HALOGAS: Extraplanar gas in NGC 3198
G. Gentile, G. I. G. Jozsa, P. Serra, G. H. Heald, W. J. G. de Blok, F. Fraternali, M. T. Patterson, R. A. M. Walterbos, T. Oosterloo
AAstronomy & Astrophysics manuscript no. n3198˙arxiv c (cid:13)
ESO 2018July 1, 2018
HALOGAS: Extraplanar gas in NGC 3198
G. Gentile , , G. I. G. J´ozsa , , P. Serra , G. H. Heald , W. J. G. de Blok , , F. Fraternali , , M. T. Patterson , R. A. M.Walterbos , and T. Oosterloo , Sterrenkundig Observatorium, Universiteit Gent, Krijgslaan 281, B-9000 Gent, Belgiume-mail: [email protected] Department of Physics and Astrophysics, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium Netherlands Institute for Radio Astronomy (ASTRON), Postbus 2, NL-7990 AA Dwingeloo, The Netherlands Argelander-Institut f¨ur Astronomie, Auf dem H¨ugel 71, D-53121 Bonn, Germany Astrophysics, Cosmology and Gravity Centre (ACGC), Astronomy Department, University of Cape Town, Private Bag X3, 7700Rondebosch, Republic of South Africa Astronomy Department, University of Bologna, Bologna, Italy Kapteyn Astronomical Institute, University of Groningen, AD Groningen, the Netherlands Department of Astronomy, New Mexico State University, PO Box 30001, MSC 4500, Las Cruces, NM 88003, USAPreprint online version: July 1, 2018
ABSTRACT
We present the analysis of new, deep H i observations of the spiral galaxy NGC 3198, as part of the HALOGAS (Westerbork HydrogenAccretion in LOcal GAlaxieS) survey, with the main aim of investigating the presence, amount, morphology and kinematics ofextraplanar gas. We present models of the H i observations of NGC 3198: the model that matches best the observed data cube featuresa thick disk with a scale height of ∼ i mass of about 15% of the total H i mass; this thick disk also has a decrease inrotation velocity as a function of height (lag) of 7–15 km s − kpc − (though with large uncertainties). This extraplanar gas is detectedfor the first time in NGC 3198. Radially, this gas appears to extend slightly beyond the actively star-forming body of the galaxy (astraced by the H α emission), but it is not more radially extended than the outer, fainter parts of the stellar disk. Compared to previousstudies, thanks to the improved sensitivity we trace the rotation curve out to larger radii. We model the rotation curve in the frameworkof MOND (Modified Newtonian Dynamics) and we confirm that, with the allowed distance range we assumed, fit quality is modestin this galaxy, but the new outer parts are explained in a satisfactory way. Key words.
Galaxies: halos - Galaxies: ISM - Galaxies: kinematics and dynamics - Galaxies: individual (NGC 3198) - Galaxies:structure
1. Introduction
The last couple of decades have seen a wealth of observationaldata revealing the presence of material outside the plane of diskgalaxies (see Sancisi et al. 2008 and Putman et al. 2012 for areview), e.g. hot X-ray emitting gas (e.g. T¨ullmann et al. 2006, Liet al. 2008), ionised gas (Collins & Rand 2001, Rossa & Dettmar2003), dust (Howk & Savage 1999, M´enard et al. 2010), andneutral hydrogen (Fraternali et al. 2002; Oosterloo et al. 2007;Heald et al. 2011). This material is found to be often (but notalways, in the case of the H i ) closely associated to star formationin the galaxy disk.The galactic fountain is an example of a framework that at-tempts to explain the presence and properties of extraplanar gas:gas is expelled to large distances from the disk by supernovaexplosions and adiabatic expansion, and then it falls back ontothe galaxy disk due to radiative cooling (Shapiro & Field, 1976;Bregman, 1980). For the conservation of angular momentum,there must be a decrease of the rotation velocity (also known as“lag”) with distance from the plane. However, ballistic models ofthe interplay between the thin disk and the extraplanar gas (e.g.Fraternali & Binney 2006) show that there must be additionalmechanisms governing the kinematics of extraplanar gas, be-cause these simple models obtain lags that are too shallow com-pared with observations. A possibility is the interaction betweenthe uplifted gas and a corona of slowly-rotating hot gas, which would explain the magnitude of the observed lags (Marinacci etal. 2011). Other explanations for the properties of extraplanargas involve e.g. thermal instabilities in the corona (Kaufmann etal. 2006, but see Binney et al. 2009), pressure gradients or thee ff ect of the magnetic field (Benjamin 2002).Observational evidence for extraplanar lagging H i gas comesfrom both edge-on and moderately inclined galaxies. Edge-ongalaxies allow us to assess the extent of this gas and the mag-nitude of the lag, whereas in moderately inclined galaxies onecan investigate the local connection between star formation andextraplanar gas, and one can study in detail its kinematics.Observationally, in moderately inclined galaxies the very ex-istence of the lag allows the disentangling of extraplanar gasfrom gas that resides in the plane, e.g. in NGC 2403 (Fraternaliet al. 2002), NGC 4559 (Barbieri et al. 2005), and NGC 6946(Boomsma et al. 2008).In Heald et al. (2011) we presented the HALOGAS(Westerbork Hydrogen Accretion in LOcal GAlaxieS) survey, asystematic investigation of extraplanar gas in 22 spiral galaxies,using very deep H i observations. The main goal of HALOGASis to investigate the amount and properties of extraplanar gas.The target galaxies were selected to be nearby and to fall in oneof two categories: edge-on galaxies (inclination i ≥ ◦ ) andmoderately-inclined galaxies (50 ◦ ≤ i ≤ ◦ ).NGC 3198 is one of the moderately-inclined galaxies in theHALOGAS sample. It is a late-type spiral that is traditionally a r X i v : . [ a s t r o - ph . C O ] A p r . Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198 Fig. 1.
