Extremely broad Lyman-alpha line emission from the molecular intergalactic medium in Stephan's Quintet: evidence for a turbulent cascade in a highly clumpy multi-phase medium?
P. Guillard, P. N Appleton, F. Boulanger, J. M. Shull, M. D. Lehnert, G. Pineau des Forets, E. Falgarone, M.E. Cluver, C.K. Xu, S.C. Gallagher, P.A. Duc
DD raft version F ebruary
16, 2021Typeset using L A TEX twocolumn style in AASTeX63
Extremely broad Ly α line emission from the molecular intergalactic medium in Stephan’s Quintet: evidence for aturbulent cascade in a highly clumpy multi-phase medium? P. G uillard , P. N A ppleton , F. B oulanger , J. M. S hull , M. D. L ehnert , G. P ineau des F orets , E. F algarone , M.E. C luver , C.K. X u , S.C. G allagher , and P.A. D uc Sorbonne Universit´e, CNRS, UMR 7095, Institut d’Astrophysique de Paris, 98bis bd Arago, 75014 Paris, France Institut Universitaire de France, Minist`ere de l’Enseignement Sup´erieur et de la Recherche, 1 rue Descartes, 75231 Paris Cedex 05, France IPAC, Caltech, MC 100-22, 1200 E. California Blvd., Pasadena, CA 91125, USA Laboratoire de Physique de l’Ecole Normale Sup´erieure, ENS, Universit´e PSL, CNRS, Sorbonne Universit´e, Universit´e de Paris, 75005 Paris, France CASA, Astrophysical and Planetary Sciences Dept., University of Colorado, UCB-389, Boulder, CO 80309, USA Observatoire de Paris, PSL University, Sorbonne Universit´e, LERMA, 75014 Paris, France Universit´e Paris Saclay, CNRS, Institut d’Astrophysique Spatiale, 91405 Orsay, France Centre for Astrophysics and Supercomputing, Swinburne University of Technology, John Street, Hawthorn 3122, Victoria, Australia Department of Physics and Astronomy, University of the Western Cape, Robert Sobukwe Road, Bellville, South Africa Chinese Academy of Sciences South America Center for Astronomy, National Astronomical Observatories, CAS, Beijing 100101, China Department of Physics and Astronomy, University of Western Ontario, London, ON N6A 3K7, Canada Laboratoire AIM, CEA / DSM - CNRS - Universit´e Paris Diderot, DAPNIA / Service d’Astrophysique, CEA / Saclay, F-91191 Gif-sur-Yvette Cedex, France (Received February, 2021; Revised ?; Accepted ?)
Submitted to ApJABSTRACTWe present
Hubble Space Telescope
Cosmic Origin Spectrograph (COS) UV line spectroscopy and integral-field unit (IFU) observations of the intergalactic medium (IGM) in the Stephan’s Quintet (SQ) galaxy group. SQhosts a 30 kpc long shocked ridge triggered by a galaxy collision at a relative velocity of 1000 km s − , wherelarge amounts of cold (10-100 K) and warm (100-5000 K) molecular gas coexist with a hot, X-ray emittingplasma. COS spectroscopy along five lines-of-sight, probing 1 kpc-diameter regions in the IGM, reveals verybroad (FWHM ≈ − − ) and powerful Ly α line emission with complex line shapes. By stackingthe spectra at the five positions, we also detect the C iv λλ α line profiles are oftensimilar to, or sometimes much broader than line profiles obtained in H β , [C ii ]157.7 µ m and CO (1-0) emissionalong the same lines-of-sight. In these cases, the extreme breadth of the Ly α emission, compared with H β ,implies resonance scattering within the observed structure. Line ratios of Ly α / H β for the two COS pointingsclosest to the center of the shocked ridge are close to the Case B recombination value, suggesting that at thesepositions Ly α photons escape through scattering in a low density medium free of dust. Some Ly α spectrashow suppressed velocity components compared with [C ii ] and H β , implying that some of the Ly α photons areabsorbed. Scattering indicates that the neutral gas of the IGM is clumpy, with multiple clumps along a givenline of sight. Remarkably, over more than four orders of magnitude in temperature, the powers radiated by themulti-phase IGM in X-rays, Ly α , H , [C ii ] are comparable within a factor of a few. We suggest that both shocksand mixing layers co-exist and contribute to the energy dissipation associated with a turbulent energy cascade.This may be important for the cooling of gas at higher redshifts, where the metal content is lower than in thislocal system, and a high amplitude of turbulence more common. Keywords: galaxies: high-redshift – galaxies: formation and evolution – galaxies: kinematics and dynamics –galaxies: ISM – galaxies: active – ISM: general – ISM: structure – turbulence INTRODUCTIONGalaxy interactions are important phases of galaxy evolu-tion, often involving high-speed shocks and huge amountsof kinetic energy. Many of these interactions are observed in infrared (IR) and visible light to trigger bursts of star-formation. The dissipation of kinetic energy a ff ects the gascooling and how, when, and where star formation proceeds.To make headway in our understanding of the conversion of a r X i v : . [ a s t r o - ph . GA ] F e b G uillard , A ppleton , B oulanger , S hull et al .molecular gas to stars, it is crucial to determine the mecha-nism and rate of gas cooling.Stephan’s Quintet (HCG 92, hereafter SQ) is a compactgroup of five interacting galaxies (Arp 1973) with a com-plex dynamical history, involving multiple galaxy collisions(Moles et al. 1997; Sulentic et al. 2001; Renaud et al. 2010;Hwang et al. 2012). It is an ideal laboratory for the study ofgalaxy interactions and their impact on the physical state andenergetics of the gas, especially the dissipation of merger-driven turbulence isolated against the dark sky (Guillardet al. 2009). It is one of the few extragalactic sourceswhere one can spatially separate star forming regions frompure shocked gas. When excluding the large foregrounddwarf galaxy NGC 7320, the main group is dominated byfour large massive galaxies. Three of them, NGC 7317,NGC 7318a and NGC 7319, have heliocentric radial veloci-ties in the range V helio = − − , and togetherthey define the main barycentric velocity of the group (ataround 6600 km s − ). A fourth galaxy, NGC 7318b (V helio = − , is often described as an intruder galaxy be-cause it appears to be colliding with gas in the main group ata relative velocity of ≈ − (see Fig 1 and Xu et al.2003).The collision of NGC 7318b with the IGM is believed tobe responsible for a striking feature of the group, namely agalaxy-wide shocked ridge ( ≈ ×
35 kpc ; Fig. 1) seen atmany wavelengths. A ridge of X-ray (O’Sullivan et al. 2009)and radio synchrotron (Allen & Hartsuiker 1972) emissionfrom the hot (6 × K) post-shock plasma is associatedwith the group-wide shock. Surprisingly,
Spitzer
IRS spec-troscopy revealed that the mid-IR spectrum in the IGM isdominated by the rotational line emission of H (see red con-tours on Fig. 1), although star formation is very weak (Ap-pleton et al. 2006; Cluver et al. 2010; Appleton et al. 2017).The weakness of the mid-IR star-forming indicators (dustfeatures or ionized gas lines) relative to the H lines suggeststhat, despite the large H mass estimated to be ≈ × M (cid:12) (Guillard et al. 2012a), the star formation rate is on aver-age very low in the shock ( < .
