHdoublets were taken from Polehampton et al. (2003); the posi-tions of the OH119 and OH84 doublets, expected to havethe strongest signatures, are indicated in Figs. 2a and c. Thereis no evidence in the spectra for either absorption or emissionattributable to OH, whose transitions fall between those of thetwo more abundant isotopologs.The abscissas in Fig. 2 indicate the velocity relative to theshorter-wavelength component (hereafter, blue component) ofeach doublet, and are calculated for a redshift of z = . OH doublets are indicated withblack arrows, while those of potentially contaminating lines ofother species (discussed in Sect. 2.2) are indicated with blue ar-rows. In panel b, three independent spectra of the OH79 doubletare shown for comparison; they are listed in Table 1.To characterize the molecular outflow traced by the OH lines,it is important to have a flat baseline that minimizes the uncer-tainties in the continuum level, and thus in the velocity extent ofthe line wings. The central spatial pixel (spaxel) of the 5 × / N) spec-trum, and was adopted whenever the continuum level was flatand the baseline was well characterized. However, the contin-uum level from the central spaxel shows low-level fluctuationsat some wavelengths that probably result from small pointing
Table 1.
Herschel / PACS observations of OH in Mrk 231.
Transition Rest wavelengths a Program b Obs. ID c Number of σ ( µ m) spatial pixels d (Jy)OH Π / − Π / − . − .
441 GT 1342186811 1 0 . Π / − Π / − . − .
181 GT 1342186811 1 0 . . . Π / − Π / − . − .
351 OT2 1342253530 1 0 . Π / − Π / − . − .
597 OT2 1342253537 3 0 . Π / − Π / − . − .
279 GT 1342207782 1 0 . . . Π / − Π / − . − .
216 OT2 1342253534 3 0 . Π / − Π / − . − .
057 OT2 1342253530 1 0 . Π / − Π / − . − .
950 OT2 1342253531 3 0 . Π / − Π / − . − .
397 OT1 1342223369 1 0 . a The two values correspond to the two Λ − components of the doublets; each one is the average of the two hyperfine transitions that make up acomponent. b GT and OT indicate “guaranteed time” and “open time”. c Identification number. d Number of spatial pixels used to generate the spectrum. In case of 1, the central pixel was used; in case of 3, the spectra from the three brightestspatial pixels were coadded. In all cases, the absolute fluxes were scaled to match the total flux as obtained by coadding the fluxes from the 25pixels. drift motions. In these cases, we coadded the flux densities re-sulting from the three brightest spaxels, which resulted in flatbaselines at the expense of a lower S / N. Regardless of the num-ber of spaxels used, the resulting spectra were re-scaled to thecontinuum level of all 25 spaxels combined (to account for pointspread function losses and pointing uncertainties, S11). Table 1lists the number of spaxels used to generate the spectrum of eachOH doublet, and the baselines are indicated in Fig. 2 with dashedlines.We also analyzed the OH35 ground-state doublet observed inthe Spitzer IRS long-high spectrum of Mrk 231. The spectrum,presented in Fig. 3, is the result of combining eight indepen-dent observations of the source obtained between 2006 and 2009as part of the IRS calibration program (earlier versions of thespectrum can be found in Farrah et al. (2007) and Armus et al.(2007)).
The OH spectra displayed in Fig. 2 show a diversity of lineshapes. The ground-state OH119 and OH79 doublets (panelsa and b) exhibit prominent P-Cygni profiles, indicative of out-flowing gas with the absorption produced in front of, and theemission feature laterally adjacent to and behind the far-IRsource. Absorption in OH119 is found up to a blueshifted ve-locity of − − , while the case of OH79 is uncertaindue to contamination by H O. Emission in OH79 is detectedup to a velocity of ≈
770 km s − from the red OH component.The third ground-state transition within the PACS range, theOH53.3 doublet (panel e), does not show any emission feature.Its blueshifted absorption extends up to at least − − ,being contaminated by the very high-lying OH53 doublet atmore negative velocities (also shown in panel e). The SpitzerIRS OH35 spectrum (Fig. 3), which has significantly lower spec-tral resolution (500 km s − ), shows absorption that peaks at cen-tral velocities with no emission feature; the absorption on theblue side is more prominent than on the red side. Detection of OH79, OH53.3, and OH35 implies that OH119, with a muchhigher opacity (F10), is optically thick. However, the peak ab-sorption in OH119 is only 30% of the continuum, indicating thatthe OH119 doublet at a given velocity only covers a fraction ofthe total 119 µ m continuum, and / or that the 119 µ m transition isvery excited .The excited OH84 and OH65 doublets (panels c and d) donot show any emission feature either, but display prominentblueshifted absorption. It is worth noting that while OH65 showsabsorption up to ∼ − − , the less excited OH84 doublet(with lower S / N) only shows absorption up to ∼ − − .The reliability of the extreme OH65 blueshifted absorption isdiscussed in Sect. 2.3.While the peak absorption in the OH119 and OH79 doubletsis blueshifted by −
300 and −
240 km s − , respectively, relative tothe blue component of the doublet, the OH53.3 and OH35 peakcloser to central velocities. The increase in line excitation alongthe Π / ladder also shifts the peak absorption toward rest ve-locities, with the high-lying OH65 and OH53 doublets peakingat nearly v = − . The latter transition does not show anyhint of blueshifted absorption to within the S / N. These velocityshifts suggest that the excited lines trace an outflow region notentirely coincident with that probed by the ground-state OH119and OH79. The OH119 and OH65 spectra also show blueshiftedabsorption at velocities significantly higher than the line wingemission in CO and HCN, which is observed just out to ∼ − (Feruglio et al., 2010; Cicone et al., 2012; Aalto et al.,2012).Along the Π / ladder, the high-lying OH71 doublet (panel f)also peaks at rest velocities, with possible blueshifted absorptionthat is uncertain due to the proximity of a strong H O line at ≈ −
500 km s − . Hints of emission are also seen at redshiftedvelocities in the OH71 doublet. The absorption strength of the transition approaches zero if the ex-citation temperature of the transition approaches the dust temperature.3onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH
Fig. 2.
Herschel / PACS spectrum of Mrk 231 around the 9 OH doublets, with the adopted baselines (dashed lines). The velocity scaleis relative to the blue Λ − component of each doublet and has been calculated with respect to the systemic redshift of z = . ≈
50 km s − ), and there are hints of redshiftedemission at velocities higher than 700 km s − (in line with theemission feature in OH79 at extreme, positive velocities). TheOH163 doublet also shows blueshifted absorption up to −
350 km s − , as well as a relatively weak emission feature at − − that is probably attributable to ( ∼
100 km s − redshifted)CO (16-15). Potential contaminations by lines of species other than OH areindicated in Fig. 2, and Fig. 4 compares the OH65 and OH84spectra of Mrk 231, Arp 220, and NGC 4418. In OH119, theredshift component J = / + − / − of the excited N = − E lower ≈
105 K, 118 . µ m) is not expected to contam-inate the blueshifted OH absorption at − − , becausethe blue component J = / − − / + of the doublet is undetected.Likewise, there are no apparent features at the positions of thetwo indicated p-H O + lines (4 − / − / / − / + − OH doublet (F10), as the ground transition ofCH + is bright in Mrk 231 (van der Werf et al., 2010).In the OH79 profile, the H O ⁀ − − (78 . µ m, E lower ≈
250 K) is clearly detected (F10,Gonz´alez-Alfonso et al., 2010, hereafter G-A10), though theprominent blueshifted absorption apparent in the GT observa-tion is not confirmed with the OT2 observation (Table 1). On theother hand, the higher-lying H O ⁀ −
720 km s − (78 . µ m, E lower ≈
600 K) was not seen in the GT obser-vation, but a spectral feature in the OT2 observation makes thecase uncertain.In the OH84 profile, the only potential contamination is dueto NH (Fig. 4a), which is expected to generate an absorptionwing-like feature between 300 and 1300 km s − associated withthe (6 , K ) a − (5 , K ) s ) lines (G-A12). However, from the uncon-taminated (6 , K ) s − (5 , K ) a ) NH lines at 83 . − . µ m, only the(6 , s − (5 , a ) line is (marginally) detected, thus no significantcontamination by NH to OH84 is expected.In the OH65 profile, no spectral feature coinciding with thep-H O + − / − / − − is found (alsoundetected in Arp 220, Fig. 4b). The H O ⁀ ≈ − could contaminate the main OH feature (G-A12), and o-H O + − / − / ⁀ OH absorption feature.A weak contribution by the H O ⁀ −
700 km s − in the OH53.3 profile (Fig. 2e). The OH71spectral region is complex (Fig. 2f), with possible baseline cur-vature and contaminations on the blue side by H O ⁀ −
430 km s − ( E lower ≈
400 K, 71 . µ m) and possibly H O + − − + and 4 − − + (both detected in NGC 4418 and Arp 220,G-A13). These lines are expected shortward of an apparent ab-sorption pedestal & − wide, to which both H O andOH may be contributing. On the red side, absorption by the veryhigh-lying H O ⁀ − is detected.The OH56 spectrum shows strong absorption by the H O ⁀ − . The OH163 spectrum is freeof contaminations, except for the relatively weak CO (16-15)emission line. Finally, the OH35 spectrum (Fig. 3) shows the [SiII] line in emission at ∼ − , precluding the measure-ment of the OH35 doublet profile.
As mentioned in Sect. 2.1, at least the OH119, OH79, OH53.3,OH84, and OH65 doublets show high-velocity absorption wingsextending up to at least − − from the blue componentof the doublets; the possible OH71 line wing is contaminated byH O.While the OH79 and OH53.3 are contaminated at veloci-ties more blueshifted than − − , the OH119, OH84,and OH65 doublets along the Π / ladder are uncontaminated Fig. 3.
