An off-centred bulge or a satellite? Hydrodynamical N -body simulations of the disc galaxy NGC 5474
R. Pascale, M. Bellazzini, M. Tosi, F. Annibali, F. Marinacci, C. Nipoti
MMNRAS , 1–21 (2020) Preprint 26 November 2020 Compiled using MNRAS L A TEX style file v3.0
An off-centred bulge or a satellite?Hydrodynamical N -body simulations of the disc galaxy NGC 5474 R. Pascale (cid:63) , M. Bellazzini , M. Tosi , F. Annibali , F. Marinacci , C. Nipoti , INAF - Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, Via Gobetti 93/3, 40129 Bologna, Italy Dipartimento di Fisica e Astronomia, Universit`a di Bologna, via Gobetti 93/2, 40129, Bologna, Italy
Accepted 2020 November 24. Received 2020 November 24; in original form 2020 August 28
ABSTRACT
We present dynamical models of the star-forming galaxy NGC 5474 based on N -body hy-drodynamical numerical simulations. We investigate the possible origin of the compact roundstellar structure, generally interpreted as the bulge of the galaxy, but unusually off-set by (cid:39) in projection from the visual and the kinematic centres of both the star and the gasdiscs. We argue that it is very unlikely that the putative bulge is in a co-planar orbit in thedisc plane, showing that such a configuration would be hardly compatible with its smoothand regular spatial distribution, and, in case its mass is above M (cid:12) , also with the regular H I velocity field of NGC 5474. Instead, if the putative bulge is in fact an early-type satellitegalaxy orbiting around NGC 5474, not only the off-set can be easily produced by projectioneffects, but our simulations suggest that the gravitational interaction between the two systemscan explain also the warped H I distribution of NGC 5474 and the formation of its loose spiralarms. As a by-product of the simulations, we find that the peculiar over-density of old starsdetected in the south-west region of NGC 5474 may be explained with the interaction betweenNGC 5474 and a smaller stellar system, unrelated to the putative bulge, accreted in the discplane. Key words: galaxies: individual: NGC 5474 - galaxies: interactions - galaxies: kinematicsand dynamics - galaxies: peculiar - galaxies: stellar content - galaxies: structure.
Located at a distance of .
98 Mpc (Tully et al. 2013), NGC 5474is a local star forming galaxy, classified as SAcd pec, belonging tothe M 101 Group. With an absolute blue magnitude M B (cid:39) − . ,it is among the most luminous satellites of M 101, also known asthe Pinwheel Galaxy, and it is also relatively close to it, with anangular separation smaller than ◦ (corresponding to ∼
120 kpc ,in projection; Tikhonov et al. 2015).Due to its various asymmetries (Kornreich et al. 1998), NGC5474 was early recognized as peculiar (Huchtmeier & Witzel1979). The H I disc is distorted at radial distances R larger than , resulting in a change of the position angle (PA) of (cid:39) ◦ twisting from ◦ for R < to ◦ for R > . (Rownd et al. 1994, hereafter R04). The change in the PA is thoughtto be associated with a warp in the H I disc, which connects thegaseous component to the south-western edge of M 101. This H I bridge (Huchtmeier & Witzel 1979; van der Hulst & Huchtmeier1979) has been often considered as a tidal debris formed during arecent fly-by of NGC 5474 close to M 101 (Mihos et al. 2012a). (cid:63) E-mail: [email protected]
The off-set between the kinematic centre of its H I discand the optical centre of what has always been interpreted as thegalaxy bulge is another fascinating peculiarity of NGC 5474. Asan example, Fig. 1 shows a zoomed-in view of the central regionof NGC 5474 in the F814W band obtained from the LEGUS pho-tometric catalogue (Calzetti et al. 2015): the kinematic centre ismarked with a blue dot, while the off-set bulge is clearly visible tothe north of the kinematic centre. First observed by van der Hulst& Huchtmeier (1979), then validated by R04 and Kornreich et al.(2000, hereafter K00) looking at the H I emission, the off-set hasalso been confirmed from the H α kinematics (Epinat et al. 2008,hereafter E08), making the picture even more puzzling. This oddlylarge discrepancy has recursively raised the question about the truenature of such a stellar component and what mechanisms may haveproduced it.Throughout this work we will refer to the galaxy’s ‘cen-tral’ stellar component as putative bulge (hereafter PB), an unbi-ased title that reflects our ignorance about its nature: it may be anoff-set pseudo-bulge; it may have been formed in-situ, in an asym-metric burst of star formation; it may be the remnant of an exter-nal accreted dwarf galaxy; it may be an external galaxy crossingthe line of sight (R04; Mihos et al. 2012b), bound or unbound to © 2020 The Authors a r X i v : . [ a s t r o - ph . GA ] N ov R. Pascale et al. ◦ RA D E C NGC 5474
NE PB
SW over density
Figure 1.
Image of NGC 5474 obtained using observations in the F814Wband from the LEGUS photometric catalogue (Calzetti et al. 2015). Theblue dot shows the position of the H I kinematic center from R04, the orangecircle is centred on the H I kinematic center and its radius is
90 arcsec long,corresponding to at a distance D = 6 .
98 Mpc . The white contoursare separated by k/ − I max , where I max is approximately the surfacebrightness of the PB centre, while k = 0 , , , . The orange arrow pointsthe PB, while the SW over-density is to the South. North is up and East isto the left. NGC 5474. Any of these explanations would require to call the PBwith a different name. According to Fisher & Drory (2010) thiscomponent has properties more similar to a pseudo-bulge than to aclassical bulge (see Kormendy & Kennicutt 2004; Fisher & Drory2008), while Bellazzini et al. (2020, hereafter B20) showed that itsstructural parameters are also consistent with the scaling relationsof dwarf galaxies. For instance, the PB is similar to the dwarf ellip-tical galaxy (dE) NGC 205 in terms of stellar mass, V-band abso-lute magnitude and projected half-mass radius (R04; McConnachie2012; B20). B20 showed that the stellar populations of the PB arevery similar to those dominating the stellar mass budget in the discof NGC 5474 and the color-magnitude diagrams are fully compat-ible with systems lying at similar distance from us. Moreover, theyconstrained the maximum difference in radial velocity between thedisc of NGC 5474 and the PB to ∼
50 km s − . Hence, if the PB isnot a substructure of NGC 5474 it should be a satellite of it or, atleast, another member of the M 101 group.The presence of a stellar over-density to the South-West of thePB, a structure that extends for almost , adds up to NGC5474 oddities (B20, see also Fig. 1). The main population dominat-ing the over-density is older than , very similar to the stellarpopulation dominating the PB. Young stars in NGC 5474 appearnot to be correlated to the over-density, and dominate instead thespiral arms, which extend out from the centre. The spiralpattern seen in optical is also marked by the H I distribution (R04).Among the possible explanations, it does seem plausible that suchover-density may have been caused by a recent or on-going inter-action between NGC 5474 and the off-centre PB, which may be also the cause of the galaxy’s large scale, asymmetric recent starformation (B20).In this work, we study the dynamical properties of NGC 5474and investigate what the true nature of the PB may be by mak-ing use of realistic N -body hydrodynamical simulations, supportedby analytic models. In Section 2 we describe the dynamical modelbuilt to match NGC 5474 and the PB: its properties, the observa-tional data; the approach used to constrain the model. In Section 3we estimate the gravitational effects felt by NGC 5474 due to thePB to put limits on the large parameter space to be investigated withnumerical simulations. In Sections 4 and 5 we focus on simulations.After describing the method used to sample the initial conditions,we explore two different scenarios: i) a purely stellar system (with-out dark matter) orbiting within the plane of the disc of NGC 5474,and ii) a compact early type dwarf galaxy (with its own dark-matterhalo) moving on a polar orbit around NGC 5474. The latter case isobviously intended to explore the possibility that PB is a satellite ofNGC 5474 that is seen near the centre of its disc only in projection.On the other hand, the former case is the mean by which we explorethe effects of an off-centred bulge on the underlying gaseous andstellar disc. In particular we are interested to answer questions like:is the off-set position of the PB compatible with some kind of long-standing quasi-equilibrium configuration and/or a relatively regularrotation curve? Can a stellar PB resist the drag by dynamical fric-tion and/or the tidal strain from the dark-matter halo of its parentgalaxy? Is our understanding of the galaxy driven by projection ef-fects? Could a possible interaction between NGC 5474 and the PBexplain some of the other peculiarities of NGC 5474? Section 6concludes. Throughout this work we assume that the main structures that com-prise the target are a dominant disk galaxy similar to NGC 5474 andthe PB, a compact stellar component that can be either embeddedin a dark-matter halo or not.