07 M (cid:12) yr − , Cluver et al.2010). This is a factor 40 below the star formation rate ex-pected from the Schmidt-Kennicutt relation (Guillard et al.2012a). Deep HST / WFC3 H α imaging of SQ reveals manycompact H α knots spread over the shock, as well as a dif-fuse, underlying component (Gallagher et al. 2001). Ex-tensive Gemini optical spectroscopy of 50 H α knots in thegroup (Konstantopoulos et al. 2014) shows that both pho-toionization and shock excitation are present (see also Xuet al. 2003, for earlier long-slit optical spectroscopy), a re-sult that is also confirmed by IFU spectroscopy (Duarte Puer-tas et al. 2019). Some knots are star clusters (masses from10 to a few 10 M (cid:12) , Gallagher et al. 2001; Fedotov et al.2011), while others show very strong [O iii ] λ i ] λ ≈
700 km s − ), complex optical emission line profiles, consis-tent with pure shock excitation (Konstantopoulos et al. 2014).We assume a distance to Stephan’s Quintet of 94 Mpc forH =
70 km s − Mpc − , and an assume group systemic helio-centric velocity of 6600 km s − . OBSERVATIONS AND DATA REDUCTION2.1.
HST COS spectroscopy
SQ was observed with the medium-resolution far-UVG130M and G160M gratings of
HST -COS on 2014 Novem-ber 9 for a total of 15 orbits. Descriptions of the COS instru-ment and on-orbit performance characteristics can be foundin (Green et al. 2012; Osterman et al. 2011), as well as inthe COS Instrument Handbook. In order to achieve contin-uous spectral coverage across the G130M bandpass (1135-1440 Å) and to minimise fixed-pattern noise, we made ob-servations at two central wavelength settings (1291 Å and1300 Å) with four focal-plane o ff set locations in each grat-ing setting (i.e. FP-POS =
1, 2, 3, 4). This combinationof grating settings ensures the highest signal-to-noise obser-vations at the shortest wavelengths available to the G130Mmode at a resolving power of R = λ/ ∆ λ ) ≈ ,
000 (about17 km s − velocity resolution). For every sightline, COSobservations yielded a continuous spectrum spanning λ ≈ − / N) of 7–15 per resolution element(FWHM =
15 km s − ) at λ ≈ CalCOS , down-loaded from MAST. First, to remove most of the geocoro-nal airglow lines, we have filtered the data in time in orderto utilize only data taken during orbital night time. To dothat, we used the TimeFilter python module to select daytime data in the COS corrtag files, and then re-extracted thedata into x1d spectra. The data have then been aligned andcoadded with the COADD X1D.pro V3.3 IDL routine pro-vided by STScI . To take into account the strong wings ofthe Line Spread Function (LSF) of COS, we have used theCOS LSF.pro IDL routine to produce a LSF model at thenearest tabulated wavelength value. The absolute flux cali-bration steps are described in detail in the COS data hand-book , and are expected to of the order of ±
5% for G130Mand G160M (errors are dominated by fixed pattern noise in see HST COS Instrument Handbook for more details: http: // / hst / cos / documents / handbooks / current / cos cover.html. https: // justincely.github.io / AAS224 / timefilter tutorial.html available at http: // casa.colorado.edu / ∼ danforth / science / cos / costools https: // hst-docs.stsci.edu / cosdhb / chapter-3-cos-calibration road L y α line emission from the S tephan ’ s Q uintet shock Visible light IFU spectroscopy
We also present Integral Field Unit (IFU) observationsmade with the George and Cynthia Mitchell Spectrograph(hereafter GCMS, formerly known as VIRUS-P) mounted onthe 2.7 m Harlan J. Smith Telescope at McDonald Obser-vatory (Hill et al. 2008; Blanc 2013). The IFU uses a 246fiber bundle, with each fiber covering 4 . (cid:48)(cid:48)
16 on the sky, whichmakes it sensitive to faint extended emission. We used a 3-point dither pattern to completely cover the 2.82 arcmin fieldof view, and to fill in gaps between the fibers. Observationswere obtained on 1 Oct 2011 with an integration time of onehour at each of the three dither positions. We used the VP2blue grating, which has a spectral resolution of ∼ . ∼
100 km s − ) and covers the wavelength range ∼ ∼ EXTREMELY BROAD Ly α LINE AND KINEMATICSOF THE IONIZED MEDIUM3.1.