Channels maps of the H i observations of NGC 3198 presented here. The velocity of each channel map is indicated in thetop right corner of each panel (every fourth channel map is shown). The beam is indicated in the bottom left (35.2 × σ ), 2.1, 10.5, 52.5 mJy beam − . The blue contour is the 0.42 mJy beam − (2 σ ) level of the 60 arcsecresolution cube. Negative contours are dashed grey. The cross indicates the galaxy centre.considered as one of the benchmark galaxies for rotation curvestudies, because its H i kinematics is regular and symmetric, itsH i disk is very extended and its inclination angle (around 70 ◦ )is ideal for deriving the rotation curve. Many studies have fo-cussed partly or completely on the kinematics of NGC 3198, e.g. van Albada et al. (1985), Begeman (1989), Blais-Ouellette et al.(2001), Bottema et al. (2002), de Blok et al. (2008), Sellwood &S´anchez (2010), Gentile et al. (2011). The distance of NGC 3198has been determined to be 13.8 Mpc using Cepheids (Kelson et
2. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198
Table 1.
The galaxy NGC 3198 and the observations presentedhere. The beam FWHM refers to the cube modelled in this paper.References are: 1: HyperLeda; 2: Kelson et al. (1999), Freedmanet al. (2001); 3: this work.
Parameter Value ReferenceHubbe type Sc 1Adopted distance 13 . ± . . ± .
07 1Synthesised beam FWHM 35.2 × / beam to Kelvin 0.51 3On-source time 132 hr 3Velocity resolution 4.12 km s − − al. 1999, Freedman et al. 2001, see also Macri et al. 2001), usu-ally considered to be one of the most precise distance indicators.In the present paper we aim to characterise the presence,amount, and kinematics of any extraplanar gas in NGC 3198,using the most sensitive H i data available and a careful tilted-ring model based analysis. We also look for a link between thepresence of this extraplanar gas and star formation in this galaxy.Additionally, we present detailed kinematic modelling of the ob-servations, as well as a mass decomposition of the rotation curve.
2. Data acquisition and reduction i data We summarise here how the data were acquired and reduced, andwe refer the reader to Heald et al. (2011) for a more thoroughdescription of the observations and data reduction, of which weonly describe the main points. Information about NGC 3198 andthe observations presented here is given in Table 1.The observations of neutral hydrogen (H i ) were obtainedat the WSRT (Westerbork Synthesis Radio Telescope) in the“maxi-short” configuration, for 10 ×
12 hours with a bandwidthof 10 MHz divided into 1024 channels, and two linear polariza-tions. The data were reduced using the software package Miriad(Sault et al. 1995). After flagging, calibration and continuumsubtraction, the data were Fourier inverted to obtain a data cube.When inverting, a Gaussian taper with an image-plane widthof 30 arcseconds was applied to the uv-data, which were alsoHanning-smoothed to a velocity resolution of 4.12 km s − . Thedata cubes obtained in this manner have an rms noise of 0.21 mJybeam − . Avoiding to apply the taper gives a higher-resolutioncube (18.9 × ff use emission, whose signal-to-noise ratio is higherwhen applying the taper and thus somewhat degrading the reso-lution.The CLEAN deconvolution of the dirty cube was done intwo steps: first, a CLEAN mask was defined (for every chan-nel map) where presence of real emission can be establishedwith confidence. In these regions the CLEAN deconvolution wasperformed down to approximately 1 σ . Then, the channel mapswere CLEANed on the whole field, down to about 2 σ . Finally,the channel maps were restored using a Gaussian beam with aFHWM size of 35.2 × We obtained photometric observations of NGC 3198 within theframework of HALOSTARS, which is a deep optical survey of
Fig. 2.