Spitzer IRS long-high (LH) spectrum around the OH35doublet, with the adopted baseline (dashed line). The emissionfeature at ∼ − is the [Si II] line at 34.815 µ m. Thespectral resolution is 500 km s − .throughout the blue range and probe the outflowing gas at themost extreme velocities. OH119 shows absorption up to − − , but OH84 only up to − − with a lower S / N.Very intriguing then is the apparent strong absorption in themore excited OH65 doublet, up to velocities of at least ∼ − − . Figure 5 compares three spectra of the OH65 doublet.Both the GT spectra of the central spaxel (GT-cen, see Table 1)and of the three brightest spaxels (GT-avg, also shown in Fig. 2d)show the line wing covering a similar velocity range, indicat-ing that it cannot be ascribed to pointing e ff ects that wouldgenerate fluctuations of the continuum level. Furthermore, theOT2-avg spectrum, with a flat baseline (albeit with a lower S / N,though mostly at positive velocities), also shows the line wingwith a similar velocity extent. The close agreement between theOH65 GT and OT2 high-velocity wing profiles together with thesmaller extent of the OH84 profile provide strong evidence forthe presence of highly excited gas at extreme velocities.Figure 6 compares the blue line wings observed in OH119,OH79, OH84, and OH65 in more detail and shows the ratios oftheir absorption strengths. For velocities in the range −
300 to − − , the intensity in all spectra decreases smoothly,and the line ratios are nearly constant with the exception ofthe 65 /
79, which increases significantly with increasing velocityshift. At higher velocity shifts, the 65 /
84 and 65 /
119 ratios jumpto higher values. The limited S / N of the OH84 spectrum and thebaseline uncertainties do not allow us to establish an accuratelimit for the 65 /
84 ratio, but it is most likely > . P-Cygni profiles are observed in at least the OH119 and OH79doublets. The upper level of the OH119 transition is only e ffi -ciently populated from the ground-state through absorption of a119 µ m photon or through a collisional event, that is, there isno e ffi cient excitation path that involves radiative pumping to ahigher-lying level and cascade down to the Π / J = / .Under certain idealized conditions, this translates into a relation-ship between the fluxes of the absorption and emission features. Pumping the Π / J = / , / . µ m photons and subsequent decay via Π / − Π / J = / , / − / ffi cient as the latter cross-ladder transitions are relatively5onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH Fig. 4.
Spectra around (a) the OH84 doublet, and (b) the OH65doublet, in Mrk 231 ( upper spectra ), Arp 220 ( middle ), andNGC 4418 ( lower ). The vertical-dotted lines in (a) indicate thepositions of NH and OH lines. The NH lines at v < − − are relatively strong in NGC 4418, indicating contribu-tion by NH to the absorption at v ∼ − . This is not thecase of Mrk 231, so that there is no significant contamination at v ∼ − . The OH66 doublet in (b) is complex withprobable contributions by o-H O + and NH .Assuming spherical symmetry, a two-level system, pure radia-tive excitation (i.e. negligible collisional excitation), and an en- weak, but can still boost the emission feature relative to the absorptionone by ∼ Fig. 5.
OH65 doublet toward Mrk 231. Three spectra are com-pared: GT-cen is the central spaxel of the GT observations, GT-avg is the average spectrum of the three brightest spaxels in theGT observations, and OT2-avg denotes the average spectrum ofthe three brightest spaxels in the OT2 observations. All threespectra show a prominent blueshifted line wing extending up toat least ∼ − − .velope size that is much larger than the size of the continuumsource (i.e. with negligible extinction of line photons emittedfrom behind the continuum source), then statistical equilibriumof the populations and complete redistribution in angles ensuresequal number of absorption and emission events as seen by anobserver that does not spatially resolve the outflow. In that case,and due to conservation of continuum photons, the outflowinggas would have the overall e ff ect of redistributing the contin-uum photons in velocity space, generating a redshifted emissionfeature as strong as the blueshifted absorption feature (this isanalogous to the H O 6 µ m band in Orion BN / KL where theP-branch is observed in emission and the R-branch in absorp-tion, Gonz´alez-Alfonso et al., 1998). In Mrk 231, however, theemission feature in the OH119 doublet is five times weaker thanthe absorption feature, revealing ( i ) the importance of extinctionof line-emitted photons arising from the back side of the far-IRsource, and / or ( ii ) significant departures from spherical symme-try, with the outflow mainly directed toward the observer (e.g.bipolar) and / or the receding component intrinsically less promi-nent than the approaching one. These conclusions are strength-ened if collisions in a warm and dense region, suggested by the In spherical symmetry, all observers located at the same distancefrom the source would detect exactly the same spectrum, and since weassume that there is neither cooling in the line (no collisions) nor ab-sorption of line-emitted photons by dust, conservation of the contin-uum emission holds regardless of the line opacity, implying equal ratesof absorption and emission events.6onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH
Fig. 6. a)
Blueshifted line wing as observed in four OH dou-blets at 119 (black histogram), 79 (green), 65 (red), and 84 µ m(blue). The spectra are resampled to the velocity resolution of theOH119 spectrum. The velocity scale is relative to the blueshiftedcomponent of each doublet. b) Ratios of the absorption strengths(1 − F ν / F C ). The vertical dotted line at −
300 km s − marks thelimit of the outflow region for the full set of lines.bright emission observed in the ground-state transitions of sev-eral species (van der Werf et al., 2010), enhance the OH119 ex-citation in the receding emitting gas.Since the OH84 doublet does not show a redshifted emis-sion feature (Fig. 2c and 4a), we argue in Sect. 3.3.1 that thecircumnuclear outflowing gas does not significantly contributeto the OH119 emission feature (at v >
200 km s − from the redcomponent, and for the reasons pointed out above), which thenarises primarily from a spatially extended low-excitation compo-nent where the far-IR extinction is presumably low (Sect. 3.4).Assuming that collisional excitation of OH in this componentis negligible (i.e. scattering of dust-emitted photons at 119 µ mis responsible for the observed emission), the strength of theOH119 emission feature relative to the observed continuum levelthen measures, under some conditions, the clumpiness of the re-ceding gas. If the receding OH were intercepting the full 119 µ m continuum emitted toward the back 2 π sr within a velocityinterval ∆ v , and neglecting again extinction, one would expectan emission feature as strong as ∼
50% of F ∆ v ( F is theobserved continuum flux density at 119 µ m), i.e. ∼ × Jy Fig. 7.
Comparison between the OH79 and OH163 (verticallyshifted for clarity) continuum-subtracted spectra. The velocityscale is here relative to the red Λ − component of each doublet,with the aim of directly comparing the emission features. Thevertical arrows indicate the positions of the blue components ofthe doublets.km s − for ∆ v ∼
500 km s − (see Fig. 2a) . The observed emis-sion feature, however, accounts for only ∼ . × Jy km s − ,thus suggesting a covering factor of ∼
20% for the above condi-tions. If the receding OH is behind the source of 119 µ m contin-uum emission, only ∼
20% of that continuum within ∆ v = − is thus intercepted and reemitted by the OH, indicat-ing high clumpiness of the spatially extended reemitting OH.This is consistent with the clumpiness of the outflow inferred byAalto et al. (2012) from HCN emission.The ground-state OH79 doublet shows a relatively strongemission feature, with a flux that is nearly 75% of the absorp-tion flux. Since extinction at 79 µ m is higher than at 119 µ m, theprominent OH79 emission feature indicates the e ff ect of radia-tive pumping through absorption of 53.3 and 35 µ m continuumphotons, and subsequent cascading down to the ground state viathe 163 and 79 µ m transitions (G-A12). The OH163 doublet(panel h) is indeed mostly observed in emission, qualitativelymatching the pumping scheme. More quantitatively, Fig. 7 com-pares in detail the OH79 and OH163 profiles, with the velocityscale relative to the red component of each doublet. The fluxof the OH163 doublet between 0 and 210 km s − is ≈
400 Jykm s − , about 10% higher than the flux emitted in the OH79doublet in the same velocity interval (i.e. the narrow-emissionfeature, 365 Jy km s − ). Owing to the contribution to OH79 by aprominent redshifted line wing that is weak in OH163, the intrin- It is also assumed that the red component of the doublet dominatesthe emission feature and that this emission is radiatively decoupled fromabsorbing foreground gas; the redshifted emission due to the blue com-ponent of the doublet is cancelled or blocked by the blueshifted ab-sorption due to the red doublet component. Note also that we use theobserved line width of the emission feature as an (approximate) proxyfor the velocity range within which τ OH119 ≥
1. 7onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH sic flux of the narrow OH79 emission feature without the wingcontribution is estimated to be ∼
200 Jy km s − , a factor of ∼ µ m line-emittedphoton should be accompanied by a 79 µ m one (Fig. 1), but theOH79 emission is additionally boosted by direct scattering of 79 µ m dust-emitted photons (which does not involve emission inthe OH163 doublet), the di ff erence in fluxes indicates, regard-less of geometry, that indeed significant extinction a ff ects thenarrow emission feature in the OH79 doublet.The prominent redshifted line-wing observed in emission inthe OH79 doublet is also remarkable, with a flux of ≈
400 Jykm / s between 210 and 800 km / s (Fig. 7). Some hints of emissionare also found in the OH163 doublet at >
210 km s − , with anuncertain ( ± ∼