Individual COS pointings: line fluxes and kinematics
We detect strong and broad Ly α (1215.67 Å) emissionfrom all the five observed regions shown in Fig. 1. TheLy α line profiles for each of the 5 observed positions, shownin Fig.2a-e (black solid lines), have complex, multi-peakedprofiles. The observed coordinates and integrated Ly α andmolecular hydrogen (H ) line fluxes for each region are pre-sented in Table 2. Our observations involve only a few COSsight-lines and therefore provide a sparse view of the Ly α emission across the IGM. It is therefore not possible to quan-tify the total Ly α luminosity from the IGM, and we will re-strict our analysis to the comparison of line fluxes at the COSpositions.The Ly α flux averaged over the 5 COS beams amounts to ≈
40% of the warm H IR line emission (see last columnof Table 1), which is the dominant cooling channel in theshocked IGM (Appleton et al. 2017) and similar to that ofthe [C ii ] line and X-rays as well (see Table 1 in Guillardet al. 2009, for a summary of the energy budget across gasphases). Remarkably, on average, the Ly α line luminosity iscomparable to that of much cooler and hotter gas. Sect. 5.3discusses the implications of this observational result on theproperties of kinetic energy dissipation in the IGM of SQ. Wealso note that the Ly α flux varies by a factor of ≈
10, betweenthe bridge (faintest) and the ridge 1 (brightest) positions.
Table 1.
COS observation log: MAST archive name, positions, COS gratings,central wavelength and individual exposure times for each of the 5 positionsobserved.COS target RA Dec Grating λ cen Exp. time(MAST) (J2000) (J2000) [Å] [seconds]HCG92-1 22 35 59.765 +
33 58 21.33 G130M 1096 1287.872G130M 1222 1437.696G160M 1611 3180.512G160M 1623 3673.536HCG92-2 22 36 00.032 +
33 58 06.75 G130M 1096 1287.936G130M 1222 1437.760G160M 1611 3180.544G160M 1623 3673.568HCG92-3 22 35 59.439 +
33 58 34.80 G130M 1096 1429.056G130M 1222 1447.840G160M 1611 3180.576G160M 1623 3673.568HCG92-5 22 36 01.222 +
33 58 22.74 G130M 1096 1366.016G130M 1222 1359.552G160M 1611 3180.640G160M 1623 3671.584HCG92-7 22 35 58.953 +
33 58 49.96 G130M 1096 600.000G130M 1222 599.712G160M 1611 1102.656G160M 1623 1303.712
In Table 3, we gather our measurements of the widths ofthe Ly α lines for the five observed positions, as well as for thestacked spectrum. There is a large variation of the widths ofthe Ly α lines, with Full Width at Zero Intensity (FWZI) up to ≈ − . This is remarkable, given that the COS beamis sampling a small region of intergalactic space betweenthe galaxies (1.1 kpc) . This suggest that the COS beamslikely probe the same large-scale organised structure in thefilament seen at other wavelengths, rather than small individ-ual emission regions. This becomes clear when we comparethe Ly α profiles with lines observed in [C ii ]157.7 µ m (Ap-pleton et al. 2013) and H β (solid and dashed red lines re-spectively). In many cases, the complex Ly α profile shapestrack approximately the main kinematic features from theother lines which are known to show large-scale coherence(see for example Rodr´ıguez-Baras et al. 2014; Duarte Puer-tas et al. 2019).Fig. 2 shows the varying velocity, profile shapes, linewidths, and strength of the Ly α emission. The broadest Ly α Stephan’s Quintet is assumed to be at a distance of 94 Mpc G uillard , A ppleton , B oulanger , S hull et al . Figure 1.
HST
Wide Field Camera 3 F665N image of the inner SQ group with
Spitzer
IRS H (0-0) S(3) 9.6 µ m line flux contours in redfrom Cluver et al. (2010). This emission from warm molecular hydrogen highlights the North-South shocked ridge of the IGM, as well as anextension towards NGC 7319, called ”the bridge”. The blue circles are the COS apertures (2.5” in diameter) corresponding to 1.1 kpc at adistance D =
94 Mpc. The positions, observing parameters and integration times are listed in Table 1. line we observed is seen in Fig. 2a, obtained at HCG92-1(see Fig. 1, and Table 3), near the center of the giant H emitting filament. In this profile, the Ly α emission tracksquite well the shape of the H β and [C ii ] profiles at low he-liocentric velocity, but deviates strongly at higher velocities.The Ly α emission extends to at least V helio = − ,whereas both the [C ii ] and H β emission fall almost to zeroflux at velocities of no more than 7300 km s − , compared tothe barycentric systemic velocity of 6600 km s − . The excessemission seen above 7300 km s − may be evidence of reso-nant scattering and bulk motions of the scattering medium, which is common in Ly α systems (see Sect. 5.2 for a dis-cussion of the origin of the line broadening). This is sup-ported by the observation that the line profiles, taken withboth the Mitchell Spectrograph and Herschel generally oc-cupy a more limited range of radial velocities compared withthe Ly α emission, despite being taken over larger beam sam-pling areas than the COS data (4 × and (for the[C ii ]157.7 µ m) (9.4 × ; see Appleton et al. 2013).Other examples of possible resonant scattering wings inthe Ly α profiles compared with the [C ii ] and H β emissionlines are the blue wing seen in HCG-2 (Fig. 2b) at V helio < road L y α line emission from the S tephan ’ s Q uintet shock Table 2.
Observed Ly α , H β and H line fluxes and line ratios for the COS lines of sight.COS target F Ly α : Ly α Flux a F H β : H β flux b F H : H flux c F Ly α / F H β F Ly α / F H [10 − W m − ] [10 − W m − ] [10 − W m − ]HCG92-1 (ridge) 5 . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . a Uncorrected for Ly α absorption. Some of the profiles show evidence of absorption (see text). b Sum of the observed line flux estimated on a 4 × square aperture from the Mitchell Spec-trograph IFU data, scaled to the circular COS aperture of 2.5” in diameter assuming constant surfacebrightness. c Sum of the (0-0) S(0) + (0-0) S(1) + (0-0) S(2) + (0-0) S(3) pure rotational H line fluxes derivedfrom extractions of 5 . × . from Spitzer
IRS data (Appleton et al. 2017), scaled down to thecircular COS aperture of 2.5” in diameter assuming constant surface brightness.
Table 3.