Total H i map (contours) superimposed onto a false colourr’- band HALOSTARS image (see text). Contours are 0.1, 1, 5,and 15 × atoms cm − . The black arrow indicates the tenta-tively detected H i described in Section 3.the HALOGAS targets with the Isaac Newton Telescope (INT).NGC 3198 was observed in the r’-band on February 4, 2010,and the total on-source time amount was 3600 s. A coadded im-age was obtained using the THELI imaging reduction pipeline(Schirmer et al. 2003, Erben et al. 2005). After the overscan- andbias correction, the images were flat-fielded, super-flat-fieldedand defringed. We also masked non-Gaussian noise-featuressuch as cosmic rays. Using SCAMP (Bertin 2006), the resultingimages were photometrically calibrated, background-subtractedand co-added, taking the sky background variation in the individ-ual chips into account and using the 2MASS catalogue to solvefor astrometric distortion and for the relative sensitivity of thechips. Using the SDSS (DR8) catalogue, we determined the zeropoint to be 24.67 ± / swe thus find a 1- σ level of 26.4 mag over 1 square arcsec. To en-hance faint extended features, the image was convolved with aGaussian with a FWHM 3 arcsec. α imaging The H α observations of NGC 3198 were made at the Kitt PeakNational Observatory (KPNO) as part of an observing run aimedat obtaining wide-field optical images of the HALOGAS sam-ple, to investigate the relation between extraplanar gas and starformation. We observed over 4 nights in January 2012 using theMosaic 1.1 instrument on the 4-meter telescope. We used an H α filter centered at 6574.74 Å and a FWHM of 80.62 Ångstroms,which includes [NII] lines at 6548 and 6584 Å. We exposed inthe 5 dither pattern with six minutes per exposure for a total of30 minutes. R-band images (10 minutes total) were also acquiredfor continuum subtraction. http: // / theli /
3. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198
Fig. 3.
Channels maps of NGC 3198: observed (black) and modelled (red), using the rotation curve and inclination parameters fromde Blok et al. (2008). Symbols are the same as Fig. 1. For better readability we have plotted fewer contours: -0.5, 0.5 (2.5 σ ), and 20mJy beam − . Also, compared to Fig. 1, fewer channels are shown. Fig. 4.
Position-velocity diagrams of NGC 3198 along the major axis, considering a position angle of 213 ◦ : observed (black)and modelled (left and middle panel: red; right panel: green), derived using the steps outlined in section 4. Left: without laggingextraplanar gas. Middle: with a single 1-kpc thick disk with a lag. Right: with a separate disk of lagging extraplanar gas. Contoursare -0.3, 0.3 ( ∼ σ ), 3, and 30 mJy beam − . The cross indicates the galaxy centre.
3. Observed H i data The data cube is shown in Fig. 1, for the lowest contour we alsoplotted the low-resolution (60 arcsec) cube. The channel mapsshow a regularly rotating disk and some asymmetries (e.g., thesouth-west side is slightly more extended).The total H i map was obtained by computing the 0th momentof a masked version of the data cube, which was determined bymasking out spurious emission (i.e., emission that is not above2 σ in at least three consecutive channels of the 60 arcsec reso-lution cube). Fig. 2 shows the total H i map superimposed with aHALOSTARS image.From the primary beam corrected data cube we find a totalH i flux of 239.9 Jy km s − , which translates into a total H i massof 1.08 × M (cid:12) assuming a distance of 13.8 Mpc. This H i mass is slightly (6%) higher than the value given in Walter etal. (2008) from the THINGS (The H i Nearby Galaxy Survey)survey, which might partly be due to amplitude calibration un-certainties, and partly to the fact that we detect some extraplanargas.In Fig. 2 one can see that our data are more sensitive to faintextended emission than previous observations. At a resolution of 30 arcsec, we find meaningful emission down to ∼ × atoms cm − . Despite their di ff erent angular resolutions, this canbe compared with the lowest contour given by Begeman (1989)(0.5 × atoms cm − ) and de Blok et al. (2008) (1 × atoms cm − ). Note that the latter study focussed more on hav-ing a high spatial resolution than on being sensitive to extendedemission. Comparing our map with the one in Begeman (1989),we now confirm the existence of the south-west extension hintedat by Begeman’s 0.5 × atoms cm − contour (see his Fig. 4).However, the total H i map presented here extends further out, es-pecially in the south-west side of the galaxy, and we tentativelydetect some very faint emission to the north of the north-easternside (visible also in Fig. 1, channels at 493.0 km s − to 558.9 kms − ). We estimate the mass of this H i structure (indicated witha black arrow in Fig. 2) to be about 5 × M (cid:12) . We note thatthis is the only feature that clearly does not belong to the thin orthick disk. In many galaxies there is evidence for more of suchfeatures, indicating some amount of accreting H i .At a projected distance of about 120 kpc from NGC 3198 wealso detect in H i a galaxy pair, identified in NED (NASA / IPACExtragalactic Database) as VV 834. One of its two componentshas a radial velocity from the Sloan Digital Sky Survey of 566 ±
4. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198
Fig. 5.