We describe NGC 5474 as a multi-component galaxy comprisinga dark-matter halo, a gaseous disc and a stellar disc. We assumethat the dark halo is spherical with Navarro, Frenk & White (1996,hereafter NFW) density distribution ρ NFWdm ( r ) = 4 ρ s (cid:18) rr s (cid:19) (cid:18) rr s (cid:19) , (1)where r s is the halo scale radius and ρ s ≡ ρ NFWdm ( r s ) .We assume that both the gaseous and the stellar discs arerazor-thin exponential discs. The H I surface number density isgiven by n ( R ) = M gas πh m p exp (cid:18) − Rh gas (cid:19) , (2)where M gas is the gas total mass, m p is the proton mass (assuming,for simplicity, that the disc is fully composed by hydrogen atoms),and h gas is the H I disc scale length.The stellar surface density is given by Σ (cid:63) ( R ) = M (cid:63) πh (cid:63) exp (cid:18) − Rh (cid:63) (cid:19) , (3) MNRAS000
We describe NGC 5474 as a multi-component galaxy comprisinga dark-matter halo, a gaseous disc and a stellar disc. We assumethat the dark halo is spherical with Navarro, Frenk & White (1996,hereafter NFW) density distribution ρ NFWdm ( r ) = 4 ρ s (cid:18) rr s (cid:19) (cid:18) rr s (cid:19) , (1)where r s is the halo scale radius and ρ s ≡ ρ NFWdm ( r s ) .We assume that both the gaseous and the stellar discs arerazor-thin exponential discs. The H I surface number density isgiven by n ( R ) = M gas πh m p exp (cid:18) − Rh gas (cid:19) , (2)where M gas is the gas total mass, m p is the proton mass (assuming,for simplicity, that the disc is fully composed by hydrogen atoms),and h gas is the H I disc scale length.The stellar surface density is given by Σ (cid:63) ( R ) = M (cid:63) πh (cid:63) exp (cid:18) − Rh (cid:63) (cid:19) , (3) MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 where M (cid:63) is the disc total stellar mass and h (cid:63) is the stel-lar disc scale length. In our analysis, a model of NGC5474 is fully determined by the free parameter vector ξ ≡{ ρ s , r s , h gas , M gas , h (cid:63) , M (cid:63) } . To determine ξ , we fit a dataset ofobservations of NGC 5474 with our galaxy model. We anticipatethat, in the subsequent analysis of NGC 5474 through hydrody-namical N -body simulations, we will drop the approximation ofrazor-thin discs for gas and stars in favour of realistic discs withnon-negligible thickness. The first H I observations from van der Hulst & Huchtmeier (1979)have too low velocity resolution ( ∼
27 km s − ) to allow for anydetailed kinematic study. The H I data collected through VLA ob-servations by R04 provide an H I velocity field relatively smoothand symmetric in the galaxy’s central, unwarped region, with a ro-tation curve peaking at ∼
14 km s − . This is inconsistent with K00whose rotation curve is − systematically lower, even thoughthe authors use the very same H I observations. Also, both H I rota-tion curves hardly agree with the H α emission, tracing the gas kine-matics of the innermost galaxy’s region. According to E08,the H α rotation curve sharply rises up to
22 km s − at R (cid:39) and it is barely consistent with R04 further out. Nonetheless, all theaforementioned studies agree and report the same off-set betweenthe PB and the H I / H α kinematic centres. For our study we rely onthe rotation curve of R04 which provides a detailed description ofthe data reduction and has a large radial coverage.The dataset used to constrain the NGC 5474 galaxy modelthen consists of: i) the H I rotation curves derived starting from theanalysis of R04 for the approaching and receding arms using a tiltedring model; ii) the observed H I column density profile of R04; iii)the stellar disc parameters resulting from the stellar disc/bulge de-composition of Fisher & Drory (2010).R04 provides a collection of N v = 18 points { R k , v a k , v r k , v t k } , with k = 0 , ..., N v , where R k is the dis-tance of the observed point from the H I disc’s kinematic centre, v a k ( v r k ) is the corresponding velocity of the approaching (receding)arm as a result of the fit with the tilted ring model using halfring, while v t k is obtained using a complete ring. To rederive the H I rotation curve and determine a reliable error δv k , accountingfor the asymmetries of the two arms and the uncertain and lowinclination we proceed as follows. For each radial bin k i) we compute σ k ≡ | v a k − v r k | / , as a measure of the velocityasymmetry between the two arms.ii) With a Monte Carlo approach, we sample M = 20000 newvelocities v jk , with j = 1 , ..., M , from a Gaussian distribution withmean and standard deviation equal to v t k and σ k , respectively.iii) Each velocity v jk is deprojected assuming a different inclina-tion i , drawn from a uniform distribution over the range of inclina-tions [17 ◦ , ◦ ] , consistent with the estimate of R04.iv) We now have M realizations of the intrinsic rotation velocitywhich we use to build the probability distribution of the rotationvelocity of the considered bin. For each bin, we adopt the th per-centile of the distribution as rotation velocity and the minimum dis-tance between the th and th , and the th and th percentilesas error.The upper panel of Fig. 2 shows the H I rotation curve thatwe obtained. We mark the separate contributions of our referenceNGC 5474 galaxy model with different colors, as we shall discussin details in the following sections. At least for R < . , v c [ k m s − ] NGC 5474
Flat rotation curveUn-corrected R [kpc] N H I [ c m − ] TotalHaloStellar discHI disc
Figure 2.
Top panel: H I rotation curve derived in Section 2.1.1 (squareswith error bars), superimposed to the model rotation curve. We mark withdifferent colors and line types the contributions to the total circular speed(black dotted curve) of the halo (red dash-dotted curve), the stellar disc(orange dashed curve) and the gaseous disc (blue solid curve). For com-parison, we show also the uncorrected rotation curve (triangles with errorbars). Bottom panel: H I observed column density profile as derived in Sec-tion 2.1.1 (squares with error bars) superimposed to the model profile (bluesolid curve, equation 2). the rotation curves of the two arms from R04 are quite simi-lar, with σ (cid:39) − , so the velocity distribution is relativelysmooth and symmetric. The deprojected rotation curve peaks at R ∼ − , where v (cid:39)
42 km s − , and it slowly decreasesout to . For distances larger than , the rotation curverises up to v (cid:39)
50 km s − . As pointed out by R04, the rise in theoutermost regions is likely due to the H I disc’s warp. We have noreason to believe that it would rise again in the outer parts, espe-cially if this is a warped region. Since we are interested in getting awell motivated dynamical mass for NGC 5474, which can be usedas a starting point for our simulations, we impose to our derivedrotation curve to remain flat beyond .The H I column density profile has been derived from the H I column density map of R04. We first plot the column density ofeach pixel as a function of the distance from the disc’s kinematiccentre. We build 15 radial bins, each containing the same numberof pixels (approximately 100). For each bin, we compute the distri-bution of the H I surface density. We take the th percentile of thedistribution as a measure of the bin’s H I column density, and themaximum interval between the th and th , and th and th percentiles as error. We assume that the disc is razor-thin, and cor-rect for the inclination according to the relation n int = n obs cos i ,where n int and n obs are, respectively, the intrinsic and observedsurface density, i = 21 ◦ (the mean inclination of the range usedto estimate the rotation curve). We exclude the innermost bin, cor-responding to a galactocentric distance of R = 1 kpc , for whichwe do not have any kinematic information. At the end of the pro- MNRAS , 1–21 (2020)
R. Pascale et al. cedure, the H I column density profile consists of N n = 14 triplets { R i , n k , δn k } , with k = 0 , ..., N n , which denote, respectively,the galactocentric distance ( R k ), the observed profile ( n k ) and theassociated error ( δn k ).To derive a motivated range of stellar scale lengths andmasses, we start from the stellar disc/bulge decomposition of Fisher& Drory (2010), who find that the best fit stellar disc modelhas log h (cid:63) / kpc = 3 . ± . , assuming a distance d =5 .
03 Mpc . This value, converted to our distance d = 6 .
98 Mpc (Tully et al. 2013), gives h (cid:63) = 1 . ± .
23 kpc . The estimate ofFisher & Drory (2010) is based on data in the 3.6 µ m band, a goodtracer of stellar mass, with a weak dependence on age and metal-licity. B20 found that the bulk of the stellar mass in the disc shouldbe provided by intermediate to old age populations. According tothe theoretical models by R¨ock et al. (2015), the 3.6 µ m mass-to-light ratio Υ for a Kroupa IMF and in the metallicity range relevantfor NGC 5474 (Z ∈ [0 . , . ; B20) is Υ ∼ . for a old population and Υ ∼ . for a
10 Gyr old population. In thefollowing we take these two values as our reference.Starting from the absolute magnitude in the 3.6 µ m band M disc3 . = − . ± . as in Fisher & Drory (2010), for the adopteddistance d = 6 .
98 Mpc and assuming M . , (cid:12) = 2 . from Ohet al. (2008), we convert the disc stellar luminosity into mass ob-taining M (cid:63) = 2 . ± . × M (cid:12) and M (cid:63) = 4 . ± . × M (cid:12) , (4)for Υ ∼ . and Υ ∼ . , respectively. We conclude that, forthe stellar disc, a plausible range of total stellar mass, accountingfor all the uncertainties in the estimates provided above, is M (cid:63) ∈ [2 . , . × M (cid:12) , given a stellar disc scale length of h (cid:63) = 1 . ± .
23 kpc . We note that our estimate of M (cid:63) is consistent with Skibbaet al. (2011), who measure a total stellar mass M (cid:63) = 5 × M (cid:12) using far-infrared imaging from the Herschel Space Observatoryfor galaxies in the KINGFISH project. The log-likelihood of a galaxy model ln L ( ξ |D ) , defined by theparameter vector ξ , given the data D is ln L = ln L v + ln L n . (5)The first term in the r.h.s. is ln L v = − N v (cid:88) k =0 (cid:18) v c ( R k ; ξ ) − v k δv k (cid:19) , (6)where v c is the model circular speed given by v = v h + v + v (cid:63) , (7)with v h ≡ πr s ρ s G (cid:20) ln(1 + r/r s ) r/r s −
11 + r/r s (cid:21) (8)the contribution to the circular speed due to the halo, and v = 2 GM cmp h cmp y × [ I ( y cmp ) K ( y cmp ) − I ( y cmp ) K ( y cmp )] (9) Fisher & Drory (2010) model the stellar disc of NGC 5474 with a razor-thin disc model, as in equation (3). the contribution to the circular speed of any of the discs, measuredin the discs plane, with cmp = (cid:63) for the stellar disc and cmp = gas for the gaseous disc. In equations (8) and (9), G is the gravitationalconstant and I n and K n are Bessel’s functions of the n -th order. Inequation (5), the latter term is ln L n = − N n (cid:88) k =0 (cid:18) n ( R k ; ξ ) − n k δn k (cid:19) , (10)where n and n k are the model (equation 2) and observed H I sur-face density profiles. Since the discs’ and halo contributions maybe highly degenerate, we fix the stellar disc parameters to observa-tionally motivated values, consistent with the aforementioned esti-mates. We adopt h (cid:63) = 1 .
71 kpc as stellar disc scale length and theaverage value M (cid:63) = 4 . × M (cid:12) as stellar mass.We perform a parameter space search using a Markov ChainMonte Carlo (MCMC) method. We run 16 chains, each evolvedfor 7000 steps, using a classical Metropolis-Hastings sampler(Metropolis et al. 1953; Hastings 1970) to sample from the pos-terior. We adopt flat priors on the free parameters. After a burn-inof 3000 steps (which we eliminate as a conservative choice), we usethe remaining steps to build the posterior distributions over ξ . Fig. 3shows the marginalized one- and two-dimensional posterior distri-butions over the models’ parameters. We estimate the uncertaintieson the models’ free parameters using the 16 th , 50 th and 84 th per-centiles of the corresponding marginalized one-dimensional distri-butions.We note that since the PB is off-set with respect to the disckinematic centre we cannot include it in our axisymmetric modelof the rotation curve. However, after computing the PB mass fromits stellar population, in Section 3 we try to estimate the possibleeffects of the PB on the gas kinematics.Figure 2 shows the newly derived H I rotation curve super-imposed to that of the model. We highlight with different col-ors the contributions of the different components. The dark-matterhalo dominates over the stellar and gaseous components at all theradii covered by the kinematic data. Even if the total stellar massis lower than the total H I mass, in the central regions the starsprovide a significant contribution, dominant over the H I disc for R ≤ , due to the very different gas and stellar scale lengths.As a reference, we estimate M (cid:63) /M gas | (cid:39) within , and M (cid:63) /M gas | (cid:39) . at the larger distance . Also, we mea-sure an H I mass M gas | (cid:39) × M (cid:12) within R = 9 kpc ,consistent with the estimate of R04. The bottom panel of Fig. 2shows the H I disc column density as a function of the galactocen-tric distance, superimposed to the model. R04 reports that the discsurface density flattens in the central parts, which is consistent withan exponential disc model with a large scale length (equation 2,Fig. 3). If the inner H I surface density were constant and equal tothe innermost point of the observed profile, our exponential modelwould overestimate the H I mass within by only 10%, which,given the uncertainties on the observed profile and the model as-sumptions, we consider of negligible impact.We define our reference model as the model (i.e. a set of ξ )with the maximum likelihood (equation 5) a posteriori. Table 1 liststhe main parameters of disc and halo here derived, and the param-eters of the reference NGC 5474 model. In our analysis the PB can be either embedded in a dark-matter haloor not. We represent its stellar component with a spherical S´ersic
MNRAS000
MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 log ρ s = 6 . +0 . − . − . . . l og r s log r s = 0 . +0 . − . . . . . l og h ga s log h gas = 0 . +0 . − . . . . log ρ s . . . . l og M ga s − . . . log r s . . . . log h gas . . . . log M gas log M gas = 9 . +0 . − . Figure 3.