185 Jy km s − . This flux is weakerthan that measured in the OH79 redshifted wing, thus indicat-ing that the emission in this OH79 wing is not significantly ex-tincted and that resonant scattering of 79 µ m dust-emitted pho-tons probably dominates the OH79 wing-emission feature. It isthus plausible that the OH79 redshifted wing is more extendedthan the source of the far-IR emission. The velocity extent ofthe OH79 redshifted emission feature is similar to that of othermolecular lines at millimeter wavelengths (Feruglio et al., 2010;Cicone et al., 2012; Aalto et al., 2012), suggesting a similar spa-tial origin.The ground-state OH53.3 doublet (Fig. 2e) does not showan emission feature, due in part to the slightly higher chancefor an OH molecule in the Π / J = / µ m transition instead of directly emitting a 53 . µ m pho-ton (Fig. 1), but also further indicating the role of extinction.Similarly, the OH35 doublet profile (Fig. 3) only shows absorp-tion, as expected given that essentially all molecules pumped tothe upper Π / J = / µ m pho-ton along the Π / ladder (instead of emitting a 35 µ m photon).There is significant redshifted absorption in OH35, indicatingthat there is still 35 µ m continuum emission behind part of thereceding gas.It is then intriguing that the high-lying OH71 doublet(Fig. 2f) shows hints of redshifted emission, with a flux of 150 Jykm s − . The reliability of this emission feature is uncertain, how-ever, as it shows di ff erent strengths in the central-spaxel and av-eraged spectra. The line should be formed very close to a warmsource of far-IR radiation, which is probably optically thick atthese wavelengths. If the feature is not an artifact of the base-line, inhomogeneities of the dust extinction in the nuclear regionand geometry e ff ects would be required to account for it. In Fig. 2, the excited OH lines (other than OH163) showstrong absorption at central velocities, similar to NGC 4418 andArp 220 (G-A12). This reveals the presence of a very excited,non-outflowing component in the nuclear region of Mrk 231.However, no trace of a relatively narrow absorption feature isfound at central velocities in the OH119 and OH79 doublets.This is conspicuous, because extinction at 119 / µ m wouldstrengthen the line absorption relative to the emission feature,as argued in Sect. 2.4 for the outflowing gas. In a quiescent com-ponent, the absorption and reemission occur at the same centralvelocities, so that one would expect a resulting central absorp-tion feature in spherical symmetry. The other ground-state tran- Fluxes in units of Jy km s − are proportional to the rate of de-excitation events in the line. sitions, the OH53.3 and OH35 doublets, do show strong absorp-tion at central velocities.The lack of measurable absorption at central velocities in theOH119 transition may be partially due to the fact that the en-closed dust is very warm ( >
100 K) and compact, thus emittingweakly at 119 µ m compared with the total emission at this wave-length. In this case, the OH119 absorption would be strongly di-luted within the observed 119 µ m continuum emission, whosemain contribution would arise from more extended regions de-void of quiescent OH. In addition, collisional excitation in awarm and dense circumnuclear component would also excite theOH molecules to the level of near radiative equilibrium with thedust, thus producing negligible absorption. A potentially impor-tant e ff ect is also resonant scattering of dust-emitted photons inthe OH119 doublet in a flattened structure (e.g. a torus or disk)seen nearly face-on or moderately inclined, which would tendto cancel the absorption produced toward the strongest contin-uum source. It is worth noting that since the OH119 transitionis ground-state, this process could work on spatial scales signif-icantly larger than the region responsible for the high-excitationabsorption observed at systemic velocities (see also Sect. 3.2).While in the OH79 transition the pumping via the 53 . µ m transitions enhances the reemission, in the OH53.3 dou-blet the upper level is higher in energy and hence more di ffi cultto excite collisionally, and reemission is less favored becauseof the competing de-excitation path via the OH163 transition(Fig. 1).The OH119 and OH79 profiles of Mrk 231 are in this respectvery di ff erent from those observed in Arp 220, which shows inthese doublets strong absorption at central velocities (G-A12).The OH spectra of the high-lying lines are more similar at centralvelocities (Fig. 4), indicating that both sources have highly ex-cited OH. This indicates that the components that are responsiblefor the ground-state absorption in Arp 220 at central velocities,that is, C halo and C extended (G-A12), are absent in Mrk 231, whichis consistent with the face-on view of the disk at kpc scales. The equivalent widths of the OH35 and OH53.3 doublets are ≈
45 and ≈
120 km s − , respectively. For optically thin absorp-tion, ignoring reemission in the lines and assuming that the OHmolecules are covering the whole continuum source at the cor-responding wavelengths, the equivalent width of a doublet (inunits of velocity) is given by W eq = λ g u A ul N OH , gr π g l , (1)where λ is the wavelength, A ul is the Einstein coe ffi cient forspontaneous emission, g u ( g l ) is the degeneracy of the upper(lower) level, and N OH , gr is the OH column density in the twolambda-doubling states of the ground Π / J = / W eq (OH53 . / W eq (OH35) ≈ .
4, whilethe observed ratio is ∼
3. This indicates that opacity e ff ects areimportant even in the OH53.3 doublet, which is less opticallythick than the OH79 doublet.Using the OH35 doublet, the lowest optical depth ground-state doublet, in eq. (1) gives N OH , gr ≈ . × cm − , whichis a lower limit for N OH because ( i ) a significant fraction ofmolecules is in excited levels; ( ii ) extinction at 35 µ m only en-ables the detection of OH in the external layers of the continuumsource; and ( iii ) the OH may not be covering the whole 35 µ mcontinuum source. OH Up to three OH doublets are detected within the PACS range,at 120, 85, and 66 µ m. While the OH120 doublet may be par-tially contaminated by CH + , and OH66 by NH and H O + , the OH85 doublet is free from contamination, with an integratedflux about five times weaker than the OH84 doublet. This con-firms the strong enhancement of OH in Mrk 231 (F10). It isalso worth noting that while the OH84 profile shows a dip in ab-sorption between the two lambda-doubling components, a nearlycontinuous bridge of absorption is seen between the OH com-ponents (probable contamination makes the case uncertain in OH65, where the absorption peaks in between the two lambdacomponents). This may suggest a relative enhancement of OHin the outflowing gas.
3. Models and interpretation
To estimate the physical properties of the molecular outflow asderived from the OH doublets, we analyzed the OH line profilesand fluxes using radiative-transfer models. We used the code de-scribed in Gonz´alez-Alfonso & Cernicharo (1999), which cal-culates in spherical symmetry the line excitation due to the dustemission and collisions with H , and includes opacity e ff ects (i.e.radiative trapping), non-local e ff ects, velocity gradients, extinc-tion by dust, and line overlaps (Gonz´alez-Alfonso & Cernicharo,1997). For a given model, the code first calculates the statistical-equilibrium populations in all shells that make up the source, andthen the emerging line shapes are computed, convolved with thePACS spectral resolution, and compared directly with the obser-vations.As shown below, at least three components (two outflowingcomponents with di ff erent velocity fields, spatial extents, andfar-IR radiation sources, and one relatively quiescent compo-nent with little or no outflowing motion, hereafter referred to asthe QC) are required to obtained a reasonable match to the ob-served line profiles. The di ff erent components are modeled sep-arately, and the corresponding emerging flux densities are thensummed up together (i.e. it is assumed that the di ff erent compo-nents do not simultaneously overlap along the line of sight andin the projected velocity). Figure 8 depicts the generic modelfor a given outflowing component (the QC component describedbelow is modeled as in G-A12). A central source of far-IR ra-diation is characterized by its radius R int , dust temperature T dust ,and optical depth at 100 µ m along a radial path τ . This issurrounded by an envelope of outflowing molecular gas withexternal radius R out . The OH is mixed with the dust in the en-velope, where T dust ∼ r − . (e.g. Adams, 1991) and τ = . R int and R out6 . The gas is outflowing in spherical sym-metry with velocity and H density profiles v ( r ) and n ( r ), respec-tively. To decrease the number of free parameters, we imposed aconstant velocity gradient (i.e. v ( r ) = v int + dv / dr ( r − R int ), where This only applies to the high-velocity component (HVC) discussedbelow, where a column of N H ∼ cm − for the outflowing shell isestimated (Table 4). A value of τ ≈ .
25 across the outflowing gasis then expected (for N H ∼ × cm − per unit of τ , G-A12). Wedoubled that number to roughly simulate illumination by an externalradiation field and / or emitting clumps mixed with the outflowing gas,though this has a weak e ff ect on results because the excitation is domi-nated by the central far-IR source. Fig. 8.
Schematic representation of a given source component. Itconsists of ( i ) a central source of far-IR emission characterizedby its radius ( R int ), dust temperature ( T dust ), and optical depth at100 µ m along a radial path ( τ ), and ( ii ) the surrounding (out-flowing) gas, with external radius R out , which is mixed with dust.The gas expands radially with a velocity field v ( r ) and H nucleidensity n ( r ), such that n OH × r × v is constant (i.e. the mass-outflow rate is constant). The dust in the outflowing envelope has τ = . T dust ∼ r − . . In some models,departures from spherical symmetry are simulated by calculat-ing the emerging fluxes only for impact parameters lower than p f (i.e. between the two dashed lines). The blue and red curves(and grayscale) show the isocontours of line-of-sight velocities(blue: approaching; red: receding) for a decelerating outflow; thedarkest colors correspond to the highest (approaching or reced-ing) velocities. In our best-fit models for the excited OH lines(the HVC component in Sect. 3.1.2), the outflowing envelope isless extended than in this representation ( R out / R int . .
5) and iscollimated ( p f < R out ). dv / dr = ( v out − v int ) / ( R out − R int ) is constant) and a constant mass-outflow rate (mass conservation then implies that n OH × r × v isindependent of r ).A constant OH abundance relative to H nuclei, X OH = . × − , was adopted (S11), as derived to within a factor of ∼ X OH so that ourresults can be easily rescaled.According to the results shown below, strict spherical sym-metry is not an accurate approach in some models, and gas out-flowing along two approaching and receding cocoons (i.e. withlittle gas expanding along the plane of sky) is favored. This isroughly simulated by including the free parameter p f , such thatthe emerging fluxes are calculated only for impact parameters p < p f (i.e. for rays within the cylinder depicted with dashedlines in Fig. 8) . Finally, the continuum-subtracted emergingprofiles of a given component can be multiplied by a factor f ≶
1, which represents either partial OH covering of the far- Note that this is only an approximation, as the level populations arecalculated in spherical symmetry. 9onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH
Table 2.
Parameters for the modeling of the OH outflow.
Parameter Units Meaning Explored range (HVC) R int pc Radius of the far-IR continuum source a e T dust K Dust temperature of the far-IR continuum source 90 − τ Continuum optical depth at 100 µ m along a radial ( R int ) path 0 . − R out / R int Radius of the outflowing envelope relative to R int . − . f v int km s − Gas velocity at R intb , c − g v out km s − Gas velocity at R outb , c − g N OH cm − OH column density from R int to R outb , c (0 . − × p f pc Limiting impact parameter for the calculation of emerging fluxes R int − R outh f Scaling factor da It coincides with the inner radius of the OH envelope. b A uniform velocity gradient is adopted, so that the velocity field is given by v ( r ) = v int + dv / dr ( r − R int ). c A constant mass-outflow rate is adopted, so that n OH × r × v is independent of r . d Representing either partial coverage by OH of the continuum source (a clumpy outflow, f < f > M scales as p f . f is not a fitting parameter, but indicates that the modeled source size is e ff ective. Nevertheless, we argue in Sect. 4 that f ∼ f & .