Observed Ly α line kinematical properties for theCOS lines of sight.COS target W (Ly α ) a W (Ly α ) a W (Ly α ) a [km s − ] [km s − ] [km s − ]HCG92-1 1910 ±
40 1690 ±
20 1290 ± ±
40 1080 ±
20 690 ± ±
40 1770 ±
20 570 ± ±
40 930 ±
40 350 ± ±
40 530 ±
40 220 ± ±
40 1820 ±
20 595 ± a Widths at 10%, 20% and 50% of maximum flux.N ote — W has been computed after smoothing the spec-tra to 80 km s − to increase the SNR in the line wings,and is close to the Full Width at Zero Intensity of theline. In case of the presence of several velocity compo-nents, only the widest component is listed. − and the blue wing in HCG-5 (Fig. 2d). In bothcases, the Ly α emission extends significantly blueward of theH β line by velocities of up to 200-300 km s − .3.2. Comparison with CO (1-0) line profiles
Our CO (1-0) observations with the Plateau de Bure Inter-ferometer (PdBI, now called NOEMA), which will be pre-sented in a companion paper (Guillard et al. in prep.), showgiant, kpc-scale molecular complexes of a few 10 M (cid:12) in theshock, some of them associated with H α -emitting regions. From the cleaned CO (1-0) line emission map, we have ex-tracted spectra within beams centered on the COS positions,using a synthesized beam of 4 . × . , of position an-gle P.A. = ◦ . We used the IRAM GILDAS mapping suiteof routines to perform the extractions, and then exported thespectra into fits files. The line parameters are gathered inTable 2. The CO (1-0) line intensity was computed by in-tegrating the line profile, and the central velocity and linevelocity dispersion were computed after a Gaussian fit to theprofile. The comparison between the CO (1-0) and Ly α lineprofiles is shown in Fig. 3, with the caveat that the NOEMAbeam is twice as large as the COS beam. It is striking thatthe CO lines detected with NOEMA are much narrower thanthe Ly α lines, except in the Northern star-forming region(SQ-A, position 7) where the main CO velocity componentat 6700 km s − matches the Ly α profile. Those molecularcomplexes are much larger than the small scale structure ofthe neutral gas through which Ly α photons scatter out of theIGM. However, we note that there may be a more di ff use,extended molecular component, filtered-out by the interfer-ometer, since single-dish IRAM 30m data exhibit very broadCO line profiles (Guillard et al. 2012a). Those giant molec-ular complexes could well break down into much smallerclumps with sometimes large shear motions between them( ≈
100 km s − ).3.3. Overview of the Large-scale motions of the IGM gas
Part of the complexity of the Ly α line profiles can beunderstood when the full picture of the ionized gas (as https: // / IRAMFR / GILDAS / doc / pdf / map.pdf G uillard , A ppleton , B oulanger , S hull et al . Table 4.
CO (1-0) line properties extracted at the positions of the COS apertures and optical line properties derived from Gemini spectroscopyat the nearest position of the HST COS apertures.COS target I CO a v CO a σ CO a Gemini Id b F(H α ) c [N ii ] / H α c [O i ] / H α c H β / H α c [Jy km s − ] [km s − ] [km s − ] [10 − W m − ]HCG92-1 (ridge) 2 . ± . ∗ . ± . . ± .
001 0 . ± .
002 0 . ± . . ± . . ± . . ± .
001 0 . ± .
002 0 . ± . . ± . ∗ . ± . . ± .
001 0 . ± .
001 0 . ± . . ± . . ± . . ± .
002 0 . ± .
001 0 . ± . . ± . . ± . . ± .
001 0 . ± .
001 0 . ± . a Parameters estimated from the CO (1-0) spectrum extracted from the IRAM NOEMA interferometer data (Guillard et al. in prep.) on a beamsize 4 . × . : integrated intensity, central velocity and velocity dispersion. b Closest Gemini slit ID from Konstantopoulos et al. (2014). The star indicates when the Gemini slit is slightly o ff set from the COS aperture. c H α flux and optical line ratios from Konstantopoulos et al. (2014). measured optically) is explored. Other authors have pre-sented 2-dimensional spectral maps of the optical emis-sion lines in Stephan’s Quintet (Iglesias-P´aramo et al. 2012;Konstantopoulos et al. 2014; Rodr´ıguez-Baras et al. 2014;Duarte Puertas et al. 2019), but we will use our own datafrom the GCMS spectrograph to provide an overview andcontext for the observed profiles. In Fig. 4, we show maps ofthe H β emission at three velocity ranges for the whole innerSQ group, ranging from low velocity (blue contours), inter-mediate velocities (green contours) and higher-velocity (redcontours). The blue contours represent velocities closer tothe radial velocity of the intruder galaxy, NGC 7318b (V sys = − ) which is thought to be entering the group frombehind with a discrepant velocity of almost 1000 km s − rel-ative to the rest of the group members, and the group-widegas (Xu et al. 2003; Hwang et al. 2012). A component ofthe emission shown by the blue contours in Fig. 4a followsthe spiral arm and HII regions seen in NGC 7318b, as ex-pected if some of the gas was part of that galaxy. The redcontours (Fig. 4c) are clearly associated with gas in NGC7318a and NGC 7319, but with significant emission from themain N / S filament. Gas at intermediate velocities (Fig. 4b)is spread along the filament, but also in the bridge betweenNGC 7318b and NGC 7319. Fig. 4d shows the range of ve-locities integrated over in a, b and c. This figure highlightsthe complexity of the morpho-kinematics of the shockedIGM, with many clumps of ionized gas appearing at di ff erentplaces along the structure, and over a wide range of veloci-ties. 3.4. Stacked spectra and the detection of theC iv λλ In Fig. 5 we show the averaged spectra of both Ly α andthe C iv λ iv line de-tected on the stacked spectra is centered around the helio-centric recession velocity of the intra-group gas ( ≈ − ) and is very broad (FWHM > − ). Theprofiles of the stacked Ly α and C iv lines also show somesimilarities, with, for Ly α , a brighter low-velocity compo-nent around 6000 − − and a fainter shoulder around6300 − − . We note that the noise in the stackedspectrum is not Gaussian, indicating that some low-level pat-tern noise is present, which makes the estimate of the S / N ra-tio of the line uncertain. The individual spectra are not veryuseful and show weak detections ( ≈ σ ) for all five positions. ABSORPTION AND SCATTERING OF Ly α PHOTONS IN THE IGMBy comparing the Ly α , H β , and [C ii ] spectra, we can de-duce at which velocities along the line of sight the Ly α pho-tons are mostly absorbed. There are four possible examplesof where Ly α absorption is taking place. In Fig. 2b and cwe see that the high velocity component of the double profileseen in both H β and [C ii ] is significantly suppressed com-pared with the low-velocity component. For Fig. 2b in par-ticular, the peak in the H β profile falls close to a strong dip inthe Ly α profile, which appears as two small wings on eitherside of the dip. For Fig. 2c, the feature seen in both [C ii ] andH β is largely suppressed. Also, in Fig. 2f, the Ly α profile isshifted blueward of the main [C ii ] and H β peaks, suggestiveof absorbing gas centered at V helio = − . Asym-metric Ly α profiles like this are often associated with radialoutflows in galaxies (e.g. Heckman et al. 2011). The COSpointing (HCG92-5) samples the gas in the so-called ”AGNbridge” (Cluver et al. 2010), a linear H filament that is ap- road L y α line emission from the S tephan ’ s Q uintet shock Figure 2.