One- and two-dimensional marginalized posterior distributions over the model free parameters ( ρ s , r s , h gas , M gas ) . The black curves in the two-dimensional marginalized distributions correspond to regions enclosing, respectively, 68%, 95% and 99% of the total probability. The orange vertical linesin the one-dimensional marginalized distributions correspond to the 16 th , 50 th and 84 th percentiles, used to estimate the uncertainties over the models freeparameters. The vertical grey lines in the marginalized one-dimensional distributions, and the squares in the marginalized two-dimensional distributions markthe position of the reference model (see also Table 1). Table 1.
Main parameters of the analytic NGC 5474 models of halo, H I and stellar discs. ρ s and r s : reference density and scale radius of the NFW dark-matterdensity profile (equation 1); M gas : total mass of the H I disc; h gas : H I disc’s scale length (equation 2); M (cid:63) : total mass of the stellar disc; h (cid:63) : stellar discscale length (equation 3). The parameters ( ρ s , r s , h gas , M gas ) are determined as described in Section 2.1.2, while we fixed the parameters of the stellar discto the ones derived in Section 2.1.1. The middle row lists the σ uncertenties over the model free parameters, while the bottom row lists the parameters of thereference model adopted throughout this work.NGC 5474 DM halo HI disc Stellar discparameter ρ s [ M (cid:12) kpc − ] r s [ kpc ] M gas [ M (cid:12) ] h gas [ kpc ] M (cid:63) [ M (cid:12) ] h (cid:63) [ kpc ]value . +45 . − . . +1 . − . . +2 . − . . +6 . − . [2 . , .
1] 1 . +0 . − . reference model .
95 1 .
51 1 .
82 6 .
76 4 . . Table 2.
Main parameters of the PB relevant for this work. R e : PB effective radius from B20; M PB : PB total stellar mass as derived in Section 2.2; m : PBS´ersic index (equation 11) from B20; ρ s , PB /M PB : PB dark-matter halo scale density-to-stellar mass (equation 14); r s , PB and r t , PB : PB dark-matter haloscale and truncation radii, respectively (see equation 14).Putative bulge Stars Dark-matterparameter R e [ kpc ] M PB [ M (cid:12) ] m ρ s , PB /M PB [ kpc − ] r s , PB [ kpc ] r t , PB [ kpc ]value .
484 [0 . ,
2] 0 .
79 5 . × , 1–21 (2020) R. Pascale et al. (1968) model, whose surface brightness profile is I ( R ) = I e exp (cid:20) − b m (cid:18) RR e (cid:19) m (cid:21) , (11)where I e = b m πm Γ(2 m ) L tot R . (12)In equations (11) and (12), Γ is the Gamma function, L tot is the to-tal PB luminosity, m is the S´ersic index, related to b m as in equation18 of Ciotti (1999), and R e the effective radius (i.e. the distance onthe plane of the sky from the PB’s centre that contains half of thetotal PB’s luminosity L tot ).We adopt m = 0 . and R e = 0 .
484 kpc from B20 (see alsoTable 2) and we infer the stellar mass from the total 3.6 µ m luminos-ity by Fisher & Drory (2010) using the same mass-to-light ratios weadopted for the disc. The absolute magnitude in the 3.6 µ m band es-timated by Fisher & Drory (2010) is M . = − . ± . , which,converted assuming a distance d = 6 .
98 Mpc , gives M . = − . ± . . We follow Forbes et al. (2017) and, assuming the σ limits in the luminosity and adopting M . , (cid:12) = 2 . from Ohet al. (2008), we get M PB = 0 . +0 . − . × M (cid:12) and M PB = 1 . +0 . − . × M (cid:12) , (13)for Υ = 0 . and Υ = 0 . , respectively . Based on the afore-mentioned estimates, we take the slightly wider M PB ∈ [0 . , × M (cid:12) as a reasonable range of stellar masses to explore for thePB. The main parameters ( m, R e , M PB ) relevant to this work arelisted in Table 2. At any rate, we should recall that any derivation ofthe PB and disc masses from Υ depends on the adopted IMF, andthat switching, e.g., from a Kroupa to a Salpeter’s IMF can changethe resulting masses by a factor of 3 (see Fig.s 13 and 14 of R¨ocket al. 2015).When present, the PB dark-matter halo has density distribu-tion ρ dm , PB ( r ) = ρ s , PB rr s , PB (cid:18) rr s , PB (cid:19) e − (cid:32) rr t , PB (cid:33) (14)i.e., a truncated NFW model, where ρ s , PB and r s , PB are, respec-tively, the halo scale density and the characteristic radius, while r t , PB is the halo truncation radius.According to estimates of the stellar-to-halo mass relation(Read et al. 2017), for a galaxy with stellar mass M PB ∈ [0 . , × M (cid:12) , one would expect a virial-to-stellar-mass ra-tio M vir , PB /M PB (cid:39) . However, since we consider the PBas an external galaxy orbiting around NGC 5474, we expect thePB halo to be less massive than estimated and to be truncatedwell before its nominal virial radius because of tidal interactions.Also, the structural properties of its stellar component resemble theones of a typical dE galaxy (see also B20), which we do not ex-pect to be significantly dominated by dark matter (McConnachie2012). As such, we use the above M vir , PB /M PB only as a refer-ence value to estimate r s , PB . Adopting the halo mass-concentrationrelation from Mu˜noz-Cuartas et al. (2011), from which we estimatea concentration log c = 1 . − . , we get a halo scale radius r s , PB = 2 . − .
91 kpc . We set r s , PB = 2 . , but we do As for the stellar disc, the mass-to-light ratios Υ are in the 3.6 µ m band. not expect our results to depend substantially on r s , PB . To derive anew mass scale we impose that the dynamical-to-stellar-mass ratio M dyn , PB /M PB , evaluated at the stellar half-mass radius r h , is M dyn , PB M PB (cid:12)(cid:12)(cid:12)(cid:12) r h ∼ , (15)which is approximately the ratio expected for a dE of sizes andstructure similar to the PB (McConnachie 2012). We truncate thePB halo at r t , PB = 15 kpc . As we will later discuss, such a value isslightly less than the initial distance we set between the PB and theNGC 5474 centres in the simulations of Section 5.2.2 and it avoidsthe halo PB to be unreasonably massive. With these choices, thePB has a dark-matter halo 20 times as massive as its stellar compo-nent. As a reference, the most massive PB halo, corresponding toa stellar mass M PB = 2 × M (cid:12) , has a total dark-matter mass (cid:39) × M (cid:12) , approximately the same as the halo virial massof NGC 5474. Table 2 lists the relevant parameters of the PB usedthroughout this work. By means of the analytic model derived in Sections 2.1 and 2.2, wequantify the mutual effects that the galaxy and the PB may haveon each other when the latter is placed within the discs’ equatorialplane. This allows us to put further constraints on the large param-eter space we will explore with hydrodynamic N -body simulationsin the following sections. We start analyzing the two scenarios of a PB with M PB = 0 . × M (cid:12) and M PB = 2 × M (cid:12) , respectively the lower and upperlimits of the mass range derived in Section 2.2. We assume no darkmatter since we do not consider the PB as an external galaxy, andwe first place it on the discs’ plane, away from the kinematiccentre, with its current size and S´ersic index ( R e = 0 .
484 kpc , m = 0 . , see Table 2). We refer to the two analytic models as,respectively, model 0.5 and model 2, where the number indicatesthe PB mass, in units of M (cid:12) . Using these models, we roughlyestimate the minimum and maximum distortions we may expect tosee in the H I disc’s velocity field due to the presence of the PB onplane.Figure 4 shows the total gravitational potential map ofmodel 0.5 (left panel) and model 2 (right panel) on a portion ofthe equatorial plane. The total gravitational potential has been com-puted summing the separate contributions of the discs, halo and PB.The PB of model 2 contributes to the total gravitational potential sointensely that the potential well shifts towards the PB centre. In thiscircumstance, it is hard to imagine an equilibrium configuration inwhich the PB and the H I disc kinematic centre are off-set.A different perspective is given by the small insets of Fig. 4,where we show the circular speed of the NGC 5474 model (discsand halo), superimposed to the circular speed computed frommodel 0.5 (left panel) and model 2 (right panel). Of course, thesesystems have lost their cylindrical symmetry, so the concept of cir-cular speed makes no strict sense, but, at least in model 0.5 wherethe PB contribution to the potential is sub-dominant, this exercisestill helps to quantify the magnitude of the H I disc’s velocity fieldperturbations. Calling the ( x, y ) -plane the equatorial plane, and MNRAS000
484 kpc , m = 0 . , see Table 2). We refer to the two analytic models as,respectively, model 0.5 and model 2, where the number indicatesthe PB mass, in units of M (cid:12) . Using these models, we roughlyestimate the minimum and maximum distortions we may expect tosee in the H I disc’s velocity field due to the presence of the PB onplane.Figure 4 shows the total gravitational potential map ofmodel 0.5 (left panel) and model 2 (right panel) on a portion ofthe equatorial plane. The total gravitational potential has been com-puted summing the separate contributions of the discs, halo and PB.The PB of model 2 contributes to the total gravitational potential sointensely that the potential well shifts towards the PB centre. In thiscircumstance, it is hard to imagine an equilibrium configuration inwhich the PB and the H I disc kinematic centre are off-set.A different perspective is given by the small insets of Fig. 4,where we show the circular speed of the NGC 5474 model (discsand halo), superimposed to the circular speed computed frommodel 0.5 (left panel) and model 2 (right panel). Of course, thesesystems have lost their cylindrical symmetry, so the concept of cir-cular speed makes no strict sense, but, at least in model 0.5 wherethe PB contribution to the potential is sub-dominant, this exercisestill helps to quantify the magnitude of the H I disc’s velocity fieldperturbations. Calling the ( x, y ) -plane the equatorial plane, and MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 − x [kpc] − − − y [ k p c ] model 0.5 − x [kpc] − − − y [ k p c ] model 2 R [kpc] δv c , max = 6 km s − v c [km s − ] R [kpc] δv c , max = 28 km s − v c [km s − ] Figure 4.