Channels maps of NGC 3198: observed (black) and modelled (green), using our final model parameters. Symbols are thesame as Fig. 1. For better readability we have plotted fewer contours: -0.5, 0.5 (2.5 σ ), and 20 mJy beam − .256 km s − . We find an H i mass of 2.1 × M (cid:12) for the galaxypair. The centres of the two galaxies are only ∼
19 arcsec apart,and the H i emission is barely resolved, with a total extent ofabout 1.5–2 arcmin and spreading over 80 km s − , from ∼ − to ∼
620 km s − .
4. Models
To investigate in detail the morphology and kinematics of theH i in this galaxy, we resorted to making models of the wholedata cube: it is the most thorough approach that can distentan-gle subtle di ff erences between various representations of the H i disk. The uncertainties on the derived parameters are discussedin Section 4.1.
5. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198
Fig. 6.
Position-velocity diagram (pvd) of NGC 3198 along themajor axis, considering a position angle of 213 ◦ . The observedpvd is shown in black and the model is in red. The model shownhere is derived with a thin disk only (plus noise), convolved withthe dirty beam, CLEANed and restored as described at the end ofsection 4. Contours are -0.3, 0.3 ( ∼ σ ), 3, and 30 mJy beam − . Fig. 7.
Position-velocity diagram (pvd) of the high-resolutioncube (18.9 × ◦ . Contours are -0.29, 0.29( ∼ σ ), 0.58, 1.16, ... mJy beam − .The first model we made was based on the parameters de-rived by de Blok et al. (2008): as can be seen in Fig. 3, the gen-eral features of the data cube are reproduced, but these parame-ters cannot account for the detailed structure of the emission ineach channel map, which is partly expected as in the present pa-per we detect more di ff use emission which typically has di ff erentkinematics.To improve the model data cube and obtain a better matchto the data, we decided to follow the strategy described below,based on successive approximations.First, we derived a rotation curve based on a tilted-ring mod-elling on the intensity-weighted velocity field (Begeman 1989),leaving the position angle, the inclination, the systemic velocityand the rotation velocity as free parameters for each ring. Eachside (approaching and receding) was treated separately and thebest-fit systemic velocity was found to be 657 km s − . Usingthe intensity-weighted velocity field can potentially underesti-mate the inner rotation curve, especially if the velocity gradient Fig. 8.
Radial dependence of the parameters of the best-fit modeldata cube presented in this paper. Blue represents the approach-ing (north-east) side and red is the receding (south-west) side.“pa” is the position angle, “incl” is the inclination, “V rot ” is therotation velocity, and “H i surf dens” is the H i surface density.is high, as is the case here in the inner regions of NGC 3198.Thus we fixed the position angle and inclination from this fit,and fitted the rotation curve on the velocity field derived usingthe WAMET method (Gentile et al. 2004), a modified version ofthe envelope tracing method. For galaxies with a high inclination(such as NGC 3198) or with a poor resolution it is a more reli-able choice than the more traditional intensity-weighted mean,which in NGC 3198 might give velocities that are slightly bi-ased towards the systemic velocity. The orientation parametersfrom the intensity-weighted velocity field were also used to de-rive the H i surface density profile by averaging the total H i mapover ellipses. Concerning the gas distribution in the vertical di-rection, we used a sech distribution, initially with a scale heightof 0.2 kpc.Then, these parameters were used as an initial guess to makea fit of the whole data cube using the TiRiFiC (Tilted RingFitting Code) software (J´ozsa et al. 2007), leaving the inclina-tion, the position angle and the rotation velocity free for eachring, separately for each side, and the velocity dispersion (in-cluding instrumental e ff ects) free as a global parameter. Thebest-fit velocity dispersion found by TiRiFiC is 11.7 km s − .Because of the very high number of degrees of freedom, theparameters of the best-fit model data cube from TiRiFiC tend tohave some strong discontinuities and unphysical “jumps” as afunction of radius. Also, by construction TiRiFiC uses a χ min-imisation, which naturally (for a constant noise) gives a higher
6. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198
Fig. 9.