Comparison of COS Ly α , Herschel [C ii ] spectra, and GCMS (VIRUS-P) IFU Spectrograph H β line profiles for the 5 positionsobserved. The COS spectra have been smoothed to a velocity resolution of FWHM =
40 km s − . G uillard , A ppleton , B oulanger , S hull et al .parently separate from the main collisional shock in SQ. Astrong shear in the velocity field of the [C ii ] emission wasnoted by Appleton et al. (2013) in that region. The moleculargas responsible for the absorption of the Ly α emission mayblock the escape of Ly α photons on the red side, becauseof the strong shearing motions in the obscuring molecularwhich are not present in the Ly α emitting gas. The spectrumof the CO (1-0) emission from that direction (see Figure 3f),indeed shows significant CO emission at the high velocityside of the profile, which would be consistent with absorp-tion. Finally, HCG92-7 shows, in Fig. 2e, the opposite e ff ect.In this case, again considering the high-velocity componentof the double-horned profile only, we see that the Ly α is sig-nificantly redshifted with respect to the H β emission, with asharp drop in emission as one approaches the peak of the H β (around V helio = − ). This may be another ex-ample of asymmetric absorption, with resonant scattering tothe red-side of the wings of the kinematics. In summary, wesee that two regions in the main emission-line filament arealmost free of absorption, whereas other regions show caseswhere strong absorption is taking place. Even in those case,at least some of the Ly α emission is able to resonantly scatter,and undergo many scatterings within the gas before eventu-ally escaping into the wings of the velocity profile where theoptical depth is much lower. In Sect. 5.2 we estimate theescape fraction of the Ly α photons and the number of scat-terings, and we discuss the multi-phase structure of the IGMgas.Does this interpretation make sense in terms of the ex-pected line ratios for the hydrogen lines? In Table 2 wealso present the H β line fluxes and Ly α to H β flux ratios inte-grated over the 5 di ff erent sets of spectra. It is interesting thatHCG92-1 and 2 both show ratios (31 and 22 respectively)consistent with little or no absorption when compared withCase B recombination (Case B predicts for T = K andtypical interstellar densities a flux ratio F(Ly α ) / F(H β ) of ∼ α and H β line profileshapes. The slightly lower Ly α to H β ratio is consistent withthe increased absorption in the red component of the Ly α pro-file. For both HCG92-3 and 5, F(Ly α ) / F(H β ) are significantlylower than Case B, suggesting stronger absorption, whichis again consistent with the line profiles. Finally, HCG92-7 (which is associated with the extragalactic H ii regions andcontains significant star formation and dust), shows the high-est deviation from Case B (F(Ly α ) / F(H β ) ∼
2) suggestingthat, at that position, most of the Ly α photons are absorbedby dust. ORIGIN AND PROPERTIES OF THE Ly α EMISSIONIN THE INTERGALACTIC MEDIUM OF SQ 5.1.
Shocks and turbulent mixing layers
Strong Ly α line emission, with typical FWHM of 200-1500 km s − , is often observed in high redshift galaxies (e.g.Tapken et al. 2007), where Ly α photons produced by power-ful starbursts scatter o ff neutral gas carried in outflows. Simi-lar broad Ly α profiles are sometimes seen in the inner regionsof Ly α nebulae associated with luminous high-z quasars (e.g.Ginolfi et al. 2018). However, these extreme conditions arethe antithesis of those seen in the SQ filament, where the starformation activity is very weak and, globally, the Ly α emis-sion is mainly powered by dissipation of mechanical energy.In this section, we argue that both radiative shocks and turbu-lent mixing layers may contribute to powering the observedLy α line emission.Shocks having velocities high enough to reach temper-atures capable of collisionally exciting electronic states ofatomic Hydrogen (above 10 K) are a strong source of Ly α photons (e.g. Shull & McKee 1979; Dopita & Sutherland1996; Lehmann et al. 2020). Due to collisional ionization ofhydrogen, Ly α and H β photons are mostly produced in gasat temperatures smaller than T = K, with collisional ex-citation dominating recombination for T > K (Raga et al.2015). In the IGM of SQ, it is likely that a wide distributionof shock velocities is present (Guillard et al. 2009). In thispaper we do not attempt at a detailed modelling of the lineemission from shocks. We rather aim at qualitatively deter-mining which shocks contribute the most to the Ly α and C iv line emission.To do so, we use the results from the MAPPINGS V shockcode library. The physics of the models is fully describedin Sutherland & Dopita (2017) and the data in Alarie &Morisset (2019). These shock models are based on Allenet al. (2008), but the new models extend the predictions forshocks and radiative precursors to shock velocities smallerthan 100 km s − , as well as up to 1500 km s − . These modelsinclude expanded atomic cooling lines and comprise a widerange of shock precursor conditions, from completely neu-tral gas through partially ionized and fully ionized. Magneticfields are also included as they can strongly impact the com-pression and temperature of the post shocked gas. Figure 6shows results from the MAPPINGS V shock models for aneutral atomic pre-shock gas with density n H =
10 H cm − and a pre-shock magnetic field intensity of B = µ G. Thepre-shock gas ionisation fraction is computed from the UVemission generated by the shock. We only show models forshock velocities above V s >
50 km s − because modelling ofH line emission shows that, in SQ, lower velocity shocks aremolecular shocks (Guillard et al. 2009). road L y α line emission from the S tephan ’ s Q uintet shock Figure 3.