Left-hand panel: total (halo, stellar disc, H I disc and PB) gravitational potential map in the discs’ equatorial plane when M PB = 0 . × M (cid:12) (model 0.5). The shades of blue, from light to dark, mark regions of increasing potential. The PB is located at ( x, y ) = (1 kpc , (black point). The orangecircles show distances corresponding to R e (inner) and R e (outer), with R e the PB’s effective radius (Table 2). Small inset at the bottom of the left panel:‘circular speed’ as a function of the distance from the H I disc’s kinematic centre of NGC 5474 without (black curve) and with (orange curve) the PB. In thelatter case the circular speed is computed following equation (16) along the principal axis y = 0 . The black points with errorbars show the rotation curvederived in Section 2.1.1. Right-hand panels: same as the left-hand panels, but for M PB = 2 × M (cid:12) (model 2). As a further comparison, in the inset ofthe left-hand panel we also show the circular speed obtained in the intermediate case of a PB with M PB = 10 M (cid:12) placed away from the galaxykinematic center (blue curve). Φ tot the model’s total gravitational potential, we define the ‘cir-cular speed’ v c ≡ (cid:115) x (cid:12)(cid:12)(cid:12)(cid:12) ∂ Φ tot ∂x (cid:12)(cid:12)(cid:12)(cid:12) y =0 . (16)The latter is computed for the models with and without the PB,along the line y = 0 , where the y = 0 axis is aligned with the PBcentre, when present. Then we measure δv c , max , i.e. the maximumdifference in circular speed between the models with and withoutthe PB, in a region of approximately R e (cid:39) . around thePB’s centre.In model 2 (Fig. 4, right panel) the PB produces distortions ashigh as δv c , max (cid:39)
28 km s − , which is not even consistent with therotation curve derived in Section 2.1.1 from the H I kinematics. Thisis not surprising since the PB centre is very close to the minimumvalue of the gravitational potential. While it is highly unlikely thatthe PB as in model 2 can be located onto the H I disc, given the largedistortion it generates over the wide area covering approximately R e , this is not excluded for model 0.5, especially due to the lackof kinematic information within R ∼ (Fig.s 2 and 4). In theinset in the left panel of Fig. 4 we also show the circular speedobtained when we place a PB with M PB = 10 M (cid:12) awayfrom the galaxy kinematic centre. The maximum distortions the PBgenerates in this case are as high as δv c , max (cid:39)
11 km s − , but stillthe overall profile is marginally consistent with the observed onewithin the errorbars.On this basis, we will restrict the range of possible PB stellarmasses to M PB ∈ [0 . , × M (cid:12) in any further analysis, sincewe expect more massive PB to have critical effects on the H I discof NGC 5474. We now focus on cases in which the PB moves along orbits co-planar with the galaxy discs. When on plane, we expect the dy-namical friction and the tidal force field of NGC 5474 to be themain drivers of the PB evolution. While the former makes the PBsink towards the galaxy centre on a relatively short timescale (fordetails, see Section 5.1.1), the latter is responsible for the PB massloss and the development of possible non-equilibrium features.We quantify the effects of the tidal force field of NGC 5474on the PB by computing its tidal radius r t . Often the tidal radius r t is estimated as (Binney & Tremaine 2008, equation 8.91) r t = R PB (cid:18) M PB M tot ( R PB ) (cid:19) / , (17)assuming that the size of the PB is negligible with respect to itsdistance from NGC 5474. In the above equation M tot ( R PB ) is thetotal mass of NGC 5474 enclosed within R PB , and R PB is the dis-tance of the PB from the centre of NGC 5474. For a PB with mass M PB = 0 . × M (cid:12) and R e = 0 .
484 kpc , the tidal radius is r t (cid:39) r h at R PB = 7 kpc . Considering that the PB mass enclosedwithin r h is (cid:39) of its total mass, we expect the tidal field ofNGC 5474 to be of small impact on the PB structural properties atany R PB (cid:38) .In this on-plane scenario it sounds then legitimate to take R PB = 7 kpc as PB initial position and v c ( R PB ) (cid:39)
42 km s − inthe azimuthal direction as initial velocity, i.e. a circular orbit. Evenif we considered a larger R PB , dynamical friction would anywaymake the PB sink within the disc eventually reaching withnegligible or minimal mass loss because of a larger r t . Whether thePB is the galaxy’s pseudo-bulge, or it is the remnant of an externalgalaxy, or it has formed in-situ in a burst of star formation happenedat least ago (compatibly with its dominant stellar population, MNRAS , 1–21 (2020)
R. Pascale et al.
B20), the observed, present-day PB would be the end-state of theorbital decay of any of these configurations.Given that the systems are extended, to follow the evolutionof the PB tidal radius from R PB = 7 kpc to the central regionswe rely on a more realistic approach. In a reference frame wherethe PB and the galaxy kinematic centre are aligned along the x -axis, we estimate the PB’s tidal radius r t by computing the position x = ( r t < R PB , , (i.e. on the equatorial plane and along the x -axis), where the effective potential Φ eff ( x ) = Φ tot ( x ) + Φ PB ( x − R PB ) −
12 (Ω r t ) (18)has a saddle point. Here, R PB ≡ ( R PB , , and Ω ≡ v c ( R PB ) /R PB , i.e. the angular speed obtained from the NGC 5474model circular speed (Fig. 2) at a distance R PB .Figure 5 shows the PB tidal radius as a function of the distancefrom the discs’ centre. We compute r t according equation (18) and,as a comparison, we also show r t computed according the classical(17). In addition to M PB = 0 . × M (cid:12) , we examined the casein which M PB = 10 M (cid:12) . For each mass value, we consideredthree cases with R e = 0 .
484 kpc , R e = 0 .
320 kpc , and R e =0 .
161 kpc , because, as an effect of the tidal force field, the PB maybecome more extended (see, e.g. Iorio et al. 2019). We notice thatequations (17) and (18) do not account for the PB mass loss, soeven if r t decreases along the orbit, the PB total mass is the sameas the initial one.According to equation (18), while the tidal radius of the lessextended PBs is always larger than at least three half-mass radii,in the remaining cases the tidal radius shrinks fast to less than twoPB half-mass radii at . We note that the tidal radius computedas in its classical formulation (equation 17) and as in equation (18)gives approximately the same results when the PB is sufficientlyfar from the galaxy’s centre, while equation (18) provides an esti-mate of r t sensibly lower when the PB is close to the galaxy centre.For the less massive ( M PB = 0 . × M (cid:12) ) and most extendedPB ( R e = 0 .
484 kpc ), the effective potential (18) does not evenhave a saddle point, meaning that the truncation radius is formallyzero. We do not show predictions for R PB < since we ex-pect the PB to have lost such a significant amount of mass to makeineffective also the use of equation (18). At least for the most ex-tended and least massive PB we may expect any effect due to thetidal force field of NGC 5474 to be extremely intense.To conclude, on the grounds of these analytic models, it seemsvery unlikely that the PB would be as massive as × M (cid:12) ,if placed on the discs’ plane of NGC 5474. If that would be thecase, we should be able to see strong distortions in the H I velocityfield map that are, instead, unseen. Moreover, although we haveconsidered the PB as only made of stars, we can interpret this resultas an upper limit on the total PB dynamical mass since the effectson the H I disc on which we have focused are purely gravitational.Even if we may expect the chance of survival of large off-centred PBs to be low, due to the galaxy’s intense tidal field force,we cannot either exclude or prove that more favorable initial con-ditions, as a less extended initial PB, may produce the required off-set, a smooth H I velocity field and a regular PB spatial distribution.Starting from the predictions of the analytic models we just exam-ined, we consider more quantitatively these scenarios through ourhydrodynamic N -body simulations, which are described in the fol-lowing sections. R PB [kpc] r t / r h M PB = 0 . × M (cid:12) R eff = 0 .
161 kpc R eff = 0 .
320 kpc R eff = 0 .
484 kpc R PB [kpc] M PB = 10 M (cid:12) Figure 5.
Left panel: ratio between tidal radius r t and the half-mass ra-dius r h of a PB with total stellar mass M PB = 0 . × M (cid:12) (and nodark matter) as a function of the distance from the system’s kinematic cen-tre. The tidal radius is computed following equation 17 (dashed curves)and according to equation 18 (solid curves). Each color refers to PBswith R e = 0 .
161 kpc (black curves), R e = 0 .
320 kpc (orange cuvers) R e = 0 .
484 kpc (purple curves). Right panel: same as the left panel, butfor a PB with mass M PB = 10 M (cid:12) . AREPO code
All our simulations are performed using the moving-mesh hydrody-namic code
AREPO (Springel 2010), as implemented in its publicly-released version (Weinberger et al. 2020). AREPO combines theadvantages of both Lagrangian smoothed particle hydrodynamics(SPH) and Eulerian hydrodynamics on an unstructured mesh withadaptive mesh refinement (AMR). The mesh is constructed froma Voronoi tessellation of a set of discrete points, used to solve thehyperbolic hydrodynamic equations with a finite-volume techniqueand it is free to move with the fluid flow. In our case, the mesh gridsizes are refined in such a way to ensure that each gas cell has ap-proximately constant mass, allowing one to sample with a largenumber of cells high density regions. The gas moving mesh is cou-pled to a particle-mesh algorithm and oct-tree approach (Barnes &Hut 1986) to solve the Poisson equation and compute both colli-sional and collisionless particles accelerations.