Comparison total H i maps: observed (black), with all theinclination of our best model plus two degrees (red) and minustwo degrees (blue). Contours are 3 × and 1.5 × atomscm − . Fig. 10. H α image overlaid with contours of the total H i map.The contour levels are chosen to enhance the central parts of theH i disk: 1 × , then (1, 5, 7.5, 10, 12.5, 15, 17.5, 20, 22.5, 25) × .weight to high signal-to-noise regions. However, these regionsare not necessarily the most discriminant between models, thatis why the TiRiFiC fitting process has to somehow be manuallyguided.Therefore we used this set of parameters as a basis to ad-just the parameters manually when necessary, in order to obtainat the same time a good match of the data cube and a realisticset of parameters. The inclination and the position angle wereadjusted based on the total H i map, and the rotation curve was Fig. 11.
Rotation curve derived here (full red points) comparedwith the rotation curve derived by de Blok et al. (2008) from theTHINGS data (open black points) and by Begeman (1989) usingless deep WSRT data (blue squares).adjusted based on the position-velocity diagram along the majoraxis. Changes of at most 5 km s − were made to obtain at bettermatch to the observations.The emission in some channels is skewed (channels 525.9 –591.8 km s − and 707.2 – 740.2 km s − , see Fig. 1), without anapparent change in the global position angle, and with no sign ofposition angle change at other azimuths at these radii.Spekkens & Sellwood (2007) introduced a formalism to fitvelocity fields, focussing on bisymmetric distortions, which theydeem as being more realistic than the more commonly used ra-dial flows. The reason is that while these two models give simi-larly good results, they consider radial flows to be less physicallyplausible as they would have no clear physical origin, and be-cause of the epicyclic approximation they are restricted to smallnon-circular motions. Therefore, we interpret the skewness ofthe emission in channels 525.9 – 591.8 km s − and 707.2 – 740.2km s − as a bisymmetric distortion of the velocity field, possiblydue to an elongation of the potential, ignoring the radial term thatis not needed by the present data. In such a model the velocityV model at a given position in the galaxy disk is modelled as: V model = V sys + sin i [ V rot cos θ − V , rot cos (2 θ b ) cos θ ] (1)where V sys is the systemic velocity, V rot is the rotation veloc-ity, V , rot is the m = θ is the azimuthal anglerelative to the projected major axis, and if φ b is the angle be-tween the bisymmetric distortion and the projected major axis,then θ b = θ − φ b . Such a distortion, when included in the model,indeed contributes to improving the match between the modeland the observations. It turns out that the amplitude V , rot mustbe ∼
15 km s − beyond a radius of approximately 510 arcsec, and φ b must be ∼ ◦ to avoid having skewed emission in the centralchannels (around the channel at 657.8 km s − ) but to have it atchannels 525.9 – 591.8 km s − and 707.2 – 740.2 km s − . Amodel with radial motions (limited to a wedge around the majoraxis, to avoid having skewed emission in the central channels)gives almost identical results to this bisymmetric distortion.A position-velocity diagram along the major axis with thismodel is shown in the left panel of Fig. 4. As clearly visible from
7. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198 the position-velocity diagram, to reproduce the observations it isnecessary to add some lagging gas. The first attempt was a sin-gle thick disk with with a scale-height of 1 kpc and a lag of 18km s − kpc − in the approaching side and 8 km s − kpc − in thereceding side (Fig. 4, middle panel). This model reproduces theobservations better than the single thin disk, but it is not satisfac-tory: by increasing the rotation curve in the inner parts (to matchthe outermost contours), there would be lagging gas missing; byincreasing the amplitude of the lag the outer regions would notbe reproduced correctly.Therefore we resorted to a two-disk model, with a a thickdisk that rotates more slowly than the main (thin) disk, with thesame orientation parameters and run of surface density. This isshown in the right panel Fig. 4, where a model with a laggingthick disk component with a scale-height of 3 kpc is a betterrepresentation of the observations. The lag in the model was de-rived to be 15 km s − kpc − in the approaching side and 7 km s − kpc − in the receding side. The thick disk contains about 15% ofthe total H i mass and its surface density distribution is a scaleddown version of the total H i density distribution.The comparison between the final model data cube and theobserved one is shown in Fig. 5. Because the lagging extraplanargas emission is only 1–2% of the peak in each channel map, wenow investigate whether it could be an artifact due to the CLEANdeconvolution and improper modelling of the dirty beam. To thisaim, we build a model data cube based on our final model datacube, only without the thick disk. We then convolve it with thedirty beam, we add some Gaussian noise (with an rms chosen inorder to obtain the same rms noise as in the observations), andwe CLEAN and restore this