45 for the HVC. e For a given model, R int is fixed to give the correct absolute fluxes. f See Fig. 11. g Accelerating velocity fields have been tried as well, but they yield poor fits to both the line profiles and the flux ratios. h p f = R out in spherical symmetry, while p f = R int simulates an outflow directed mainly toward the observer. IR source (i.e. a clumpy outflow, f < f >
1, see below).The free parameters for each component are then R int , T dust , τ , R out / R int , v int , v out , N OH , p f / R out , and f , and are listed inTable 2. The data that constrain the fit are the line profiles andfluxes of the nine OH doublets. The line ratios essentially de-pend on T dust , N OH , and R out / R int , while the absolute fluxes alsodepend on f R . The radial column density of H nuclei in agiven component is N H = X − Z R out R int n OH ( r ) dr . (2)The mass-outflow rate per unit of solid angle is d ˙ Md Ω = f m H X − n OH ( R int ) R v int , (3)and the total mass-outflow rate is˙ M = π g ( p f ) d ˙ Md Ω , (4)where g ( p f ) ≤ g < p f < R out ), and is estimated inAppendix A. For reference, if f = X − n OH =
700 cm − at r =
70 pc, and v = − , then d ˙ M / d Ω ≈
90 M ⊙ yr − sr − . The momentum flux and the mechanical power, ˙ Mv and0 . Mv , are higher in this approach for the highest velocity gas.The sizes we report below ( R int , R out ) should be considerede ff ective. Results identical to a given model are obtained by scal-ing R int and R out to higher values as R / p f ( f <
1) while decreas-ing the densities as n H p f (i.e. keeping the same column density)and decreasing the continuum-subtracted spectra as f × F ν . Thiswould approximately simulate partial covering by the outflow ofthe far-IR source (covering factor f ). Conversely, the emergingprofiles can also be interpreted as produced by an ensemble of f clouds ( f >
1) each of radius R int / p f . In both cases, the mass-outflow rate scales as p f . A lower limit on f is set by the con-dition that the modeled far-IR continuum cannot exceed the ob-served level. We argue below (Sect. 4 and 3.3.1) for nearly com-plete covering ( f ∼ . −
1) for both the high-excitation quiescent component (QC) and the high-velocity component (HVC), andgive below all parameters ( R int , R out , n H ) for f =
1. For the low-excitation extended component (LEC) discussed below, f ≈ . R out ) andthus high mass-outflow rates. The need for several gas components is illustrated in Fig. 9for the OH119, OH84, and OH65 doublets. Specifically, theOH84 and OH65 line profiles (Fig. 2c-d) together with their ratio(Fig. 6) are primarily used to define the gas components whereOH is excited, while an additional low-excitation component isrequired to fully match the ground-state OH119 and OH79 dou-blets. One of our best-fit composite models is compared with allof the OH profiles in Fig. 10, where the red curves indicate thetotal absorption and emission as generated from all components. -The quiescent component (QC) : the spectra of the excitedlines and also the OH53.3 doublet indicate the presence of highlyexcited gas with the lines peaking at central velocities. Themodel for the QC is shown with blue curves in Figs. 9 and 10. -The high- and low-velocity components (HVC and LVC) : inour simplest model, most of the absorption in the blueshiftedline wing of the OH84 and OH65 excited doublets was simu-lated with a single outflowing component with a negative veloc-ity gradient, the HVC. This component is indicated with light-blue curves in Figs. 9 and 10 and, in addition to the wing inOH65 and OH84, it contributes significantly to the absorptionand emission in all other doublets, except for the high-excitationOH53 and OH56. Details of our best-fit model for the HVC arediscussed and characterized in more detail in Appendix A.The OH84 absorption at low velocities ( ∼
200 km s − ) is notfully reproduced with the HVC alone, and additional absorp- Fig. 9.
Illustration of the need for several OH components in Mrk 231, as inferred from the OH119, OH84, and OH65 doublets.In panels d-i, red curves indicate the absorption and emission by all considered components. a-c)
An outflow-free, high-excitationcomponent (QC) generates absorption in the high-lying OH lines (blue curves), but cannot account for the blue wings in the threedoublets or the redshifted emission in the OH119 transition. e-f)
The HVC (light-blue curves) and LVC (dashed light-blue curves)reproduce the blue wings in the OH85 and OH65 doublets, but fail to account for both the full blueshifted absorption and redshiftedemission in the ground-state OH119 doublet. g) A low-excitation component (LEC, green curves) is therefore required to matchthe ground-state OH119 (and also OH79) blueshifted absorption and redshifted emission (panel g). The composite fit to all lines isshown in Fig. 10.tion at these velocities is proposed to arise from another moreextended low-velocity component (LVC). The LVC contributesslightly to the reemission in OH119, OH79, OH84, and OH163. -The low-excitation component (LEC) : the joint absorptionand emission from the above components fails to account forthe total absorption and emission observed in the ground-stateOH119 and OH79 doublets, therefore the additional LEC com-ponent (green curves in Fig. 10) that produces significant ab-sorption in these doublets but not in excited lines is required.Because of the low excitation that this component represents, itis also the most extended one, presumably tracing the outflowregion probed at millimeter wavelengths by CO and HCN. Theparameters of this component are less constrained than those ofthe QC and HVC.The inferred model parameters for the well-constrained com-ponents are listed in Table 3. We also list in Table 4 the densities,hydrogen columns, and masses associated with the QC and theHVC, as well as the energetics that characterize the HVC.
The QC component is modeled as in G-A12; we adopt a “mixed”approach (i.e. the OH molecules and the dust are coexistent), andsimulate the line broadening with a microturbulence approach
Table 3.
Probable values of the parameters involved in the OHmodeling a . Parameter QC HVC LVC b R int (pc) c −
73 65 −
80 65 − T dust (K) 95 −
120 90 − ∼ τ − . − . . R out / R int − . . ∼ . − v int (km s − ) − ∼ v out (km s − ) − ∼ N OH (10 cm − ) 5 − d . − ∼ . p f / R oute ∼ . ∼ a Parameters for the LEC (low-excitation component) are not wellconstrained (see Sect. 3.4) and are omitted. b Uncertain parameters from the present data. c For f = d Column density per unit of τ (G-A12). e p f / R out = p f / R out < ( v tur =
90 km s − ). The line ratios depend on T dust , τ , and N OH /τ (G-A12), for which the explored ranges are 80 − Fig. 10.
Model fit for the OH doublets in Mrk 231. The blue, light-blue, dashed light-blue, and green curves show the contributionsby the quiescent component (QC), high-excitation outflow component (HVC), low-velocity component (LVC), and low-excitationcomponent (LEC), respectively. Red is the total emission and absorption due to all components. In this specific model, the parametersfor the HVC are T dust =
105 K, R int =
74 pc, R out / R int = . N OH = . × cm − , v int = − , v out =
100 km s − , and p f = . × R int , and the mass-outflow rate per unit of solid angle is d ˙ M / d Ω ≈
100 M ⊙ yr − sr − .K, 0 . −
3, and (2 − × cm − . In our best-fit model, theline ratios are reproduced with T dust =
110 K, τ = .
5, andan OH column of 8 × cm − per unit of τ (blue curves inFig. 10). For the above parameters, an e ff ective ( f =
1) radius of R out ≈
64 pc is obtained. Similar model fits are also obtained bydecreasing (increasing) T dust , and increasing (decreasing) both N OH /τ and R out ; the most plausible ranges are T dust = − N OH /τ = (16 − × cm − , and R out = −
55 pc,respectively (Table 3). For a hydrogen column of N H = × cm − per unit of τ (G-A12), N OH /τ = × cm − gives X OH = × − , in close agreement with our adopted value. Both T dust and N OH /τ are similar to those inferred in the nuclear re-gion of Arp 220 (G-A12). The total OH column along a radialpath (assuming uniform conditions) is ∼ × cm − . The con-tinuum optical depth is similar to that of the HVC (Sect. 3.3), τ = .
5, and corresponds to N H ∼ × cm − . Valuesof τ > Table 4.
Densities, column densities, masses, and energetics a Parameter QC HVC n H (10 cm − ) 1 − b . − . c N H (10 cm − ) 1 . − . − . M gas (10 M ⊙ ) 2 . − . . − . M (M ⊙ yr − ) − − P (10 g cm s − ) − ∼ − d , e L mech (10 L ⊙ ) − ∼ − d , e T mech (10 erg) − ∼ − ea Assuming X OH = . × − relative to H nuclei and f =
1. Valuesscale inversely with the OH abundance relative to this assumedvalue. Only the best-constrained components, the QC and the HVC,are considered (see Sect. 3.4 for the LEC). b Average density (the medium is probably clumpy). c The two values correspond to the highest and lowest outflowingvelocities (1700 and 100 km s − , respectively). d Varies with velocity; values are given for v = − . e Values are given for ˙ M =
850 M ⊙ yr − and R out / R int = . − . doublet at central velocities , while τ < f =
1) and most likelywarmer than the “warm component” needed to reproduce theH O lines detected in Mrk 231 with SPIRE (95 K and 120 pc,G-A10). Preliminary models for H O indicate, however, that theH O absorption lines are fitted with a model source size as smallas that of the QC. This behavior suggests that while the H O ab-sorption is primarily produced in front of the compact far-IR op-tically thick cores, e ffi cient H O submillimeter line emission isgenerated in the surrounding more extended regions with lowerextinction, but with a far-IR radiation density su ffi ciently high toproduce the observed excitation. The relationship between theH O absorption and emission lines will be explored in a futurestudy.To avoid too strong absorption at central velocities in theground OH119 and OH79 doublets, significant collisional ex-citation with a density of 2 . × cm − (much higher thanthe average density of (1 − × cm − , Table 4) was used.As discussed in Sect. 2.5, however, significant departures fromspherical symmetry (i.e. reemission from a flattened and possi-bly more extended structure viewed nearly face-on or with mod-erate inclination) could also account for the weakness of theselines. E ffi cient reemission at central velocities could take placein the surrounding region primarily responsible for the H O sub-millimeter emission, or at even the larger scales of the face-ondisk (Downes & Solomon, 1998). In any case, the above den-sity is e ff ective because collisional excitation with electrons andatomic hydrogen in an environment with a high-ionization frac-tion would additionally relax the required value. Nevertheless,the density cannot be lower than several × cm − , indicat-ing that the QC is clumpy. In these e ff ective models, collisionalexcitation primarily a ff ects the ground-state OH119 and OH79transitions. The H O absorption lines, however, favor τ ∼
3, as will be re-ported in a future work.