Comparison of COS Ly α and CO (1-0) line profiles for the 5 positions observed. The CO (1-0) PdBI spectra are extracted at the COSpositions over an ellipsoid beam of 4 . × . , P.A = (cid:12) . uillard , A ppleton , B oulanger , S hull et al . Figure 4.
A representation of the H β emission from Stephan’s Quintet based on the Mitchell Spectrograph observations. In each of the figures(a, b, c) we show contours of emission covering the blue (5580-6296 km s − ; ∼ intruder velocity), green (6296-6656 km s − ) and red (6653-7733 km s − ) spectral channels of the data cube, (d) shows an integrated spectrum of the main shock region showing how we have integratedeach of three wavelength windows-ascribed blue, green and red-in the previous panels. The bridge between the main filament and the Seyfertgalaxy NGC 7319 mainly appears in the intermediate velocity (green) channels. Note a possible bi-polar outflow from NGC 7319. All of thecontours are projected against the WFC3 F665N (H α ) greyscale image of the system. road L y α line emission from the S tephan ’ s Q uintet shock Figure 5.
HST COS Ly α (red line) and C iv (black line) spectrastacked over all 5 observed positions in the intra-group medium ofStephan’s Quintet. The flux of the C iv line has been multiplied by afactor of 10 for clarity. The widths at 20% and 50% of the stackedC iv peak flux are, respectively, 1100 ±
60 and 900 ±
40 km s − . The left panel of Figure 6 shows some optical and UV linesfluxes normalized to the total radiative flux of the shock , L T = ρ V s , where ρ = . n H m H is the gas mass density,as a function of the shock velocity V s . The solid lines showthe fractional line luminosities from the shock only, whiledashed lines include the contribution of the precursor forshock velocities V s >
100 km s − . Due to collisional ion-ization of Hydrogen atoms, the fraction of the shock emis-sion accounted for by Ly α photons is the highest for shockvelocities smaller than 100 km s − for which Hydrogen ex-citation is mainly collisional. Faster shocks do contributeto Ly α emission but to a lesser fraction of the total radiatedpower. For shock velocities typically above 150 km s − , theLy α shock emission comes mainly from the photo-ionisationof the post-shock gas that has cooled down to ∼ K. TheUV emission produced in fast shocks is also in part processedinto Ly α photons by hydrogen photo-ionization in the radia-tive precursor (Sutherland & Dopita 2017). The C iv and O vi line fractions peak at shock velocities higher than for Ly α ,120 and 200 km s − respectively. The contribution of photo-ionization to the Ly α emission could thus be significant in theIGM of SQ. This contribution would still be associated withthe dissipation of mechanical energy since in SQ the UV fluxis mostly produced by shocks.In the right panel of Fig. 6 we show model predictions forstrength of various emission line fluxes normalized to the H β In planar shock geometry, when the gas passes through the shock, the de-crease of the flux of kinetic energy is balanced by changes in the enthalpyflux and the radiative flux. line, and we compare them to the observed values, averagedfor the 5 positions. The H α / H β ratio in the models is alwaysabove the case B value (2.86), and slightly below the stackedvalue (H α / H β = α / H β = α / H β at low velocities ( V s < − ) is due to col-lisional excitation in the post shock gas, since the precursorgas entering the shock is neutral (Sutherland & Dopita 2017).We also note that both the low C iv / H β observed ratio and theO vi / H β upper limit point to shocks below 150 km s − .Another contribution to the generation of Ly α emissioncould be irradiated molecular shocks at lower velocitiesthan those presented in Fig. 6. Shocks at velocities 30–50 km s − driven into molecular gas at typical densities n H = cm − produce a strong Ly α radiation (Lehmann et al.2020). However, the Ly α / L T ratio in such molecular shocksis lower (10–20%) than for the shocks in atomic gas at simi-lar velocities presented in Fig. 6. In addition to shocks, radia-tive turbulent mixing is likely to contribute to the emission ofUV line emission (e.g., Slavin et al. 1993; Kwak & Shelton2010) and to the overall energy dissipation in the multiphasemedium of SQ. Turbulent mixing layers may also explain thehigh C iv / O vi line flux ratio ( > . Scattering of Ly α photons in a highly clumpy mediumwith large bulk motions as a source of line broadening Neufeld (1991) and Charlot & Fall (1993) first empha-sized the importance of the clumpiness and multi-phase na-ture of astrophysical media in a ff ecting the observed Ly α linestrength and spectral shape. This paper does not attempt at aquantitative modelling of the Ly α emission in the multiphaseIGM of SQ. We only introduce here some qualitative sugges-tions in the framework of Ly α radiative transfer in a clumpymedium, as introduced by Zheng & Miralda-Escude (2002);Verhamme et al. (2015); Gronke et al. (2016, 2017).Comparing the mean ratio between Ly α and H β fluxesmeasured on stacked spectra (Table 2) with the intrinsic value( ≈
30) expected from hydrogen recombination and colli-sional excitation for temperatures of a few 10 K, we esti-mate the escape fraction of Ly α to ∼
30% (see right panel ofFig. 6). If the Ly α line emission is coming mainly from lowvelocity ( V s < − ) shocks, the escape fraction couldbe as low as f esc = . N sc can be es-timated from the observed frequency shift ∆ ν = ν − ν , where ν and ν are the observed and rest frequencies of the Ly α line,as the following: N / sc = ∆ ν/ ∆ ν D ≈ ( a τ ) / , where ∆ ν D isthe thermal Doppler broadening, a the damping parameter,and τ the line-center optical depth (see Neufeld 1990). As-suming T = K and a typical velocity shift of 100 km s − ,2 G uillard , A ppleton , B oulanger , S hull et al . Figure 6.