cube. If the lagging extraplanar gashad been an artifact, in Fig. 6 we would have seen some “pseudoextraplanar gas”. However, we do not, therefore we concludethat the observation of lagging extraplanar gas is genuine and isnot an artifact due to the dirty beam. This is also confirmed bythe hint of a signal from the extraplanar gas that we see in thehigh-resolution cube (see Fig. 7).The best-fit parameters are plotted in Fig. 8 (only the param-eters with a radial dependence): the two sides are quite symmet-ric. Uncertainties on the parameters that describe the model datacubes were estimated by varying the parameter in question un-til the model data cube becomes only marginally consistent withthe observed one, focussing on a particular projection of the datacube or a particular region that is particularly sensitive to that pa-rameter.First, the inclination and position angle are better derived bycomparing total H i maps: a variation in inclination or positionangle of only 2 ◦ makes the model total H i map barely consistentwith the observed one (see Fig. 9 for the inclination).In a similar way, the (global) velocity dispersion is ideallyconstrained by position-velocity diagrams. It turns out that thehigh-velocity contours of the position-velocity diagram alongthe major axis become too spaced out if the velocity dispersionof 11.7 km s − changes by more than 2 km s − .The uncertainties on the rotation curve were assumed to behalf the di ff erence between the two sides. To avoid having unre-alistically small errors when the velocities of the two sides areclose, we also considered a minimum error of 4.12 / sin( i ) km s − where i is the inclination angle and 4.12 km s − is the channelincrement of the data cube we used. The amplitude of bisym- metric distortion can be constrained from the channel maps: wefind that it is quite uncertain (15 ±
10 km s − ).How robust are the figures about extraplanar gas? For thethin disk we assumed a scale-height of 0.2 kpc, and for thethick disk we tried di ff erent values. For a given inclination (con-strained from the total H i map, see above), the central channelsare particularly sensitive to the scale-height, and we find a valueof 3 ± − kpc − . These uncertainties includethe degeneracy between scale-height and lag amplitude: a smallscale-height can be partly compensated by a large lag value andviceversa. However a close inspection of the data cube helps re-solving this ambiguity and the above uncertainties take it intoaccount.There are however two additional uncertainties in the com-putation of the lag value. First, the lag amplitude might varywith radius (e.g. NGC 891, Oosterloo et al. 2007): we cannotexclude a radially varying lag amplitude with values up to 10km s − kpc − higher in the inner parts. Second, the value of thelag is also somewhat degenerate with the velocity dispersion ofthe thick disk: a higher value of the latter can be compensatedby a lower value of the lag. However, the velocity dispersion ofthe thick disk cannot be higher than about 20 km s − (and cor-respondingly a lag about 25% lower than the abovementionedvalues), in which case the emission in the central channels be-comes too extended. Lastly, a thick disk with no vertical gradientin the lag but whose rotation velocity is 20 km s − than the thindisk yields a model that is almost as good as the model with avertical gradient in the lag.
5. Discussion
We find that the presence of anomalous gas is required to matchthe observed and model data cubes. The anomalous gas (definedas the modelled thick disk) has a mass of 15 ±
5% of the total H i mass, which is comparable to previous studies that use very deepH i observations to study extraplanar gas, e.g. Zschaechner et al.(2011) and Fraternali et al. (2002). The lag is also a requiredfeature of the model, meaning that the extraplanar gas is pro-gressively rotating more slowly with increasing vertical distancefrom the mid-plane. The value of the lag is comparable with pre-vious estimates, e.g. Oosterloo et al. (2007) and Zschaechener etal. (2011).The H α image of section 2.3 is shown in Fig. 10. There isa clear correspondence between the brightest H i peaks and H α emitting regions. How does the radial extent of extraplanar gascompare with optical properties? HyperLeda gives an R of 194arcsec, whereas the RC3 gives 255 arcsec. From the position-velocity diagram along the major axis (Fig. 4) there is evidencefor extraplanar gas out to 5–6 arcmin (receding side) and at least6–7 arcmin (approaching side). This can also be compared tothe extent of H α emission (see Fig. 10 and Fig. 12, left panel),and to our deep r’-band image from the HALOSTARS survey(Fig. 2, Fig. 12, right panel). Since a precise determination ofthe extent of the extraplanar gas is problematic because it is quitemodel-dependent, an accurate comparison of the various radialextents (extraplanar gas, stars, H α emission) is hardly possiblein a quantitative manner. From the surface brightness profiles(Fig. 12) it appears that the H α emission extends out to ∼ .
8. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198
Fig. 12.
Azimuthally-averaged surface brightness profiles of the H α image (left) and our HALOSTARS r’-band image (right). Fig. 13.