In our simplest approach, the observed absorption in the linewings of the excited OH84 and OH65 doublets at v < − − are simulated with a single outflow component, the HVC.It is characterized by a decelerating velocity field with v int = − and v out =
100 km s − , with hydrogen column den-sities in velocity intervals of 100 km s − , as shown in Fig. A.1a.We used two values for T dust ; 105 K, which is close to the valueused for the QC, and 90 K, closer to the value used for the warmcomponent in G-A10. A moderately high τ = . How extended is the outflowing gas in the HVC (as seen inOH) as compared with the source of far-IR emission that excitesthe OH? In Fig. 11a, the OH65-to-OH84 equivalent-width ratio(solid curves) is plotted as a function of the adopted R out / R int for two combinations of ( T dust , N OH ). The observed ratio ( ≈ . − .
5) can be reproduced either with ( T dust =
105 K, N OH = . × cm − ), or with ( T dust =
90 K, N OH = . × cm − ),as long as the thickness of the outflowing shell is small in com-parison to the radius of the far-IR source (i.e., R out / R int . . R out / R int > .
5) the OH be-comes less excited, the predicted OH65 / OH84 ratio drops, andhigher columns are then required. However, an extended HVCwould have an observable e ff ect on the line shapes. In spher-ical symmetry, one would expect an emission feature at red-shifted velocities (see Fig. 11c), arising from the limb of the far-IR source where the continuum optical depth is relatively low(for impact parameters p higher than R int , Fig. 8). This mod-eled emission feature, especially prominent in the OH84 dou-blet, is not seen in the spectra. Reemission in OH84 is not oc-curring at high velocities, that is, not in the HVC (for the LVC,see below), indicating that the projected surface where reemis-sion by the excited OH is generated is not significantly largerthan the surface where the absorption is produced. This suggeststhat either the HVC is compact around the optically thick far-IRcontinuum source, or that the outflow is collimated ( p f ∼ R int ,Fig. 8), flowing just toward (and possibly in the opposite direc-tion of) the observer. In an extended / collimated HVC, however,the line shapes would di ff er significantly from the observations.The covering factor as a function of the line-of-sight velocitywould have little contrast between moderately low and high ve-locities, thus predicting relatively flat blueshifted line wings thatare hardly compatible with the observed steep OH84 blueshiftedwing. The model fit for an extended outflow grossly overpre-dicts the OH84 absorption at ∼ − − relative to OH65(light-blue curve in Fig. 11b). While some degree of collimationis probably present (see Sect. 2.4 and below), the observed lineshapes and high OH excitation argue in favor of a componentwhere the high-velocity OH gas is piled up into a relatively nar-row region, tracing excited gas blowing out (along with the warmdust) from the warm far-IR continuum source against which wesee the OH absorption. We therefore favor R out / R int . . f = R int ∼ −
80 pc, and the outflow size (diameter) is ∼
200 pc. A lower limit, f & .
45, is set by the constraint thatthe continuum flux density at 30 − µ m, 7 . − . Fig. 11.
Model results for OH84 and OH65 in the HVC compo-nent. a) The ratio of the OH65-to-OH84 equivalent widths in theblueshifted wing as a function of the thickness of the outflowingshell, R out / R int . N OH and T dust are given by 1 . × cm − , 105K (green curve), and 3 . × cm − , 90 K (blue curve), yield-ing very similar results. Other model parameters are τ = . v int = − , and v out =
100 km s − . b) Correspondingmass-outflow rates per unit solid angle for full coverage of thefar-IR source ( f =
1, corresponding to R int =
75 pc; see alsoeq. A.1 and Table 3). The assumed OH abundance relative to Hnuclei is X OH = . × − . c) Observed (black histograms) andpredicted OH84 and OH65 line profiles ( N OH = . × cm − )for R out / R int = . . . . . × R int and an outflow size of up to ∼ T dust =
105 K, N OH = . × cm − ), and with ( T dust =
90 K, N OH = . × cm − ), il-lustrative of degeneracies in the models when constrained onlyby these two transitions. However, significant di ff erences be-tween the two models are seen especially in the ground-stateOH79 and OH53.3 doublets, which are overpredicted by the highcolumn-density solution. On the other hand, the high strengthof the OH85 doublet favors high OH columns (Sect. 3.6), sothat the column density is probably within the range N OH = (1 . − . × cm − .Even with a compact shell with R out / R int = .
3, our sphericalmodels overpredict the reemission at redshifted velocities in theOH84 doublet profile, so that p f / R out < ∼ . In most of the generated models for the HVC, the OH84 absorp-tion at low blueshifted velocities (200 −
300 km s − ) is underpre-dicted. Broadening of the absorption by the QC due to rotationof the circumnuclear structure (torus or thick disk) could accountfor some of this missing absorption. However, the responsiblegas is less excited than in the HVC, as little additional absorp-tion in OH65 is required for a good fit of the profile. Therefore,we tentatively associate this absorption with an increase of thecovering factor of the continuum by the OH at these velocities.Even if this additional low-velocity absorption is probably pro-duced by a spatial extension of the HVC in the plane of sky, withvelocities lower than predicted by the HVC, it is modeled in ourspherically symmetric models by means of a separate compo-nent, the LVC (dashed light-blue lines in Fig. 10). The LVC ismore extended than the HVC, generating some reemission inOH119, OH79, OH84, and OH163. In general, the inferred pa-rameters of the LVC are rather uncertain because the associatedabsorption overlaps with that produced by the HVC; we mod-eled it with R out =
150 pc and N OH = × cm − . This is aminor component, contributing little to the observed spectra andonly at low velocities. The OH65 / OH84 ratio in the blueshifted line wing is relativelyflat for v > −
900 km s − , and tends to increase (or at least remainsimilar) with higher velocity shifts. This dependence providesclues about the relative location of the gas at di ff erent velocitieswith respect to the source of excitation. If the OH excitation wereindependent of velocity, saturation of the OH84 doublet at lowvelocities (Fig. A.1c) would enhance the OH65 / OH84 ratio atthese velocities. This is contrary to the observed trend, suggest-ing that the OH gas with the highest velocity shift is more excitedthan the low-velocity outflowing gas. The increasing excitationwith increasing velocity shift is, in our models, generated by lo-cating the higher velocity gas closer to the far-IR exciting sourcethan the lower velocity gas (Fig. A.1b), thus suggesting an over-all decelerating velocity field. We also tried to model the HVCwith accelerated velocity flows, but found that the modeled lineshapes and line flux ratios were inconsistent with observations.In our models, the LVC is more extended and less excited thanthe HVC, supporting the same decelerating scenario. We note, however, that this solution relies on our simple spherical geom-etry (where the successive shells are concentric) and may notbe unique; for example, the high- and low-velocity gas may beflowing from di ff erent regions of a circumnuclear torus or disk,characterized by di ff erent T dust and possibly with di ff erent pro-jection e ff ects as the outflow widens. Nevertheless, some decel-eration is most likely taking place because CO and HCN, whichtrace larger regions, show wings up to a velocity of ∼
800 km s − from the line center (Feruglio et al., 2010; Cicone et al., 2012;Aalto et al., 2012), significantly lower than OH. Spoon & Holt(2009) also inferred a decelerating velocity field from the ion-ized gas outflows traced by the [NeII], [NeIII] and [NeV] lines ina sample of ULIRGs, though not in Mrk 231; the fine-structuremid-IR lines trace an outflow on a significantly smaller spatialscale, however.The very strong velocity gradient used in our model fits, withthe gas velocities varying from 1700 to 100 km s − in a relativelyshort path ( .
40 pc), may be indicative of high clumpiness andturbulence within the flow, but also favors a nonconcentric ori-gin of gas at di ff erent velocities. Nevertheless, strong shocks inswept-up gas of high density and column could in principle pro-duce a strong deceleration of the previously accelerated gas. Itis also worth noting that the LEC described below also indicatesthe presence of high-velocity gas (up to ∼
900 km s − ) detachedfrom the nuclear region, representing high-velocity gas that es-capes from the nuclear region along paths of least resistance. For N OH = . × cm − , R int =
70 pc, and R out / R int . .