AREPO has been extensively employed to deal with a largenumber of astrophysical problems, such as AGN winds and feed-back (Costa et al. 2020), stellar evolution and interstellar mediumenrichment processes (van de Voort et al. 2020), spiral arms for-mation mechanisms (Smith et al. 2014), and run state-of-the-artlarge volume cosmological simulations of galaxy formation such asthe latest IllustrisTNG simulations (Naiman et al. 2018; Pillepichet al. 2018; Springel et al. 2018; Nelson et al. 2018; Marinacci et al.2018), or zoom-in cosmological magneto-hydrodynamical simula-tions as the ones from the Auriga project (Grand et al. 2017). For adetailed review see Weinberger et al. (2020) and references therein. https://arepo-code.org/ . MNRAS000
AREPO has been extensively employed to deal with a largenumber of astrophysical problems, such as AGN winds and feed-back (Costa et al. 2020), stellar evolution and interstellar mediumenrichment processes (van de Voort et al. 2020), spiral arms for-mation mechanisms (Smith et al. 2014), and run state-of-the-artlarge volume cosmological simulations of galaxy formation such asthe latest IllustrisTNG simulations (Naiman et al. 2018; Pillepichet al. 2018; Springel et al. 2018; Nelson et al. 2018; Marinacci et al.2018), or zoom-in cosmological magneto-hydrodynamical simula-tions as the ones from the Auriga project (Grand et al. 2017). For adetailed review see Weinberger et al. (2020) and references therein. https://arepo-code.org/ . MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 We sample the halo’s and discs’ initial conditions (hereafter ICs)following Springel et al. (2005). The dark-matter halo density fol-lows the Hernquist (1990) model ρ dm ( r ) = M tot π ar r + a ) , (19)whose mass profile is given by M dm ( r ) = M tot r ( a + r ) , (20)where a is the Hernquist’s scale radius and M tot the halo totalmass.Dark-matter haloes are often represented with NFW models(equation 1). However, the use of an Hernquist model (which hasthe same inner slope as the NFW model) is motivated by the factthat it has finite mass, and an analytic, ergodic distribution function(hereafter DF), which simplifies the sampling of the halo particlevelocities. To link the Hernquist model to the NFW profile we re-quire that the Hernquist total mass is equal to the NFW halo virialmass M vir (i.e. the enclosed mass within the virial radius), and weimpose that the two profiles share the same normalization in thecentral parts. As a consequence, provided, for instance, M vir and r s for the NFW halo, the corresponding Hernquist halo is fixed withparameters M tot = M vir a = r s (cid:112) c ) − c/ (1 + c )] . (21)(see Fig. 1 of Springel et al. 2005)The radial density profiles of the gaseous and stellar discs fol-low from equations (2) and (3), respectively, where for the gas wehave switched from particle number density to mass density. Wedrop the razor-thin disc approximation and let the discs have a non-negligible thickness. For the stellar disc, the vertical profile strati-fies with radially constant scale height z (cid:63) , so that its full three di-mensional density distribution is given by ρ (cid:63) ( R, z ) = M (cid:63) πh (cid:63) z (cid:63) e − (cid:0) Rh(cid:63) (cid:1) sech (cid:18) zz (cid:63) (cid:19) . (22)The vertical profile of the H I disc is determined from the ver-tical hydrostatic equilibrium ∂ Φ tot ∂z = − ρ gas ∂P∂z . (23)In the above equation Φ tot is the total gravitational potential (stellardisc, H I disc and dark matter), P is the thermal pressure of the gas,assumed to be isothermal. For any chosen Φ tot , ρ gas is constrainedrequiring Σ gas ( R ) = (cid:90) + ∞−∞ ρ gas ( R, z ) d z, (24)where Σ gas is an exponential disc model as in equation (2), ex-pressed in terms of surface density. The total potential is deter-mined iteratively, following the scheme of Springel et al. (2005). For simplicity, only for the dark-matter halo, we assume Φ tot =Φ dm (i.e. the dark-matter potential) and we draw the halo phase-space positions directly from the Hernquist analytic isotropic DF. We assume that the stellar disc DF depends only on the energy E and the third component of the angular momentum L z . Hence,the only non-vanishing moments of the stellar disc DF are (cid:104) v R (cid:105) = (cid:104) v z (cid:105) ≡ σ , (cid:104) v φ (cid:105) and (cid:104) v φ (cid:105) . We compute σ and (cid:104) v φ (cid:105) from the Jeansequations, and, to sample the radial and vertical components of thestellar particle velocities, we assume that the velocity distributionsin these two directions are Gaussian, with dispersion equal to σ .To compute the streaming velocity (cid:104) v φ (cid:105) we rely on the epicyclicapproximation, so (cid:104) v φ (cid:105) = (cid:115) (cid:104) v φ (cid:105) − (cid:104) v R (cid:105) η , (25)where η = 4 R ∂ Φ tot ∂R (cid:18) R ∂ Φ tot ∂R + ∂ Φ tot ∂R (cid:19) − . (26)The azimuthal component of the stellar disc particles are then sam-pled from a Guassian with (cid:104) v φ (cid:105) as mean and the r.m.s. velocity (cid:113) (cid:104) v φ (cid:105) − (cid:104) v φ (cid:105) as standard deviation.The gas velocity field is instead composed only by the az-imuthal component v φ, gas which satisfies the stationary Euler equa-tion v φ, gas = R (cid:18) ∂ Φ tot ∂R + 1 ρ gas ∂P∂R (cid:19) . (27) The reference model derived in Section 2.1 (see Table 1) fixes al-most all the degrees of freedom needed to set the NGC 5474-likemodel of our simulations. We further adopt a stellar scale height z (cid:63) = 0 . h (cid:63) (Kregel et al. 2002; Oh et al. 2015) while the gasis mono-atomic, with T /µ = 3400 K ( T the gas temperature and µ the gas mean molecular weight). As a reference, for a neutral,hydrogen gas, this corresponds to T = 3400 K .We require all particles to have the same mass m part =5000 M (cid:12) . With this choice, given M vir = 3 . × M (cid:12) ,M gas = 1 . × M (cid:12) ,M (cid:63) = 4 . × M (cid:12) , (28)it follows N halo = 708750 ,N gas = 364000 ,N (cid:63) = 82200 , (29)where N halo , N gas and N (cid:63) indicate, respectively, the number ofparticles of the halo, the H I disc and the stellar disc .Since the dark-matter halo particles of NGC 5474 have beensampled as if the halo were in isolation (i.e. not accounting forthe contribution of the discs to the total gravitational potential), thestellar disc was built in Maxwellian approximation and the discsprovide a non-negligible contribution to the total gravitational po-tential, we expect all the models components to be close to equilib-rium, but not exactly in equilibrium. To check how they respond tothe presence of each other, and to let them shift towards an equilib-rium state, we first run a simulation where NGC 5474 is evolved in The dark halo of NGC 5474 is not sampled at radii larger than r vir , sothe total mass represented with particles is (cid:39) . M vir .MNRAS , 1–21 (2020) R. Pascale et al. isolation. The main features and the results of this simulation aredescribed in Appendix A. In all the following hydrodynamic N -body models, we will take as ICs of NGC 5474 the output of thesimulation of Appendix A after .
98 Gyr . The phase-space positions of the PB stellar and dark-matter parti-cles are drawn directly from the components ergodic DFs. Start-ing from equation (11), through an Abel inversion we retrievethe intrinsic density distribution ρ (cid:63), PB of the stellar component.We complete the stellar and dark-matter density-potential pairs( ρ (cid:63), PB , Φ (cid:63), PB ) and ( ρ dm , PB , Φ dm , PB ) solving the Poisson equa-tion ∇ Φ ∗ , PB = 4 πGρ ∗ , PB , where ∗ = ( (cid:63), dm) .By means of an Eddington inversion (Binney & Tremaine2008) we compute numerically the ergodic DFs of stars ( f (cid:63), PB )and dark matter ( f dm , PB ) as f ∗ , PB ( E ) = 1 √ π dd E (cid:90) E d Φ PB √ E − Φ PB d ρ ∗ , PB d Φ PB , (30)where ∗ ∈ { (cid:63), dm } , while Φ PB is the total potential Φ PB =Φ (cid:63), PB + Φ dm , PB . In case the PB is only made by stars, Φ PB =Φ (cid:63), PB and ∗ = (cid:63) .Since in our simulations we will consider PBs with and with-out dark matter, and with stellar components ranging over differ-ent sizes and masses, for clarity, we will separately list in Sections5.1.1 and 5.2.1 the PB parameters and the number of particles usedin each simulation. In our first set of simulations we explore cases in which the PBmoves within the galaxy discs’ plane, and we check whether theoff-set can be reproduced as an outcome of the simulations keepingthe shape of the PB smooth and regular and the kinematics of thegalaxy’s gaseous component unperturbed as observed.Starting from the conclusions of Section 3: • we consider a PB made only of stars, without dark-matterhalo; • the PB centre of mass is at an initial distance R PB = 7 kpc from the galaxy centre, on a circular orbit with an initial azimuthalvelocity v c ( R PB ) , co-rotating with the H I and stellar discs. We ex-pect the PB orbit to shrink because of dynamical friction and thusto reach R PB ≈ ; • we focus on a PB with total initial stellar mass M PB =0 . × M (cid:12) , M PB = 10 M (cid:12) and M PB = 1 . × M (cid:12) .For each mass, we consider PBs with initial R e = 0 .
484 kpc , R e = 0 .
320 kpc and R e = 0 .
161 kpc . The case of M PB =1 . × M (cid:12) is intended to account for the fact that, due tomass loss, the PB can reach with less than the upper limitof M (cid:12) that we estimated in Section 3.While we have required that the particles of all the componentsof NGC 5474 must have m part = 5000 M (cid:12) , we relax this con-dition on the PB and we set its particles to be three times lessmassive. This allows us to sample the PB’s phase space with asufficiently large number of particles and to avoid an overwhelm-ingly high number of particles per simulation, which would just be computationally expensive with no particular gain in terms of ac-curacy. The PBs with M PB = 0 . × M (cid:12) , M PB = 10 M (cid:12) and M PB = 1 . × M (cid:12) are sampled with N PB = 30000 , N PB = 60000 and N PB = 90000 particles, respectively, follow-ing the scheme of Section 4.3. When these PBs are evolved in iso-lation for
10 Gyr they keep their equilibrium configuration.We expect the PB to sink towards the system centre due to dy-namical friction on a timescale t fric , which we estimate as (Binney& Tremaine 2008) t fric = 1 . M tot ( R PB ) M PB t cross , (31)where R PB is the PB distance from the centre, ln Λ is the Coulomblogarithm, t cross ≡ R PB /v c ( R PB ) is the crossing time at R PB and M tot is the total mass enclosed within R PB . At R PB = 7 kpc ,with v c ( R PB ) (cid:39)
42 km s − , we get t fric = 1 . − . when M PB = 0 . × M (cid:12) ,t fric = 0 . − .
67 Gyr when M PB = 10 M (cid:12) ,t fric = 0 . − .
15 Gyr when M PB = 1 . × M (cid:12) , (32)The lower and upper limits over t fric are obtained assuming thetypical values ln Λ ∼ and ln Λ ∼ , respectively. According tothese estimates we may expect the PB to decay towards the centerof NGC 5474 on a very short timescale.We run a total of 9 simulations and, in each of them, the ICs ofNGC 5474 correspond to the equilibrium simulation of Appendix Aafter .
98 Gyr . After .
98 Gyr the center of mass of NGC 5474 isin the origin of the system reference frame. Details on simulationsand NGC 5474 parameters (e.g. softenings, number of particles)are listed in Tables 1 and A1. We will refer to these simulationsas PB Re X M Y , where X = 484 , , indicates the PB ef-fective radius in pc, and Y = 0 . , , . is the PB mass in unitsof (see also Table 3). The simulations run for . and weuse an adaptive timestep refinement with typical timestep values (cid:39) . . Figure 6 shows the trajectories of the PBs in each of the nine sim-ulations. Each orbit (red curve) is obtained connecting the centreof mass of the PB from consecutive snapshots and, alongside theorbit, each panel also shows the projected density distribution ofthe PB taken at few representative snapshots. The ( X, Y ) -plane isthe plane of the orbit, co-planar with the discs’ plane. We point outthat in Fig. 6 we have assumed the galaxy’s symmetry axis as lineof sight, but, as long as the PB evolves on the discs plane, the term cos i provides a negligible correction for i = 21 ◦ , and we can any-way project the galaxy in such a way to align the off-set with oneof the galaxy’s principal axes.As expected, the PBs with the largest initial R e are distortedthe most by the galactic tidal field. At R (cid:39) , the PB of modelPB Re484 M0.5 has: i) lost approximately 60% of its mass; ii) lostits spherical symmetry in favour of the formation of an elongatedstructure that has just started to wrap the galaxy centre; iii) de-veloped extended and massive tidal debris, formed from the verybeginning of the simulations. We find a similar outcome also inmodels PB Re484 M1 and PB Re484 M1.5, notwithstanding thehigher PB mass which should make, in principle, the PB more re-sistant against the galaxy tidal force field. To estimate the massloss we consider as particles belonging to the PB those that remainwithin 3 r h ( r h is the PB stellar half-mass radius in the ICs). As MNRAS000
98 Gyr the center of mass of NGC 5474 isin the origin of the system reference frame. Details on simulationsand NGC 5474 parameters (e.g. softenings, number of particles)are listed in Tables 1 and A1. We will refer to these simulationsas PB Re X M Y , where X = 484 , , indicates the PB ef-fective radius in pc, and Y = 0 . , , . is the PB mass in unitsof (see also Table 3). The simulations run for . and weuse an adaptive timestep refinement with typical timestep values (cid:39) . . Figure 6 shows the trajectories of the PBs in each of the nine sim-ulations. Each orbit (red curve) is obtained connecting the centreof mass of the PB from consecutive snapshots and, alongside theorbit, each panel also shows the projected density distribution ofthe PB taken at few representative snapshots. The ( X, Y ) -plane isthe plane of the orbit, co-planar with the discs’ plane. We point outthat in Fig. 6 we have assumed the galaxy’s symmetry axis as lineof sight, but, as long as the PB evolves on the discs plane, the term cos i provides a negligible correction for i = 21 ◦ , and we can any-way project the galaxy in such a way to align the off-set with oneof the galaxy’s principal axes.As expected, the PBs with the largest initial R e are distortedthe most by the galactic tidal field. At R (cid:39) , the PB of modelPB Re484 M0.5 has: i) lost approximately 60% of its mass; ii) lostits spherical symmetry in favour of the formation of an elongatedstructure that has just started to wrap the galaxy centre; iii) de-veloped extended and massive tidal debris, formed from the verybeginning of the simulations. We find a similar outcome also inmodels PB Re484 M1 and PB Re484 M1.5, notwithstanding thehigher PB mass which should make, in principle, the PB more re-sistant against the galaxy tidal force field. To estimate the massloss we consider as particles belonging to the PB those that remainwithin 3 r h ( r h is the PB stellar half-mass radius in the ICs). As MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 Table 3.