MOND fit of the rotation curve derived in the presentpaper, using a distance free within the uncertainties of theCepheids method (13.8 ± ∼ The rotation curve we find (Fig. 11) is roughly consistent withthe one derived by de Blok et al. (2008) and Begeman (1989).The former focussed more on the high resolution of their data,whereas in the present paper the data have a lower angular res-olution but a higher sensitivity to extended emission. In particu-lar, we confirm the velocity decrease by 10–15 km s − between200 and 250 arcsec, and we find a meaningful rotation curve outto (at least) 720 arcsec. The last useful radius is not an unam-biguous quantity to determine. In order to fully exploit the data,but at the same time to avoid overinterpreting them, we defined(rather conservatively) the last reliable radius as the average be-tween the last radius where the average surface density is above1 × atoms cm − (650 arcsec) and the last radius where thetilted-ring fit on the velocity field converged. At these large radii(720 arcsec correspond to 48 kpc for a distance of 13.8 Mpc)there is still no sign of decrease of the rotation curve. Note thatin Begeman (1989) the last few points of the rotation curve werederived by assuming the same inclination and position angle asthose at a radius of 9.5 arcmin. Moreoever, at these radii in thereceding side, Begeman’s fit to the velocity field is only basedon positions outside the major axis. The MOND (Modified Newtonian Dynamics) paradigm was in-troduced by Milgrom (1983) as an explanation (alternative todark matter) for the absence of Keplerian decline in the ob-served kinematics of galaxies. MOND has a remarkable pre-dictive power on galactic scales (see e.g. Famaey & McGaugh2012 and references therein), even though on larger scales alsoMOND needs some invisible mass. The galaxy studied in thepresent paper, NGC 3198, was claimed to show tension withMOND (Bottema et al. 2002, Gentile et al. 2011). It is an idealcase study for MOND because of its inclination (perfectly suitedfor kinematical studies, see Begeman 1989), its relative close-ness, the fact that it is a late-type spiral galaxy with regular andsymmetric kinematics, and the accurate determination of its dis-tance using Cepheids (Kelson et al. 1999, Freedman et al. 2001).Indeed, the MOND fit is extremely sensitive to the assumed (or
9. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198 fitted) distance. In MOND, the gravitational acceleration g N pro-duced by the visible matter is linked to the true gravitational ac-celeration g through the interpolating function µ : µ (cid:32) ga (cid:33) g = g N , (2)where µ ( x ) ∼ x for x (cid:28) µ ( x ) ∼ x (cid:29)
1. Recentlypreference has been given to the so-called “simple” µ function: µ ( x ) = x + x .The contribution of the stellar disk was taken from de Bloket al. (2008), where is was derived from 3.6 µ m observations (in-cluding a correction for a possible mass-to-light ratio gradient,based on the J-K colour gradient), and the contribution of thegas from the model data cube discussed above: thin disk and thethick disk were treated separately (each with its own mass andscale height), and then their velocities were added quadratically.We analyse here in some more detail the MOND fit with thedistance left free within the uncertainty of the Cepheids deter-mination. In comparison with the fit that used the THINGS data(Gentile et al. 2011, Fig. 5), the present MOND fit is of compa-rable overall quality: see Fig. 13. However, some di ff erences canbe seen: in the innermost parts MOND fits the THINGS rotationcurve better than the HALOGAS rotation curve, whereas in theouter parts the situation is reversed. We note that MOND fitseach rotation curve best where each rotation curve is expectedto be the best choice: THINGS in the inner parts (because of thehigher resolution) and HALOGAS in the outer parts (because ofthe better sensitivity to the outer, fainter emission). The best-fitvalues of the two fits are the same (best-fit 3.6 µ m mass-to-lightratio of 0.48 and best-fit distance of 12.3 Mpc, at the lower endof the allowed range). The two fits di ff er in the shape of the con-tribution of the gas (apart from, obviously, the rotation curve it-self): with the more extended data from HALOGAS, we accountbetter for the outer surface density profile of the gas, and there-fore we trace better the region where the contribution of the gasstart declining. Also, none of the two fits manage to reproducethe decrease in rotation velocity between 200 and 250 arcsec,which corresponds to the end of the brighter part optical disk.