5, the mass-outflow rate per solid angle in the direction ofthe observer (eq. A.1) associated with the HVC is d ˙ M / d Ω & p f (2 . × − / X OH ) M ⊙ yr − sr − . In spherical symmetry,this corresponds to ˙ M & g ( p f ) p f M ⊙ yr − , but we favora collimated outflow ( p f ∼ R int ) such that, for R out / R int = . g ( p f ) may be as low as ∼ . × M ⊙ yr − , andpossibly ∼ M ⊙ yr − , are inferred locally in the circumnuclearregion of Mrk 231 (Table 4). However, it is just the compact na-ture of the HVC gas that may suggest a non-steady flow, leavingopen the possibility of intermittency.In our prescription, the momentum flux in-creases with gas velocity and is given by ˙ P = . × ( ˙ M /
850 M ⊙ yr − )( v / km s − ) (2 . × − / X OH )g cm / s , or ∼ L AGN / c , adopting L AGN = . × L ⊙ (Veilleux et al., 2009). The corresponding mechanical luminos-ity is L mech ∼ × L ⊙ . The uncertainty in these parameters( ˙ M , ˙ P , and L mech ) is as high as a factor ∼ ff ects and the uncertainty in the OH abundance.While our estimates for the rates ( ˙ M , ˙ P , and L mech ) are roughlyconsistent with those inferred from CO by Feruglio et al. (2010),our integral values ( M gas and T mech ) are much lower due to thecompactness of the HVC. While the joint emission / absorption from the above three (QC,HVC, and LVC) components properly describes the observedabsorption in the excited doublets, the ground-state OH119 andOH79 lines remain underpredicted. An additional low-excitationcomponent (LEC) that accounts for the remaining OH119 andOH79 flux, but does not significantly contribute to the excitedOH doublets, was therefore included in the model. The LEC is Fig. 12. a)
OH119 and OH79 doublets after subtracting themodel for the QC + HVC + LVC, thus tentatively isolating thecontribution of the low-excitation component (LEC) to the ab-sorption and emission. The green curves show our simple spher-ically symmetric model for the LEC (Sect. 3.4). b) Inferred OHcolumn density of the LEC per unit of line-of-sight velocity in-terval across the blue absorption wing after correcting for thecovering factor at each velocity, but not corrected for the ree-mission in the lines (see text). The integral gives a total LECcolumn of N OH ≈ × cm − , in agreement with the detailedmodels.expected to be more spatially extended than the source of far-IR emission so that the OH molecules remain essentially in theground-state, and is also expected to be primarily responsible forthe emission features detected in OH119 and OH79 at redshiftedvelocities. Because this component is traced by the ground-statedoublets, no additional constraints on the spatial extent can beinferred from the OH data. Nevertheless, it is reasonable to as-sume that the LEC probes the relatively extended outflowingemission measured at millimeter wavelengths (Feruglio et al.,2010; Cicone et al., 2012; Aalto et al., 2012).Figure 12a shows the OH119 and OH79 profiles after sub-tracting the modeled emission of the QC + HVC + LVC (i.e. themodeled components for the excited OH). If our compositemodel for the excited OH is su ffi ciently accurate, the profiles inFig. 12a thus isolate the contribution by the LEC to the observedprofile. We note, however, that the emission at v ∼ −
200 km s − may still have a substantial contribution from circumnuclear gas.In OH119, even collisional excitation in a warm, dense regionmay take place at these moderately redshifted velocities wherethe CO 16-15 line appears to peak (Fig. 10h) . The LEC contri- The ground-state lines of OH + , CH + , and HF are all detected inemission (van der Werf et al., 2010), indicating the importance of colli-sional excitation in these ground transitions; the observed emission inOH119 at central velocities may also have a substantial contributionfrom collisionally excited gas in the same warm / dense region. 15onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH bution in Fig. 12a is thus tentative at low velocities. It is never-theless interesting that the OH119 / LEC shows a nearly symmet-ric line shape with an emission feature only ∼
20% weaker thanthe absorption feature, and with similar velocity extents on theblue and red sides. Within the model uncertainties and accord-ing to the discussion in Sect. 2.4, this result is consistent with aroughly spherical distribution of the LEC with negligible extinc-tion e ff ects at 119 µ m, indicating a wide opening angle of theflow at the corresponding spatial scales.Since detection of OH79 in the LEC ensures that the OH119doublet is optically thick, the absorption of the LEC OH119 nor-malized spectrum directly gives the covering factor at each line-of-sight blueshifted velocity ( f v = − F v / F c , where F v / F c is thecontinuum-normalized spectrum in Fig. 12a), uncorrected forthe reemission in the line. The OH column per unit of velocityinterval was estimated from the OH79 doublet (also uncorrectedfor the line reemission), and is shown in Fig. 12b. The integralof this spectrum gives N OH ≈ × cm − , in agreement withthe model for the LEC discussed below that accurately takes intoaccount the reemission in both doublets.Models for the LEC have significant degeneracies becauseof ( i ) the uncertainty in the shape and strength of the far-IR con-tinuum field as seen by the absorbing and emitting OH, and ( ii )the lack of constraints on the spatial scale. Our simple modelfor the LEC (green curves in Fig. 12a and Fig. 10a,b,e, and h)assume the following: ( i ) the LEC surrounds the whole sourceof far-IR emission, which is described by a spherical sourcewith R int =
490 pc, T dust =
55 K, and τ = . µ m; ( ii ) weadopted an external radius of R out =
800 pc (corresponding tothe ∼ e − level of the FWHM = . . Finding a reasonable match to the dou-blet shapes again requires a decelerating flow, with v int = − and v out =
200 km s − . The gas velocity fields as derivedfrom OH119 in other sources will be explored in a future work.The model fit in Fig. 12a uses N OH ≈ . × cm − (inclose agreement with Fig. 12b), p f = R out (i.e. strict sphericalsymmetry) and a covering factor of f = .
20 (as discussed inSect. 2.4). The latter value is significantly lower than the cover-ing factor of the compact HVC ( f & . n H ∼
30 cm − is too low to excite theCO 1 − M gas ∼ × × (2 . × − / X OH ) M ⊙ , and most of the mechanicalenergy, T mech ∼ × erg (compared with values of the HVCin Table 4). Eq. (A.1) gives ˙ M ∼ × (2 . × − / X OH ) M ⊙ yr − for the above parameters. Within the uncertainties in the analy-sis of OH and CO (Feruglio et al., 2010; Cicone et al., 2012), theenergetics inferred from both species appear to be consistent, es-pecially if the OH abundance drops below our adopted value atlarge distances from the circumnuclear region. In Fig. 13, the same model used to fit the far-IR OH linesobserved with Herschel / PACS (Fig. 10) is compared with the Note, however, that OH can potentially trace regions more extendedthan those traced by CO, because CO requires a minimum density to becollisionally excited while OH only needs the available far-IR radiationfield.
Fig. 13.
Model results for the OH35 doublet compared with theobserved Spitzer IRS spectrum. The model and color code arethe same as in Fig. 10.Spitzer IRS OH35 spectrum. While the absorption at central ve-locities is reproduced, the model appears to underpredict the ab-sorption at ∼ −
500 km s − as well as at redshifted velocities.Since the ground-state OH53.3 is reproduced, these discrepan-cies are most likely consequences of the uncertainties in themodeled continuum-flux density at 35 µ m relative to the fluxdensity at longer wavelengths, which is determined by the solidangle and T dust of the underlying continuum source. Specifically,we may expect a range of T dust behind the observed absorption,with the warmest and most compact components contributingsignificantly to the mid-IR continuum emission. OH One intringuing finding in the OH spectra of Mrk 231 is the rel-atively strong absorption by OH seen at 120, 85, and 66 µ m.While the OH120 doublet may be contaminated by CH + in itsblue component and the OH66 feature has a probable contri-bution by NH and H O + , the prominent OH85 is expected tobe free from contamination and shows evidence for absorptionby outflowing gas as well as by the QC component.In our model for the HVC with T dust =
105 K, we required N OH ∼ × cm − to generate the modeled blueshifted ab-sorption in the OH85 doublet (Fig. 10c). Likewise, N OH ∼ . × cm − was obtained for T dust =
90 K, correspondingto OH / OH ∼
20. Similarly, we required for the QC N OH ≈ × cm − per unit of τ , that is, OH / OH ∼
30. The over-abundance estimated for OH is then even more extreme thanwe previously reported (F10). Since we cannot exclude higherOH columns at moderate velocities in the outflow (because ofsaturation in OH84, Fig. A.1c), we favor OH / OH ∼ −
30 inboth components, with some indications that the ratio decreasesin the HVC. Models for the undetected OH were also per-formed, from which we estimate OH / OH & There are several spectral features that our modeling does not ac-count for. The high-velocity redshifted emission wing in OH79is poorly reproduced, and the redshifted emission feature inOH71 is ignored. The latter may be associated with outflowinggas more excited than modeled for the HVC.The OH163 is one of the most puzzling of the line shapes(Fig. 10h). The strength of the emission and absorption in thisdoublet is very sensitive to the continuum opacity. The ab- sorption at blueshifted velocities and the asymmetry betweenthe two lambda-doubling components indicate high continuumopacity and thus suggest a significant contribution by the HVC.However, the narrow linewidths of the emission features wouldsuggest an origin in low-velocity gas, but both the QC and theLEC predict line shapes broader than observed. The dip of emis-sion in between the two lambda-doubling components cannotbe reproduced. Since the OH163 doublet is pumped through ab-sorption of far-IR photons, part of the emission is most likelyarising from the same region that generates the submillimeterH O emission (G-A10), which is expected to surround the QC(see Sect. 3.2).
4. Discussion and conclusions
The picture that emerges from the OH observations and mod-els can be summarized as follows: a highly excited componentwhere OH peaks at central velocities, the QC, represents anoutflow-free circumnuclear component with T dust ∼
110 K, ane ff ective radius R ∼
65 pc, and a column of N H ∼ × cm − .The observed high-velocity absorption by excited OH arisesfrom a somewhat larger ( R int ∼
75 pc, R out ∼
100 pc, both ef-fective radii) radiatively excited and apparently collimated com-ponent (the HVC). This component is also associated with highfar-IR radiation density ( T dust ∼
100 K) and, given its somewhatlarger size, most likely surrounds the QC. This scenario suggeststhat the QC is feeding the outflow, in the sense that the outflow-ing gas emanates from the same circumnuclear structure that isresponsible for the central-velocity absorption. The OH columndensity in the HVC is N OH ≈ (1 . − × cm − , suggesting A v ∼
30 magnitudes of outflowing circumnuclear gas. We esti-mate a mass-outflow rate per unit of solid angle in the directionof the observer of at least ∼
70 and possibly ∼
100 M ⊙ yr − sr − for X OH = . × − . In spherical symmetry, this would corre-spond to ∼ ⊙ yr − , though significant departures from afully spherical model probably reduce the above value by a fac-tor ∼
2. The momentum flux attains ∼ L AGN / c . In our models,consisting of concentric shells of gas and dust with well-orderedradial motions, the high excitation found for the highest velocitygas was reproduced with a decelerating flow (see discussion inSect. 3.3.3). An extraordinary enhancement of OH was found(OH / OH .
30) in both the QC and the HVC.Our model for the excited OH leaves residuals in the ground-state OH119 and OH79 doublets, indicating the presence of alow-excitation component of the outflow (the LEC), with a col-umn of N OH ≈ × cm − . The LEC contribution to the pro-files (Fig. 12a) is only tentative at low redshifted velocities, butappears to show similar strengths and velocity extents for the ab-sorption and emission features in OH119. This suggests that theLEC is roughly spherical and spatially extended, in contrast withthe HVC. If the LEC is extended and surrounds the whole sourceof 119 µ m continuum emission, its covering factor is f ∼ f & / or collisionally ex-cited, the covering factor of the extended component must beeven lower than ∼ The QC has a modeled size (diameter of ∼
130 pc) remark-ably similar to that of the circumnuclear rotating structure (torusor thick disk) observed with the EVN in OH megamaser emis-sion by Kl¨ockner et al. (2003), which delineates the central re-gion of the OH megamaser complex (Richards et al., 2005)and traces the inner region of the radio / H I disk (Carilli et al.,1998) and of the star-forming region observed in the near-IR(Davies et al., 2004, 2007). From the (roughly) estimated con-tinuum optical depth ( τ ∼ .