Main input parameters of the set of simulations of Section 5.1. From the left-hand to the right-hand column: name of the model (model’s name); PBeffective radius ( R e ); PB total stellar mass ( M PB ); PB number of particles ( N PB ); softening used for the PB particles ( l (cid:63), PB ). The softening is computedrequiring that the maximum force between the PB’s particles should not be larger than the PB’s mean-field strength (Dehnen & Read 2011). We notice thatall the models components (halo, stellar disc, gas disc, PB) have different softenings. The PB particles have mass m part = 1667 M (cid:12) . In each simulation theinitial position of the PB centre of mass is at R PB = 7 kpc , with initial streaming velocity v c ( R PB ) = 42 km s − . The ICs of NGC 5474 correspond to theconfiguration of Appendix A taken after .
98 Gyr (for further details see Tables 1 and A1).Models’ name R e [ kpc ] M PB [ M (cid:12) ] N PB l (cid:63), PB [ kpc ]PB Re484 M1.5 0.484 1.5 90000 0.029PB Re484 M1 0.484 1 60000 0.033PB Re484 M0.5 0.484 0.5 30000 0.042PB Re320 M1.5 0.320 1.5 90000 0.019PB Re320 M1 0.320 1 60000 0.022PB Re320 M0.5 0.320 0.5 30000 0.028PB Re161 M1.5 0.161 1.5 90000 0.0096PB Re161 M1 0.161 1 60000 0.011PB Re161 M0.5 0.161 0.5 30000 0.014 anticipated in Section 3, and looking at Fig. 5, this result is not sur-prising given that the PB tidal radius is less than its initial half-massradius at best.For each simulation, Fig. 7 shows the projected distance be-tween the PB centre of mass and the galaxy centre, and the PBbound stellar mass fraction as a function of time. The systemsare projected as in Fig. 6, assuming as line of sight the symme-try axis, so R ≡ √ X + Y . The main driver of the PB evo-lution is the dynamical friction: in a very short time-scale (lessthan . ; see also equation 32) all the PBs reach R ∼ and, as expected, the most massive PBs sink faster than the leastmassive ones (equation 31; top panel of Fig. 7). Among the PBswith initial R e = 0 .
320 kpc , those of models PB Re320 M1 andPB Re320 M1.5 reach R = 1 kpc losing only 20% of their orig-inal mass, while that of model PB Re320 M0.5 has experiencedsubstantial mass loss already at R (cid:39) .
54 kpc . The most com-pact PBs (PB Re161 M0.5, PB Re161 M1 and PB Re161 M1.5),instead, provide a great resistance against the galaxy tidal forcefield ( r t (cid:39) R e , see Fig. 5), and reach the galaxy centre losingfrom 10% to 30% of their total initial mass, without developingsignificant tidal tails.Based on the aforementioned features, we exclude from anyfurther analysis the PBs with initial R e = 0 .
484 kpc and the onefrom model PB Re320 M0.5, which immediately depart from equi-librium and are severely distorted by the strong tidal force fieldof NGC 5474. We notice that when the initial mass is M PB =1 . × M (cid:12) , in some cases the PB reaches with less than M (cid:12) (the mass upper limit estimated in Section 3). For this rea-son, we ran additional simulations with a PB initial mass as highas M PB = 2 × M (cid:12) . Even though these configurations reachthe galactocentric distance with a bound mass of M (cid:12) ormore, when R e = 0 .
484 kpc the PB still develops pronounced tidalfeatures, when R e = 0 .
320 kpc and R e = 0 .
161 kpc it presentsthe same structural properties of its . × M (cid:12) analogs, and theyare not shown here for the sake of synthesis.To make any comparison between simulations and observa-tions coherent, in Fig. 8 we add the contribution due to the stel-lar disc of NGC 5474 to the PB projected density maps of someof the remaining models, taken when they are away fromthe galaxy centre (i.e. the orbits end-point of Fig. 6). As prototypi-cal cases, we selected models PB Re161 M0.5, PB Re161 M1 andPB Re320 M1.5. The greyscale extends over two order of magni-tudes from the highest density peak, which would be compatible with a difference of five magnitudes from the map brightest pointif we assume the same mass-to-light ratio for all the particles. Inthis case, the systems have been projected assuming i = 21 ◦ , andwe have called the plane of the sky the ( ξ, η ) -plane, with ξ ≡ X .For instance, such an image displays patterns that are somehowreminiscent of the galaxy stellar map of Fig. 1 or Fig. 5 by B20,both obtained through the LEGUS (Calzetti et al. 2015) photomet-ric catalogue based on HST ACS images. In all cases, the pertur-bation provided by the PB induces the formation of a loose spiralstructure in a region − around the discs’ centre. The spi-ral arms are similar to the observed pattern of NGC 5474 (B20),they are not necessarily symmetric with respect to the discs’ cen-tre and also the gaseous component develops a spiral structure thatfollows the optical stellar disc (R04). In the small insets of Fig. 8we show the corresponding H I line-of-sight velocity field maps asin the main panels. The velocity maps are derived using the samespatial and velocity resolutions as the one from the R04 map: pix-els .
33 kpc × .
33 kpc wide and velocity contours separated by − . The solid black lines and redder colors mark the disc’sapproaching arm, while the black dashed curves and bluer colorsindicate the receding arm. Once the resolution has been down-graded to the same one of the observations, almost all the H I ve-locity maps appear smooth and regular, and consistent with R04.A different perspective is given by the H I rotation curves shown inthe bottom panels of Fig. 8. Most of the differences that are clearlyvisible between the rotation curves from the simulations and the an-alytic model are confined within − , where we barely havekinematic information. However, as predicted in Section 3, the ro-tation curves start to show appreciable differences with respect tothe de-projected rotation curve we derived when the PB mass isclose to (cid:39) M (cid:12) .We rederived some of the structural parameters of the PBsfrom Fig. 8 fitting their projected density distribution with aS´ersic model, as in Fisher & Drory (2010) and B20, and comput-ing their average axis-ratio q through elliptical isodensity contours.We measure the axis ratio q ≡ c/a , with c and a the semi-minorand semi-major axes, respectively (see Lau et al. 2012), computingthe inertia tensor of the projected densities maps in an ellipse withsemi-major axis R e , centred on the map densest point. To producethe projected density distribution: i) we evaluate the PB centre onthe plane of the sky as the densest/brightest point; ii) we bin theparticles in 50 circular annuli, linearly equally spaced, and extend-ing out to R e , with R e as in the models’ ICs; iii) we subtracted the MNRAS , 1–21 (2020) R. Pascale et al. − . − . . . . . X [kpc] − − − Y [ k p c ] d max = 2 .
18 kpc t max = 0 .
83 GyrPB Re484 M0.5 − . − . . . . . X [kpc] d max = 1 .
54 kpc t max = 0 .
93 GyrPB Re320 M0.5 − . − . . . . . X [kpc] d max = 0 .
94 kpc t max = 1 .
03 GyrPB Re161 M0.5 − . − . . . . . X [kpc] − − − Y [ k p c ] d max = 1 .
18 kpc t max = 0 .
93 GyrPB Re484 M1 − . − . . . . . X [kpc] d max = 1 .
06 kpc t max = 0 .
83 GyrPB Re320 M1 − . − . . . . . X [kpc] d max = 0 .
77 kpc t max = 0 .
88 GyrPB Re161 M1 − . − . . . . . X [kpc] − − − Y [ k p c ] d max = 1 .
10 kpc t max = 0 .
78 GyrPB Re484 M1.5 − . − . . . . . X [kpc] d max = 1 .
00 kpc t max = 0 .
78 GyrPB Re320 M1.5 − . − . . . . . X [kpc] d max = 1 .
16 kpc t max = 0 .
78 GyrPB Re161 M1.5
Figure 6.
Trajectories of the centre of mass of the PBs (red curves) in all the hydrodynamical N -body models considered in Section 5.1. The top, middleand bottom rows of panels refer to models whose PB has an initial stellar mass M PB = 0 . × M (cid:12) , M PB = 10 M (cid:12) and . × M PB = 10 M (cid:12) ,respectively. The initial PB effective radius decreases from the left to the right column of panels. In each panel we also show the PB spatial density distributiontaken at few representative snapshots along its orbit, projected along the symmetry axis (so the ( X, Y ) -plane is the equatorial plane). In most cases, thetrajectory is drawn until the PB reaches a distance of ∼ from the centre (black circle). The PB centre is determined using the shrinking sphere method(Power et al. 2003). Details on simulation parameters in Tables 1, 3 and A1. stellar disc contribution, estimated from a wider and more distantconcentric circular annulus.A S´ersic model is specified by the parameters { R e , m, M PB } ,as in equation (11) of Section 2.2, replacing surface brightness withstellar surface mass density, assuming constant mass-to-light ratio.The model’s log-likelihood is ln L = − (cid:88) i (cid:18) Σ( R s ,i ) − Σ s ,i δ Σ s ,i (cid:19) , (33) where the points { R s ,i , Σ s ,i , δ Σ s ,i } are the PB projected densityprofile, and the sum extends over the total number or radial bins.The fit is performed using a Bayesian approach relying on MCMC,adopting the same scheme as in Section 2.1.2 and using flat priorsover the model parameters. We list the parameters resulting fromthe fit in Table 4.On the plane of the sky, the PB of NGC 5474 appears roundand regular ( q (cid:39) ). In our analysis, the only PB that keeps its MNRAS000
Trajectories of the centre of mass of the PBs (red curves) in all the hydrodynamical N -body models considered in Section 5.1. The top, middleand bottom rows of panels refer to models whose PB has an initial stellar mass M PB = 0 . × M (cid:12) , M PB = 10 M (cid:12) and . × M PB = 10 M (cid:12) ,respectively. The initial PB effective radius decreases from the left to the right column of panels. In each panel we also show the PB spatial density distributiontaken at few representative snapshots along its orbit, projected along the symmetry axis (so the ( X, Y ) -plane is the equatorial plane). In most cases, thetrajectory is drawn until the PB reaches a distance of ∼ from the centre (black circle). The PB centre is determined using the shrinking sphere method(Power et al. 2003). Details on simulation parameters in Tables 1, 3 and A1. stellar disc contribution, estimated from a wider and more distantconcentric circular annulus.A S´ersic model is specified by the parameters { R e , m, M PB } ,as in equation (11) of Section 2.2, replacing surface brightness withstellar surface mass density, assuming constant mass-to-light ratio.The model’s log-likelihood is ln L = − (cid:88) i (cid:18) Σ( R s ,i ) − Σ s ,i δ Σ s ,i (cid:19) , (33) where the points { R s ,i , Σ s ,i , δ Σ s ,i } are the PB projected densityprofile, and the sum extends over the total number or radial bins.The fit is performed using a Bayesian approach relying on MCMC,adopting the same scheme as in Section 2.1.2 and using flat priorsover the model parameters. We list the parameters resulting fromthe fit in Table 4.On the plane of the sky, the PB of NGC 5474 appears roundand regular ( q (cid:39) ). In our analysis, the only PB that keeps its MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 R [ k p c ] R eff =0.484 kpc0 . . . N r h / N t o t R [ k p c ] R eff =0.320 kpc0 . . . N r h / N t o t R [ k p c ] R eff =0.161 kpc0 . . . t [Gyr]0.20.61 N r h / N t o t Figure 7.