Results from the MAPPINGS V shock models library from Alarie & Morisset (2019), covering a range in velocities of 50-1000 km s − . Solid lines show the contribution of the shock only and dashed lines include the contribution of the radiative precursor atvelocities V s >
100 km s − . In this grid of models, the preshock Hydrogen density is n H =
10 H cm − and the pre-shock magnetic field intensityis B = µ G. Left:
Line fluxes normalized to the total radiative flux of the shock, L T = µ m H nV s , as a function of the shock velocity V s , µ = . m H the atomic mass unit. Right:
Line emissivity ratios to H β flux versus shock velocities,with the same chemical abundances (Solar) and shock parameters as Allen et al. (2008). The colored horizontal bands show the observed lineratios, averaged for the 5 positions in the IGM. The grey band spans Ly α / H β ratios from the observed value (10) to the value after correctionassuming an escape fraction of f esc =
30 %. The orange and green bands show respectively the observed H α / H β and C iv / H β values with a ±
20 % uncertainty. The blue band show the upper limit on the O vi / H β ratio. we find N sc ≈
60. In a multiphase medium, the Ly α escapefraction f esc depends on the dust optical depth of the clumps, τ d , and the covering factor f c , i.e. the average number ofclumps along the sightline (Neufeld 1991). Both the mod-elling of the dust emission in SQ (Natale et al. 2010; Guillardet al. 2010) and studies of the molecular gas content (Guil-lard et al. 2012a; Appleton et al. 2017) converge to an averagecolumn density of N H ≈ × cm − in the ridge, whichtranslates into τ d ≈ . λ = f c ≈
15 for a fiducial escape fraction f esc = . N sc ≈
80. Note that in a turbulent medium like SQ, thewidth of the Ly α line does not provide a direct constrainton the number of scatterings and the covering factor, sinceit depends on the velocity gradient over the scattering scaleof the photons. Very high spatial resolution observations areneeded to confirm the presence of such velocity gradients atsmall scales.In conclusion, this high escape fraction, combined with thespectral evidence of Ly α scattering, reflects the clumpy pic-ture that has emerged from the analysis of SQ observations,mainly the spatial correlation between the tracers of the hot,warm and cold IGM phases, and the modelling of the SQdust emission (Guillard et al. 2010). The neutral gas (domi-nated by dusty molecular gas in the ridge) is in clumps witha high velocity dispersion. The clumps are embedded in theX-ray emitting hot, dust-free, plasma. Within such a clumpymedium, Ly α photons escape through multiple scatterings o ff the clump surfaces (Neufeld 1991; Gronke et al. 2017). Theclumps fill a small fraction of the volume but Ly α scatteringindicates that the IGM is not porous to Ly α photons, i.e. theneutral gas surface filling factor must be close to unity withmultiple clumps along a given line of sight. From a quali-tative comparison with radiative transfer models, we believethat the di ff erences in spectral shapes observed for the fivepointings could be accounted for by variations in the totalcolumn of dusty neutral (molecular) Hydrogen, the numberof clumps along the line of sight and the inter-clump velocitydispersion. It is likely that the prominent blue-shifted scat-tering wings observed at Positions 2, 3 and 5 are the result ofsystematic velocity gradients related to the 3D geometry ofthe collision between the intruder and the IGM.5.3. A constant dissipation rate across many orders ofmagnitude in gas temperatures: a signpost of aturbulent cascade?
The observations presented in this paper brings anotherpiece to the cooling budget puzzle of the SQ shocked IGM.Putting together multi-wavelength line spectroscopy allowsus to combine radiative tracers which spans a wide range ofgas temperatures, from ∼ >
100 K for rotational H and [C ii ]lines, to 5 × K for X-rays. Remarkably, over more thanfour orders of magnitude in temperature, the powers radiatedby the multi-phase IGM in X-rays, Ly α , H , [C ii ] are com-parable within a factor of a few (see also Table 1 in Guillardet al. 2009, for a summary of the energy budget across gas road L y α line emission from the S tephan ’ s Q uintet shock CONCLUSIONSWe have used the COS spectrograph on HST to observeLy α emission from the intergalactic gas in SQ. The obser-vations sample five positions across the 30 kpc-wide shock.The HST data is compared with CO, [C ii ] and H β spectra.We summarize the main observational results and outline ourinterpretation of the data.We detect extremely wide Ly α lines with a full width atzero intensity of ≈ − , which exceeds the velocityrange of CO, [C ii ] and H β line emission. After stacking ofthe five HST spectra, we also detect the C iv doublet. Weobserve significant variations in the Ly α / H β spectral ratiobetween positions and velocity components. From the meanline ratio averaged over positions and velocities, we estimatethe mean escape fraction of Ly α photons to be ∼ − α lines are systematically broader than the H β onesat the same positions, which we consider as observationalevidence for scattering of Ly α photons by the SQ IGM. Thedi ff erence in velocity spread is asymmetrical and amountsto ≈