5) and size, the gas mass of theQC is ∼ × × (0 . / X dust ) M ⊙ ( X dust is the dust-to-gasmass ratio) , significantly higher than the previously estimatedvirial mass (Kl¨ockner et al., 2003), but still roughly consistentwithin the uncertainties of both estimates. Furthermore, our in-ferred T dust ∼
110 K is not far from the value calculated via( L AGN / π R σ ) / ∼
130 K, where L AGN ∼ × L ⊙ (theactual T dust will be lower due to opacity e ff ects). We thus tenta-tively identify the QC component with the circumnuclear OH-megamaser torus (thick disk or oblate spheroid geometries areequally favorable). The match in sizes also suggests that f , thecovering factor (Table 2), is of order unity for the QC, thoughmore likely f ∼ . − . ff erent areasof the warm far-IR surface.The lack of OH119 absorption at central velocities may beindicative of high densities in the quiescent component, butcould also reflect scattering (i.e. reemission in the line) tak-ing place in a flattened structure seen nearly face-on or withlow inclination (Sect. 2.5). The geometric problem is prob-ably complex, because high resolution observations indicatea tilt of the torus (Kl¨ockner et al., 2003; Davies et al., 2004;Richards et al., 2005) relative to the outer nearly face-on disk(Downes & Solomon, 1998); nevertheless, the scattering processmay be operating in the region responsible for the H O submil-limeter emission (which is more extended than the QC, G-A10)or even on relatively large ( ∼ The HVC is likely to be emanating from, or is at least associatedwith, this torus, because the highly excited outflowing absorbing
OH is seen in front of, and is excited by a strong far-IR radia-tion field most likely generated in and around that circumnuclearcomponent. The QC could also provide a reservoir of gas richin OH that feeds the outflow, but if so, then either a relativelysmooth acceleration process (e.g. successive low-velocity, non-dissociative C-shocks) allows the OH to survive, or if the OH isdestroyed in fast (J-) shocks, it must reform in the post-shockgas. There is evidence for interaction between the radio jetand the surrounding gas (Ulvestad et al., 1999; Kl¨ockner et al.,2003; Rupke & Veilleux, 2011), as well as an overall (moderate)velocity blueshift of the torus or thick inner disk relative to thesurrounding gas at larger spatial scales (Kl¨ockner et al., 2003)that could indicate a slow expansion of the torus. According tooutflow models driven by radiation pressure (Roth et al., 2012), This gives a gas mass surface density of 2 × M ⊙ pc − , which is alower limit to the total value including stars (Davies et al., 2004, 2007). The transition from a C- to a J-type shock occurs at a critical veloc-ity of 2 . B / p πρ n (e.g. Ciolek et al., 2004), yielding <
100 km s − for B ∼ µ G (Carilli et al., 1998) and a density of 10 cm − , muchlower than the OH velocities. 17onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH the low outward velocity of the gas in the torus can be a result ofhigh inertia, the drop of the radiation pressure with decreasing T dust , and gravitation.The geometry of the inner outflowing gas (HVC) relative tothe torus may be more complex than simulated in our schematicspherically symmetric models. The possibility that the molecu-lar gas is primarily flowing along the polar regions of the torushas two drawbacks; first, the tilt of the torus implies that itsaxis deviates from the direction of the observer, with the con-sequent projection e ff ects on the line-of-sight velocity of thepolar gas. Second, the gas column along the polar direction isexpected to be relatively low, while our inferred high mass-outflow rate and the requirement of absorption of and excita-tion by optically thick 84 / µ m continuum indicate large gasreservoirs behind (and associated with) the outflowing gas. The3D radiation pressure models by Roth et al. (2012) predict thehighest di ff erential mass-outflow rates for polar angles > ◦ (their Figs. 12-14), that is, not far from the equatorial plane,and it is just the tilt of the torus that in this context would pro-vide a geometry favorable for detecting high di ff erential mass-outflow rates in the direction of the observer. Conceivably, theobserved OH outflow could probe an interclump medium ofthe torus itself that is flowing past the dense clumps (possiblyprobed by the QC), permeating the whole structure. Interactionwith the high-density clumps and shadowing e ff ects (Roth et al.,2012) would decelerate the outflowing gas with increasing ra-dial distance. The highest-velocity gas could also be tracing aconical transition region between the torus and the polar direc-tions. In our model for the HVC, the densities for velocities of500 − − are n H ∼ −
500 cm − , respectively,also in rough agreement with the wind-driven outflow modelsby Faucher-Gigu`ere & Quataert (2012). The high mass-outflow rate and outflow velocities derived fromthe far-IR observations of OH strongly point toward a key roleof the central AGN, as previously argued (S11). The momentumflux of ˙ P ∼ L AGN / c is roughly consistent with that required toregulate the growth of the black hole and set the M BH − σ relation(Debuhr et al., 2012). In the framework of radiation pressure, 3Dmodels indicate that in a clumpy disk with a wide opening an-gle, the radiation tends to escape along the poles and radiationpressure becomes less e ffi cient, generally accounting for a mo-mentum deposition rate of (1 − L AGN / c (Roth et al., 2012).Still, these models predict high di ff erential mass-outflow rates( d ˙ M / d Ω >
30 M ⊙ yr − sr − ) for su ffi ciently high columns andin directions close to the equatorial plane; a high scale-height ofthe torus / disk, or a relatively high mass-concentration in the po-lar region, could additionally increase the mass-outflow rate. Inaddition, fast energy-conserving AGN winds can do work on theswept-up (molecular) gas and then strongly boost the momen-tum flux (Faucher-Gigu`ere & Quataert, 2012). It is possible thatwhile radiation pressure a ff ects the whole circumnuclear struc-ture, winds are responsible for the highest velocity wings seenin OH. On the other hand, the high rate of mass loss derivedhere, together with the narrow-shell configuration favored forthe HVC component, may suggest an intermittent (explosive)instead of a steady flow, consistent with the multiple, expand-ing, concentric supershells seen in the optical / UV at larger scales(L´ıpari et al., 2005, 2009). OH and thecircumnuclear star formation An intriguing implication of the present observations is thestrong enhancement of OH in both the QC and the HVC.Since fractionation e ff ects do not chemically enhance OH(Langer et al., 1984), the OH / OH ratio is expected to bethe same as the O / O ratio. Our results indicate that Ois enhanced by about one order of magnitude relative to theGalactic Sgr B2 (Polehampton et al., 2005), and even morerelative to the solar value. This is of interest in the contextof the high metallicities that are observed in quasar environ-ments, whose enrichment is thought to be due to star for-mation with an IMF weighted toward massive stars (see re-view by Hamann et al., 2007). Similarly, O is thought to beenriched in the ISM by partial He burning in massive stars(e.g. Wilson & Matteucci, 1992; Henkel & Mauersberger, 1993;Prantzos et al., 1996; Wouterloot et al., 2008; Kobayashi et al.,2011), and the stars in the inner disk of Mrk 231 havebeen formed in situ (Davies et al., 2004). This is consistentwith the lack of detection of OH if O is primarily pro-duced in low- and intermediate-mass stars (Sage et al., 1991;Wilson & Matteucci, 1992) that, regardless of the IMF, arenot expected to have a significant chemical e ff ect on theISM of Mrk 231 (Muller et al., 2006) given the youth ofthe circumnuclear starburst, . .
25 Gyr (Davies et al., 2007).Artymowicz et al. (1993) and Collin & Zahn (1999) have pro-posed and explored a scenario in which unstable fragmentstrapped in the accretion disk of massive black holes, grow byaccretion to ∼ −
100 M ⊙ stars and ultimately explode as su-pernovae (SNe), to explain the metal enrichment of QSO ejecta,though it is unclear whether the O would be destroyed in theseconditions by He burning, yielding Ne (e.g. Prantzos et al.,1996). Interestingly, a very low (but not so extreme) O / O ∼
50 ratio, together with a high O / O ∼
12 ratio, are also in-ferred in the arm of a spiral galaxy at z = .
89 (Muller et al.,2006, 2011). In Mrk 231, OH is detected up to a velocity shiftof ∼ −
600 km s − , although OH enhancement at higher veloc-ities is not ruled out. While the high column density of OH inthe QC indicates previous O enrichment of the swept-up gas ,the possible relative enhancement of OH that we inferred in theline wing could suggest the contribution of SNe or massive stel-lar winds to the outflow. More studies of OH / OH in galaxiesand SNe are required to fully understand the evolutionary impli-cations of these enhancements.
Acknowledgements.
PACS has been developed by a consortium of institutes ledby MPE (Germany) and including UVIE (Austria); KU Leuven, CSL, IMEC(Belgium); CEA, LAM (France); MPIA (Germany); INAFIFSI / OAA / OAP / OAT,LENS, SISSA (Italy); IAC (Spain). This development has been supported bythe funding agencies BMVIT (Austria), ESA-PRODEX (Belgium), CEA / CNES(France), DLR (Germany), ASI / INAF (Italy), and CICYT / MCYT (Spain). E.G-A is a Research Associate at the Harvard-Smithsonian Center for Astrophysics,and thanks the Spanish Ministerio de Econom´ıa y Competitividad for supportunder projects AYA2010-21697-C05-0 and FIS2012-39162-C06-01. Basic re-search in IR astronomy at NRL is funded by the US ONR; J.F. and H.W.W.S.also acknowledge support from the NHSC. S.V. thanks NASA for partial supportof this research via Research Support Agreement RSA 1427277, support from aSenior NPP Award from NASA, and his host institution, the Goddard SpaceFlight Center, and acknowledges support from the Alexander von Humboldt Scaling the results by Davies et al. (2007) for an estimated SN rateof ∼ − within the inner ∼
500 pc (Davies et al., 2004), the cumu-lative ejected mass is ∼ × M ⊙ (including OB winds and AGBstars, Davies et al., 2007), which is similar to the current stellar mass(Davies et al., 2004) and also similar to the total gas mass ( ∼ . × M ⊙ , Downes & Solomon, 1998). If a significant fraction of these ejectastill remains bound, the circumnuclear ISM is expected from thesegrounds to be deeply recycled by the ejecta of high-mass stars.18onz´alez-Alfonso et al.: The Mrk 231 molecular outflow as seen in OH Foundation for a renewed visit to Germany following the original 2009 award.This research has made use of NASA’s Astrophysics Data System (ADS) and ofGILDAS software (http: // / IRAMFR / GILDAS).