Top two panels: distance between the centre of mass of the PB andthe galaxy’s centre as a function of time (top) and fraction of mass enclosedwithin r h as a function of time (bottom), with r h the PB half-mass radiusas in the ICs. The green curves refer to PBs with initial R e = 0 .
484 kpc .Middle two panels: same as the top panels, but for PB with initial R e =0 .
320 kpc (orange curves). Bottom two panels: same as the top panels, butfor PB with initial R e = 0 .
161 kpc (purple curves). The solid, dashedand dotted curves refer, respectively, to the PBs with initial mass M PB =0 . × M (cid:12) , M PB = 10 M (cid:12) and . × M PB = 10 M (cid:12) . The curveslighten when the PBs have reached a distance from the system’s centre of ≈ (colored circles). circular symmetry is in model PB Re161 M1 ( q (cid:39) . ), while weestimate q (cid:39) . and q (cid:39) . for models PB Re161 M0.5 andPB Re320 M1.5, respectively. According to the S´ersic fit, the PBswith initial R e = 0 .
161 kpc are more extended, on average, bya factor of ∼ . with respect to the initial system, while the PBwith initial R e = 0 .
320 kpc by a factor 1.15 (see Table 4). As areference, the left panel of Fig. 9 shows the projected density pro-file of the PB from model PB Re320 M1.5, together with the bestfitting S´ersic model. In this case, the large errorbars are caused byhaving imposed circular symmetry in the derivation of the profile,even though the system is not strictly circularly symmetric.The case of model PB Re320 M1.5 is probably the most in-triguing one. Even if its PB is too flattened to look like the ob-served stellar system of NGC 5474 ( q (cid:39) . ), the fit withthe S´ersic model provides R e = 0 . ± .
02 kpc , the closest to R e = 0 .
484 kpc among the cases explored. Since in Fig. 8 wehave assumed the same mass-to-light ratio for the stellar disc andPB particles, it may be worth asking how would the PB look like ifwe made its particles more luminous. Assuming different mass-to-light-ratios Υ PB and Υ disc between stars of the PB and the galaxy’sstellar disc, respectively, the middle and right panels of Fig. 9 showthe surface brightness maps obtained from a zoom of the right panelof Fig. 8. In the 3.6 µ m band, if Υ PB ranges between 0.5 and 0.8, Table 4.
Models’ parameters as a result of the fit with the S´ersic modelof Section 5.1.2. The columns, from left to the right, list: the model’sname; the S´ersic effective radius ( R e ); the S´ersic index ( m ); the PB totalmass ( M PB ); the axis ratio ( q ). For the parameters of the S´ersic model, wetake as measure of that parameter the th percentile of its marginalizedone-dimensional distribution, while the uncertatites are estimated from the16 th , 50 th and 84 th percentiles of the corresponding marginalized one-dimensional distribution.Model PB Re161 M0.5 PB Re161 M1 PB Re320 M1.5 R e / kpc 0 . ± .
01 0 . ± .
01 0 . ± . m . +0 . − . . ± .
02 0 . ± . M PB M (cid:12) . ± .
03 7 . ± .
01 8 . ± . q .
67 0 .
90 0 . then Υ PB / Υ disc = 2 (middle panel) implies a disc stellar popula-tion (cid:39) − old (see also Sections 2.1 and 2.2). Although veryextreme, we also examined the case with Υ PB / Υ disc = 3 (rightpanel). The shades of colours are the same as in Fig. 8. For thesePBs we have rederived the axis ratios and the structural parametersresulting from a S´ersic fit, but we do not find any significant differ-ence with respect to the case with uniform Υ in axis ratio, size andconcentration.To summarize, we find that a system with a large initial sizesteps away from its equilibrium state on a timescale so short tomake very unlikely the possibility that it represents the observedPB of NGC 5474. It seems implausible that these systems may pro-duce the observed off-set as the result of orbital decay as well as itseems implausible that the PB, if misplaced from the galaxy centre,would last long enough without developing visible debris or tidaltails. Furthermore, in this latter case, the problem of the mechanismthat may have caused the off-set would still be an open question.The most compact PBs, at least over the range of masses explored,move towards the centre without producing detectable perturbationin the H I velocity field but, in spite of this, none of them reproducethe observed properties of the PB of NGC 5474. Not even the com-promise of a massive and intermediate sized system (which can inprinciple resist the tidal force field, and expand to the required sizeduring its evolution) is anyway close to the desired outcome. Infact, we find that a stellar component with these properties flattensto an extent inconsistent with the observations, independently onthe luminosity of its stellar population.We do not expect an increase of the initial mass to improvethe chances of survival of the PB: even though it would make itmore resilient against the gravitational tidal field force, accordingto Section 3, the smoothness of the H I velocity field would start tobe compromised. Instead, objects less massive than . × M (cid:12) would have a final mass (measured when the system is off-centred) less than the one we would expect on the basis of the lu-minosity of the PB stellar population, as discussed in Section 2.2.Even if we limited ourselves in studying only few specific or-bits, as long as PB moves within the galaxy’s equatorial plane, itseems likely that different orbits (for instance, more radial orbits,or with low inclination) would not behave so differently from theones that we have considered given the dominant effects of dynam-ical friction and the strong tidal force field of NGC 5474. MNRAS , 1–21 (2020) R. Pascale et al. − − ξ [kpc] − − η [ k p c ] PB Re161 M0.5 − − ξ [kpc] PB Re161 M1 − − ξ [kpc] PB Re320 M1.5 R [kpc] v [ k m s − ] R [kpc] R [kpc] model Observation Simulation − − . . . - - - − − . . . - - - − − . . . - - - Figure 8.
Top left panel: total (disc and PB) stellar projected density map computed from the configuration corresponding to the orbit’s end-point of thehydrodynamical N -body model PB Re0.161 M0.5 as in Fig. 6. The system has been projected assuming an inclination of i = 21 ◦ , as in R04. We showintensity contours equal to Σ max / n with Σ max the map’s densest peak and n = 1 , ..., . The blue and white dots show, respectively, the discs kinematiccentre and the PB centre and the full projected orbit is shown with a red line. We notice that the disc kinematic center also corresponds to the centre of the ( ξ, η ) plane. The small inset shows the H I line-of-sight velocity map, derived in the same portion as in the main panel. The H I velocity map has been obtainedas in R04, binning with pixels .
33 kpc × .
33 kpc wide, once we have assumed the distance d = 6 .
98 Mpc , and with velocity contours separated by − . The approaching arm is shown with red colors and solid black curves, while the receding arm with blue colors and dashed-black curves. Bottomleft panel: H I circular speed as a function of the distance from the galaxy kinematic center (blue points with error bars) from the same snapshot as in the topleft panel, compared to the galaxy’s deprojected rotation velocity curve computed in Section 2.1.1 (black points with error bars). The orange curve shows thecircular speed of the analytic model of NGC 5474. Middle panels: same as the left panels but for the N -body model PB Re0.161 M1. Right panels: same asthe left panels but for the N -body model PB Re0.320 M1.5. Details on simulation parameters in Tables 1, 3 and A1. . . . R [kpc] Σ [ M (cid:12) k p c − ] t int = 0 .
78 Gyr
PB Re320 M15 reference model R eff = 0 .
32 kpcReal PBsofteningSimulation − ξ [kpc] − η [ k p c ] Υ PB / Υ disc =2 − ξ [kpc] Υ PB / Υ disc =3 Figure 9.
Left panel: projected density profile of the PB computed from the density map of the right panel of Fig. 8 (black points with error bars), superimposedto the reference best fitting model, obtained as in Section 5.1.2 (red dashed curve). As a comparison, we show the S´ersic model resulting from the fit of B20(blue curve), so m = 0 . , R e = 0 .
484 kpc , while we have imposed the same total mass as the best model ( M PB (cid:39) M (cid:12) , see Table 4). The verticalblack-dashed line marks the effective radius of the reference model. Middle panel: total stellar (disc and PB) surface brightness map as in the right hand panelof Fig. 8, where we have assumed a different mass-to-light-ratio ( Υ ) for stars belonging to the stellar disc and the PB. The PB is twice as luminous as thestellar disc. The blue dot, also the centre of the ( ξ, η ) plane, shows the kinematic centre of the stellar and H I discs. Right panel: same as the middle panel, butthe PB particles are three times as luminous as the stellar disc ones. MNRAS000
484 kpc , while we have imposed the same total mass as the best model ( M PB (cid:39) M (cid:12) , see Table 4). The verticalblack-dashed line marks the effective radius of the reference model. Middle panel: total stellar (disc and PB) surface brightness map as in the right hand panelof Fig. 8, where we have assumed a different mass-to-light-ratio ( Υ ) for stars belonging to the stellar disc and the PB. The PB is twice as luminous as thestellar disc. The blue dot, also the centre of the ( ξ, η ) plane, shows the kinematic centre of the stellar and H I discs. Right panel: same as the middle panel, butthe PB particles are three times as luminous as the stellar disc ones. MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 − ξ [kpc] − η [ k p c ] stellar distribution − ξ [kpc]H i distribution − ξ [kpc] - - - H i velocity field Figure 10.
Left panel: stellar projected density map (stellar disc and PB) from the hydrodynamic N -body model PB Re484 M1 taken after t int = 0 .
83 Gyr .The system has been projected as in Fig. 8, assuming i = 21 ◦ , with ξ ≡ X . The centre of the ( ξ, η ) plane is the kinematic centre of the gaseous disc(blue dot), while the intensity contours are at Σ max / n with Σ max the map’s densest peak and n = 1 , ..., . The orange ellipse shows the position of theover-density. Middle panel: H I projected density map. The colour scale is such that yellow corresponds to high-density regions while purple to low-densityregions. As a comparison, the white curves show the stellar isodensity contours as in the left panel. For clarity, we only show stellar isodensities Σ max / n corresponding to n = 1 , . Right panel: H I velocity field map. The solid curves and the redder colours mark the disc’s approaching arm while the dashedcurves and the bluer colours the disc’s receding arms. The contours are separated by − and each pixel is .
33 kpc × .
33 kpc wide, as in R04.