300 km s − for the blue-shifted Ly α wings observed atthree of the five positions.The observations provide insight on the structure of themultiphase IGM in SQ. The high Ly α escape fraction andscattering reflect the clumpy picture suggested by the spa-tial correlation between the tracers of the hot, warm and coldphases of the SQ IGM. The neutral, mainly molecular, gas isin clumps embedded in the X-ray emitting, hot and dust-freeplasma. Ly α photons must escape through multiple scatter-ings o ff the clumps. Scattering indicates that the IGM is notporous to Ly α photons, i.e. the neutral gas surface filling fac-tor must be close to unity with multiple clumps along a givenline of sight. A quantitative comparison with Ly α radiativetransfer models is beyond the scope of this observational pa-per, but we believe that di ff erences in the spectral shapes ofthe Ly α , H β , [C ii ] and CO lines observed for the five posi-tions could be accounted for by variations in the total columnof atomic hydrogen, the number of clumps along the line ofsight and the inter-clump velocity dispersion. It is likely thatthe blue-shifted scattering wings follow from systematic ve- locity gradients related to the 3D geometry of the collisionbetween the intruder and the SQ IGM.The bulk of the Ly α emission must be powered by dis-sipation of mechanical energy because the SQ star forma-tion rate is small and the gas velocities span an exception-ally large range. Ly α photons are emitted by gas at tempera-tures smaller than the thermal energy threshold for collisionalionization ( T < K). It is likely that both collisional ex-citation and recombination of photo-ionized Hydrogen con-tribute to the observed emission. Due to collisional ioniza-tion of hydrogen atoms, the fraction of the shock emissionaccounted for by Ly α photons is the highest for shock ve-locities smaller than 100 km s − . Faster shocks do contributeto Ly α emission but to a lesser fraction of the total radiatedpower. The UV emission produced in fast shocks is in partprocessed into Ly α photons in the post-shock gas and by theradiative precursor. This contribution of photo-ionized gasto the Ly α emission, which is associated with dissipation ofmechanical energy, could be significant.The HST observations complement our view at the ener-getics of the galaxy-wide shock created by the collision ofhigh-speed intruder galaxy with the SQ IGM. The total poweremitted in the Ly α line is comparable to that of much coolergas in the mid-IR rotational H and the [C ii ] fine structurelines. The energy radiated in [C ii ], H , Ly α and X-rays rep-resents cooling from gas spanning four order of magnitudesin temperature from 100 to 10 K. The observed fluxes arecomparable within a factor of a few, which indicates thatroughly the same fraction of energy is dissipated per loga-rithmic bin of temperature. This is a remarkable result thatconstrains models of the turbulent energy cascade in SQ. Itemphasises the possible contribution from turbulent mixinglayers to energy dissipation.Following the trail of mechanical energy dissipation andgas kinematic in the turbulent gas on scales smaller than ≈ James Web Space Telescope , and future UV-optimized tele-scopes. Such observations will also help to further elucidatethe anisotropic motions in the population of clumpy molec-ular structures, which are necessary in order to explain theblue scattering wings in the Ly α profiles seen in some of theobserved positions.4 G uillard , A ppleton , B oulanger , S hull et al .ACKNOWLEDGMENTSPG thank the Centre National d’Etudes Spatiales (CNES),the University Pierre and Marie Curie, and the ”ProgrammeNational de Cosmologie and Galaxies” (PNCG) and the”Physique Chimie du Milieu Interstellaire” (PCMI) pro-grams of CNRS / INSU for there financial supports. We thankDaniel Kunth and Brigitte Rocca for very useful physicaland technical discussions about Ly α scattering. Supportfor Program number HST-GO-13321.001-A was provided byNASA through a grant from the Space Telescope Science In-stitute, which is operated by the Association of Universitiesfor Research in Astronomy, Incorporated, under NASA con-tract NAS5-26555. PA thanks Guillermo Blanc (CarnegieObservatories) and Emily Freeland (formerly Texas A&MUniversity) for providing assistance with observations, andsoftware / data reduction associated with the Mitchell Spectro-graph. This work used observations carried out under projectnumber U020 (P.I. Guillard) with the IRAM NOEMA Inter-ferometer, reduced and analysed with the GILDAS software.IRAM is supported by INSU / CNRS (France), MPG (Ger-many) and IGN (Spain). REFERENCES
Adams, J. J., Blanc, G. A., Hill, G. J., et al. 2011, The hetdex pilotsurvey. I. Survey design, performance, and catalog ofemission-line galaxies, IOP Publishing Ltd.,doi: 10.1088 / / / / / ia.01851101p.2019.55.02.21Allen, M. G., Groves, B. A., Dopita, M. A., Sutherland, R. S., &Kewley, L. J. 2008, The Astrophysical Journal SupplementSeries, 178, 20, doi: 10.1086 / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / MNRAS / STY364Green, J. C., Froning, C. S., Osterman, S., et al. 2012, \ apj, 744,60, doi: 10.1088 / / / / / / / / l26 road L y α line emission from the S tephan ’ s Q uintet shock Gronke, M., Dijkstra, M., Mccourt, M., & Peng Oh, S. 2017,Astronomy and Astrophysics, 607, 71,doi: 10.1051 / / / / / / / / / / / / / / / j.1365-2966.2005.09870.xHeckman, T. M., Borthakur, S., Overzier, R., et al. 2011,Astrophysical Journal, 730, 5, doi: 10.1088 / / / / / / j.1365-2966.2011.19847.xIglesias-P´aramo, J., L´opez-Mart´ın, L., V´ılchez, J. M., Petropoulou,V., & Sulentic, J. W. 2012, Astronomy and Astrophysics, 539,doi: 10.1051 / / / mnras / stz1248Joshi, B. A., Appleton, P. N., Blanc, G. A., et al. 2019, \ apj,doi: 10.3847 / / ab2124Konstantopoulos, I. S., Appleton, P. N., Guillard, P., et al. 2014,Astrophysical Journal, 784, doi: 10.1088 / / / / / / / / / / / ff s, R. J., Xu, C. K., et al. 2010, AstrophysicalJournal, 725, 955, doi: 10.1088 / / / / / / / s10509-011-0699-5O’Sullivan, E., Giacintucci, S., Vrtilek, J. M., Raychaudhury, S., &David, L. P. 2009, Astrophysical Journal, 701, 1560,doi: 10.1088 / / / / / / / / / mnras / stu851Shull, J. M., & McKee, C. F. 1979, The Astrophysical Journal, 227,131, doi: 10.1086 / / \ apj, 407, 83,doi: 10.1086 / / ff ects of pre-ionisationin radiative shocks I: self-consistent models,doi: 10.3847 / / aa6541Tapken, C., Appenzeller, I., Noll, S., et al. 2007, Astronomy andAstrophysics, 467, 63, doi: 10.1051 / / / ff s, R. J. 2003,The Astrophysical Journal, 595, 665, doi: 10.1086 / //