References
Aalto, S., Garc´ıa-Burillo, S., Muller, S., et al. 2012, A&A, 537, 44Adams, F. C. 1991, ApJ, 382, 544Amari, S., Zinner, E., & Lewis, R. S. 1995, ApJ, 447, L147Armus, L., Charmandaris, V., & Bernard-Salas, J. 2007, ApJ, 656, 148Artymowicz, P., Lin, D. N. C., & Wampler, E. J. 1993, ApJ, 409, 592Baan, W. A., Haschick, A. D., & Henkel, C. 1989, ApJ, 346, 680Carilli, C. L., Wrobel, J. M., & Ulvestad, J. S. 1998, AJ, 115, 928Chevalier, R. A. 1977, ARA&A, 15, 175Cicone, C., Feruglio, C., Maiolino, R., et al. 2012, A&A, 543, 99Ciolek, G. E., Roberge, W. G., & Mouschovias, T. C. 2004, ApJ, 610, 781Collin, S. & Zahn, J.-P. 1999, A&A, 344, 433Davies, R. I., Tacconi, L. J., & Genzel, R. 2004, ApJ, 613, 781Davies, R. I., M¨uller S´anchez, F., Genzel, R., et al. 2007, ApJ, 671, 1388Debuhr, J., Quataert, E., & Ma, C.-P. 2012, MNRAS, 420, 2221Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604Downes, D., & Solomon, P. M. 1998, ApJ, 507, 615Farrah, D., Bernard-Salas, J., Spoon, H. W. W., et al. 2007, ApJ, 667, 149Faucher-Gigu`ere, C.-A. & Quataert, E. 2012, MNRAS, 425, 605Fischer, J., Sturm, E., Gonz´alez-Alfonso, et al. 2010, A&A, 518, L41 (F10)Feruglio, C., Maiolino, R., Piconcelli, E., et al. 2010, A&A, 518, L155Gonz´alez-Alfonso, E., & Cernicharo, J. 1997, A&A, 322, 938Gonz´alez-Alfonso, E., Cernicharo, J., van Dishoeck, E. F., Wright, C. M., &Heras, A. 1998, ApJ, 502, L169Gonz´alez-Alfonso, E., & Cernicharo, J. 1999, ApJ, 525, 845Gonz´alez-Alfonso, E., Smith, H. A., Ashby, et al. 2008, ApJ, 675, 303Gonz´alez-Alfonso, E., Fischer, J., Isaak, K., et al. 2010, A&A, 518, L43Gonz´alez-Alfonso, E., Fischer, J., Graci´a-Carpio, J., et al. 2012, A&A, 541, A4(G-A12)Gonz´alez-Alfonso, E., Fischer, J., Bruderer, S., et al. 2013, A&A, 550, A25 (G-A13)Hamann, F., Warner, C., Dietrich, M., & Ferland, G. 2007, ASP ConferenceSeries, 373, 653Henkel, C., & Mauersberger, R. 1993, A&A, 274, 730Kl¨ockner, H.-R., Baan, W. A., & Garrett, M. A., 2003, Nature, 421, 821Kobayashi, C., Karakas, A. I., & Umeda, H. 2011, MNRAS, 414, 3231Langer, W. D., Graedel, T. E., Frerking, M. A., & Armentrout, P. B. 1984, ApJ,277, 581L´ıpari, S., Terlevich, R., Zheng, W., Garc´ıa-Lorenzo, B., S´anchez, S. F., &Bergmann, M. 2005, MNRAS, 360, 416L´ıpari, S., S´anchez, S. F., Bergmann, M., et al. 2009, MNRAS, 392, 1295Meijerink, R., Spaans, M., Loenen, A. F., & van der Werf, P. P. 2011, A&A, 525,A119Muller, S., Gu´elin, M., Dumke, M., Lucas, R., & Combes, F. 2006, A&A, 458,417Muller, S., Beelen, A., Gu´elin, M., et al. 2011, A&A, 535, 103Murray, N., Quataert, E., & Thompson, T. A. 2005, ApJ, 618, 569M¨uller, H. S. P., Thorwirth, S., Roth, D. A., & Winnewisser, G. 2001, A&A, 370,L49M¨uller, H. S. P., Schl¨oder, F., Stutzki, J., & Winnewisser, G. 2005, J. Mol. Struct.742, 215O ff er, A. R., & van Hemert, M. C. 1992, Chem. Phys. 163, 83Pickett, H. M., Poynter, R. L., Cohen, E. A., Delitsky, M. L., Pearson, J. C., &M¨uller, H. S. P. 1998, JQSRT, 60, 883Pilbratt, G. L., Riedinger, J. R., Passvogel, T., et al. 2010, A&A, 518, L1Poglitsch, A., Waelkens, C., Geis, N., et al. 2010, A&A, 518, L2Polehampton, E. T., Brown, J. M., Swinyard, B. M., & Baluteau, J.-P. 2003,A&A, 406, L47Polehampton, E. T., Baluteau, J. -P., & Swinyard, B. M. 2005, A&A, 437, 957Prantzos, N., Aubert, O., & Audouze, J. 1996, A&A, 309, 760Richards, A. M. S., Knapen, J. H., Yates, J. A., et al. 2005, MNRAS, 364, 353Roth, N., Kasen, D., Hopkins, P. F., & Quataert, E. 2012, ApJ, 759, 36Rupke, D. S. N. & Veilleux, S. 2011, ApJ, 729, L27Sage, L. J., Henkel, C., & Mauersberger, R. 1991, A&A, 249, 31Sakamoto, K., Aalto, S., Wilner, D. J., et al. 2009, ApJ, 700, L104Spoon, H. W. W., & Holt, J. 2009, ApJ, 702, L42Sturm, E., Gonz´alez-Alfonso, E., Veilleux, S., et al. 2011, ApJ, 733, L16 (S11)Ulvestad, J. S., Wrobel, J. M., & Carilli, C. L. 1999, ApJ, 516, 127van der Werf, P. P., Isaak, K. G., Meijerink, R., et al. 2010, A&A, 518, L42Veilleux, S., Cecil, G., & Bland-Hawthorn, J. 2005, ARA&A, 43, 769Veilleux, S., Rupke, D. S. N., Kim, D.-C., et al. 2009, ApJS, 182, 628Walter, F., Weiss, A., & Scoville, N. 2002, ApJ, 580, L21 Wilson, T. L., & Matteucci, F. 1992, A&AR, 4, 1Wouterloot, J. G. A., Henkel, C., Brand, J., & Davis, G. R. 2008, A&A, 487, 237 Appendix A: Some properties of the OH outflowmodels
In our models, which assume a constant mass-outflow rate andvelocity gradient within R int and R out , the relationship betweenthe mass-outflow rate per unit of solid angle ( d ˙ M / d Ω ) and N OH is given by (from eqs. 2 and 3) d ˙ M / d Ω = f m H X − v int R int N OH × x − v out / v int x − ! × " − x + v int / v out − v int x / v out − v int xv out ! − , (A.1)where x ≡ R out / R int . For given N OH , R int , v int , and v out , both ˙ M and the OH excitation increase with decreasing x (Fig. 11a), sothat higher excitation (e.g. higher OH65 / OH84 ratio) implies amore compact outflow and an increasing ˙ M in our models.The OH column density per unit of velocity interval is dN OH dv = ˙ M X OH R int π g ( p f ) f m H × x − | v out − v int | × r v ( r ) . (A.2)The corresponding N H spectrum, calculated in velocity intervalsof 100 km s − , is shown in Fig. A.1a for the model of the HVCdisplayed in Fig. 10a. In a compact decelerating outflow, the col-umn density remains nearly constant for high velocities. The vi-sual extinction at high velocities in these 100 km s − intervals isexpected to be A V ∼ N OH and of the OH excitation, and on the covering factor asa function of the line-of-sight velocity. The increasing excitationwith increasing velocity shifts is obtained in our models with adecelerating field (Fig. A.1b). The calculated OH84 and OH65optical depths along a radial path are shown in Fig. A.1c, indi-cating saturation e ff ects in the OH84 doublet mostly at moderatevelocities, but optically thin absorption in OH65. On the otherhand, the steep decrease of the OH84 absorption with increas-ing velocity shift (Fig. 10) is indicative of a decreasing coveringfactor with increasing projected velocity, as shown in Fig. A.1d.At low projected velocities ( <
400 km s − ), the covering factorexceeds unity, which generates reemission from the limb of theoutflow at significantly redshifted velocities. Since this reemis-sion is not observed in OH84, a collimated ( p f ∼ R int ) outflow isfavored.The total mass-outflow rate is given by˙ M = π g ( p f ) d ˙ M / d Ω , (A.3)where g = p f = R out . For R int ≤ p f < R out , we roughlyapproximate the geometry depicted in Fig. 8 as two cones, eachone with half opening angle sin θ / = p f / R out , and thus g ( p f ) = − q − ( p f / R out ) . (A.4)This approximation underestimates ˙ M because the model stillincludes the contribution by gas outflowing along the plane ofsky (Fig. 8). Fig. A.1.
Details of the model for the HVC shown in Fig. 10 withlight-blue curves; note that higher velocities correspond to lowerdistances to the far-IR exciting source in a decelerating field. a) Column density of H nuclei (assuming X OH = . × − ) inintervals of 100 km s − as a function of the radial velocity for R out / R int = . N OH = . × cm − , v int = − ,and v out =
100 km s − . b) The rotational temperature of the Π / J = / radial velocity. c) The OH84 and OH65 optical depths along a ray passing throughthe center, and d) the covering factor of the continuum, both asa function of the line-of-sight velocity. At low projected veloc-ities ( <
400 km s − ), the covering factor exceeds unity, whichgenerates reemission from the limb of the outflow.), the covering factor exceeds unity, whichgenerates reemission from the limb of the outflow.