6. Conclusions
We have presented new, very deep (10 ×
12 hours) H i observa-tions of the spiral galaxy NGC 3198. The observations are partof the HALOGAS (Westerbork Hydrogen Accretion in LOcalGAlaxieS) survey, see Heald et al. (2011). These new observa-tions go significantly deeper than previous H i observations.We made careful 3D models of the H i layer in NGC 3198,including not only traditional features such as a variable inclina-tion, position angle, and the rotation curve, but also newer fea-tures such as variable rotation speed as a function of distancefrom the plane, a thicker disk and a bisymmetric distorsion ofthe kinematics. In this manner we managed to obtain a modelthat matched the observed data cube very well.We revealed for the first time in this galaxy the presence ofextraplanar gas over a thickness of a few ( ∼
3) kpc. Its amount isapproximately 15% of the total mass, and one of its main proper-ties is that it appears to be rotating more slowly than the gas closeto midplane, with a (rather uncertain) rotation velocity gradientin the vertical direction (lag) of 7–15 km s − kpc − .Despite the uncertainty on its actual radial extent, the ex-traplanar gas seems to be slighly more extended than the star-forming part of the galaxy (as revealed by the H α image), butour new deep r’-band image reveals that there is stellar emission out to the end of the detected extraplanar gas. We detect thin diskH i out to a radial extent that is twice as far as the stellar disk andextraplanar layer. We also detect a faint H i complex beyond theH i disk on the northern side of NGC 3198, with an estimatedmass of ∼ × M (cid:12) .Finally, we make a mass model in the context of MOND(Modified Newtonian Dynamics) of the newly-derived rotationcurve, which is more extended than previous determinations.The fit quality is modest, similarly to previous studies, but theouter parts are explained in a satisfactory way. Acknowledgements
GG is a postdoctoral researcher of the FWO-Vlaanderen(Belgium). PS is a NWO / Veni fellow. RAMW and MP ac-knowledge support for this project from the National ScienceFoundation under grant AST-0908126. Based on observationswith the Kitt Peak National Observatory, National OpticalAstronomy Observatory, which is operated by the Associationof Universities for Research in Astronomy (AURA) under co-operative agreement with the National Science Foundation.This publication makes use of data products from the TwoMicron All Sky Survey, which is a joint project of theUniversity of Massachusetts and the Infrared Processing andAnalysis Center / California Institute of Technology, funded bythe National Aeronautics and Space Administration and theNational Science Foundation. Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, theParticipating Institutions, the National Science Foundation, theU.S. Department of Energy, the National Aeronautics and SpaceAdministration, the Japanese Monbukagakusho, the Max PlanckSociety, and the Higher Education Funding Council for England.The SDSS Web Site is http: // / . Finally, we thankRenzo Sancisi for useful comments on an early version of thismanuscript. References
Barbieri, C. V., Fraternali, F., Oosterloo, T., et al. 2005, A&A, 439, 947Begeman, K. G. 1989, A&A, 223, 47Benjamin, R. A. 2002, Seeing Through the Dust: The Detection of HI and theExploration of the ISM in Galaxies, 276, 201Bertin, E. 2006, Astronomical Data Analysis Software and Systems XV, 351,112Binney, J., Nipoti, C., & Fraternali, F. 2009, MNRAS, 397, 1804Blais-Ouellette, S., Amram, P., & Carignan, C. 2001, AJ, 121, 1952Boomsma, R., Oosterloo, T. A., Fraternali, F., van der Hulst, J. M., & Sancisi, R.2008, A&A, 490, 555Bottema, R., Pesta˜na, J. L. G., Rothberg, B., & Sanders, R. H. 2002, A&A, 393,453Bregman, J. N. 1980, ApJ, 236, 577Collins, J. A., & Rand, R. J. 2001, ApJ, 551, 57de Blok, W. J. G., Walter, F., Brinks, E., et al. 2008, AJ, 136, 2648Erben, T., Schirmer, M., Dietrich, J. P., et al. 2005, Astronomische Nachrichten,326, 432Famaey, B., & McGaugh, S. 2012, Living Reviews in Relativity, in press(arXiv:1112.3960)Fraternali, F., van Moorsel, G., Sancisi, R., & Oosterloo, T. 2002, AJ, 123, 3124Fraternali, F., & Binney, J. J. 2006, MNRAS, 366, 449Freedman, W. L., Madore, B. F., Gibson, B. K., et al. 2001, ApJ, 553, 47Gentile, G., Salucci, P., Klein, U., Vergani, D., & Kalberla, P. 2004, MNRAS,351, 903Gentile, G., Famaey, B., & de Blok, W. J. G. 2011, A&A, 527, A76Heald, G., J´ozsa, G., Serra, P., et al. 2011, A&A, 526, A118Howk, J. C., & Savage, B. D. 1999, AJ, 117, 2077J´ozsa, G. I. G., Kenn, F., Klein, U., & Oosterloo, T. A. 2007, A&A, 468, 731Kaufmann, T., Mayer, L., Wadsley, J., Stadel, J., & Moore, B. 2006, MNRAS,370, 1612Kelson, D. D., Illingworth, G. D., Saha, A., et al. 1999, ApJ, 514, 614
10. Gentile et al.: HALOGAS: Extraplanar gas in NGC 3198