The left panel of Fig. 10 shows the projected stellar density mapfrom model PB Re484 M1 after (cid:39) . We show a zoomed-in view of the discs’ centre, the line of sight is inclined of i =21 ◦ with respect to the symmetry axis, and we have used the samescheme of colours as in Fig. 8. We have shown that the PB of thehydrodynamical N -body model PB Re484 M1 is disrupted by thetidal force field of NGC 5474 and perturbs the central stellar andgas distribution of NGC 5474 while sinking towards the centre. Inthe left panel of Fig. 10, we mark with an orange ellipse what isleft of it after , when its centroid is ∼ away from thegalaxy’s centre.We cannot help pointing out the similarities between thisstructure and the SW over-density of NGC 5474. We mentionedthe SW over-density as one of the peculiarities of NGC 5474: itis a large substructure extending to the South-West of the PB ofNGC 5474 (see Fig. 1), mostly dominated by old-intermediate agestars, whose structure is not associated with the overall spiral pat-tern (B20). When the PB of model PB Re484 M1 is dismembered,its remnants shape into an elongated and wide structure appearingdenser than the stellar disc’s centre. The over-dense region partiallybrightens the galaxy’s spiral structure with its tidal tails, and thespiral arms are also slightly traced by the H I distribution (middlepanel). As noted by B20, the SW over-density seems to be corre-lated to a local minimum of the H I distribution (K00), which ispartially consistent with Fig. 10. We find a similar configurationalso in model PB Re320 M0.5.Reproducing the observed properties of the SW over-densityof intermediate-old stars of NGC 5474 is beyond the scope of thiswork and it would require a systematic search of the wide param-eter space (different orbits inclinations, eccentricity, initial veloc-ity etc). However, as a by product, our simulations have producedconfigurations that are worth commenting as a viable channel forthe formation of this substructure and, in general, of the substruc- tures traced by old-intermediate age stars in the disc of NGC5474.An interaction with M 101 may not be enough to explain the SWover-density (B20; Mihos et al. 2012b) and we showed that the hy-pothesis that the SW over-density can be the remnant of a disruptedsystem, unrelated to the PB and M 101, as proposed, for instance,by B20, is plausible and compatible with the smooth H I velocityfield map (Fig 10, right panel). In this second set of simulations, we shall consider the scenario thatwould attribute the off-set as the apparent position of the PB in thedisc to projection effects of an external system (an unbound galaxyor a bound satellite) crossing the line of sight (R04; Mihos et al.2012b; B20). By means of radial velocity measurements from emis-sion (absorption) lines of stars from the disc (PB), B20 constrainedthe maximum line-of-sight velocity difference between the two to ∼
50 km s − , less than the circular speed of NGC 5474. This fea-ture implies a similar distance and, together with the fact that discand PB have stellar populations with comparable age, probably im-plies a common history as well. As such, we explore cases wherethe PB is an external satellite galaxy, moving around NGC 5474onto motivated orbits, rather than an un-related foreground galaxy.We try to explore the possibility that some of the other peculiaritiesobserved in NGC 5474 (the H I warp, van der Hulst & Huchtmeier1979; the SW over-density, B20) may be caused by the gravita-tional interaction with such a satellite and, more generally, whetherthe observed properties of NGC 5474 are consistent with the pres-ence of an orbiting satellite.We focus on orbits that pass right above the galaxy’s sym-metry axis, starting from ( x i , y i , z i ) = (0 , ,
20 kpc) , with x, y, z the axes of a Cartesian reference frame whose ( x, y ) -plane is the MNRAS , 1–21 (2020) R. Pascale et al. − −
10 0 10 20 Y [kpc] − − − − Z [ k p c ] PBwithDM M0.5 d20 i = 21 ◦ − −
10 0 10 20 Y [kpc] − − − − Z [ k p c ] PBwithDM M1 d20 i = 21 ◦ Figure 11.
Left panel: trajectory around NGC 5474 (black curve) made by the PB of the hydrodynamical N -body model PBwithDM M0.5 d20. The PBcentre of mass starts from ( x i , y i , z i ) = (0 , ,
20 kpc) , with initial velocity v c ( z i ) = 33 km s − in the y -direction, in a Cartesian reference frame whose ( x, y ) -plane is the galaxy’s disc plane. The orbit is shown for (cid:39) . The system has been projected assuming as line of sight the x -axis, so Y ≡ y and Z ≡ z . The edges of the two orange bands show the different lines-of-sight ( i = 21 ◦ ) that would produce an off-set of on the plane of the discs,compatible with observations, when the PB centre crosses them. The stars of the PB are embedded with a dark-matter halo, whose isodensity contours aremarked in red. We show two different PBs, one corresponding to the ICs and one corresponding to a crossing of one of the lines of sight. As a comparison,we also show the projected distributions of the stellar disc of NGC 5474 (black points) and the NGC 5474 dark-matter halo isodensities (blue curves). Rightpanel: same as the left panel, but for the N -body model PBwithDM M1 d20. In the latter case, the orbit is shown for (cid:39) . galaxy’s plane of the discs. The only non-zero component of theinitial velocity of the PB centre of mass is in the y -direction, withmodulus v c ( z i ) = 33 km s − . Since the PB moves far from thediscs plane we relax the condition on its total mass and we embedit in a realistic dark-matter halo, following the scheme describedin Section 2.2. We focus on the two cases of a PB with a stellarmass of M PB = 0 . × M (cid:12) or M PB = 10 M (cid:12) . The PB haloparameters are chosen as in Section 2.2 and we highlight that the re-quirement in equation (15) implies that, when the PB stellar mass isdoubled, also the dark-matter mass is doubled, for fixed truncationradius. Since at z i = 20 kpc the effects of the dynamical frictionare still non-negligible, we expect the PB to sink slowly towardsthe galaxy centre in a few Gyr (equation 31).The PB ICs have been generated as described in Section 4.3,considering its separate luminous and dark components. We veri-fied that the PB is in equilibrium by evolving it for
10 Gyr in iso-lation. The only numerical effect is a slight decrease of the centraldensity, which is negligible for the purposes of our investigation.For clarity, we will refer to the two hydrodynamical N -bodymodels as PBwithDM M0.5 d20 and PBwithDM M1 d20, wherethe terms 0.5 and 1 indicate the PB stellar mass in units of M (cid:12) ,while d20 states that the PB centre of mass is located at an initialdistance of z = 20 kpc (see Table 5). As in the previous section,the ICs of NGC 5474 correspond to the equilibrium configurationof Appendix A after (cid:39) .
98 Gyr . Details on simulations and NGC 5474 parameters are listed in Tables 1 and A1. The simulations runfor and we use an adaptive timestep refinement with typicaltimestep values of . . Figure 11 shows the trajectories of the PBs of models PB-withDM M0.5 d20 and PBwithDM M1 d20, alongside their pro-jected spatial distribution taken at two representative snapshots.The systems have been projected assuming as line of sight thegalaxy’s x -axis, i.e. the axis perpendicular to the plane of the or-bit. The trajectories of the satellites differ the most in the num-ber of windings around NGC 5474. This different behavior iscompletely driven by dynamical friction: since the PB of modelPBwithDM M1 d20 has twice the dynamical mass of model PB-withDM M0.5 d20, its orbital time is approximately halved (equa-tion 31). While the lighter PB (left panel) completes approximatelyfour excursions above and below the equatorial plane in (cid:39) ,the stellar component of the most massive PB (right panel) devel-ops tidal tails after (cid:39) , when it has completed less than threewindings. In Fig. 11 we have marked the possible galaxy’s incli-nations ( i = 21 ◦ , as in R04) with two orange bands. The edgesof such bands are such that, when crossed by the PB centre, theyproduce an off-set of as an effect of projection on the stellardisc of NGC 5474. In both panels the orbits are interrupted when MNRAS000
98 Gyr . Details on simulations and NGC 5474 parameters are listed in Tables 1 and A1. The simulations runfor and we use an adaptive timestep refinement with typicaltimestep values of . . Figure 11 shows the trajectories of the PBs of models PB-withDM M0.5 d20 and PBwithDM M1 d20, alongside their pro-jected spatial distribution taken at two representative snapshots.The systems have been projected assuming as line of sight thegalaxy’s x -axis, i.e. the axis perpendicular to the plane of the or-bit. The trajectories of the satellites differ the most in the num-ber of windings around NGC 5474. This different behavior iscompletely driven by dynamical friction: since the PB of modelPBwithDM M1 d20 has twice the dynamical mass of model PB-withDM M0.5 d20, its orbital time is approximately halved (equa-tion 31). While the lighter PB (left panel) completes approximatelyfour excursions above and below the equatorial plane in (cid:39) ,the stellar component of the most massive PB (right panel) devel-ops tidal tails after (cid:39) , when it has completed less than threewindings. In Fig. 11 we have marked the possible galaxy’s incli-nations ( i = 21 ◦ , as in R04) with two orange bands. The edgesof such bands are such that, when crossed by the PB centre, theyproduce an off-set of as an effect of projection on the stellardisc of NGC 5474. In both panels the orbits are interrupted when MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 − . − . . . . η [kpc] − − − ξ [ k p c ] − . − . . . . η [kpc] − . − . . . . η [kpc] −
10 10 Y [kpc] − − Z [ k p c ] −
10 10 Y [kpc] − − Z [ k p c ] −
10 10 Y [kpc] − − Z [ k p c ] PBwithDM M0.5 d20
Figure 12.
Total (stellar disc and PB) stellar surface density map taken at three snapshots of the model PBwithDM M0.5 d20 assuming an inclination i = 21 ◦ .In each snapshot the PB produces, as a perspective effect, an off-set of 1 kpc on the discs’ plane. The centre of ( η, ξ ) -plane is the kinematic center of the stellardisc (blue dot), and the shades of greys extend for two order of magnitudes from the stellar density peak, corresponding, assuming a constant mass-to-light-ratio for both populations (stellar disc and PB) to a depth of five magnitudes. Indicating with Σ max the densest peak, the contours are separated by Σ max / n ,with n = 1 , ..., . The three snapshots have been taken after the system has evolved for t int = 3 .
43 Gyr , .
75 Gyr and .
28 Gyr , respectively from the leftto the right panel. In the small insets we show the full PB orbit as in Fig. 11, marking with a blue dot the position of PB along its orbit that corresponds to itsreference main panel, and with the dashed-orange curves the possible galaxy’s lines-of-sight.
Table 5.
Main input parameters of the set of simulations of Section 5.2.From top to bottom: name of the model (model’s name); PB total stel-lar mass ( M PB ); PB total dark-matter mass ( M dm , PB ); number of stel-lar particles used for the PB ( N PB ); number of dark-matter particles usedfor the PB ( N dm , PB ); softening used for the PB stellar component par-ticles ( l (cid:63), PB ); softening used for the PB dark-matter component particles( l dm , PB ). The softening is computed as in Section 5.1.1 and all the sim-ulations components (NGC 5474 dark halo, stellar disc, gas disc, PB stel-lar component and PB dark-matter halo) have different softenings. The PBstellar particles have m part = 1667 M (cid:12) , while the dark-matter particles m part = 5000 M (cid:12) . Details on how the PB dark-matter halo parametershave been fixed are in Section 2.2, while details on the PB initial positionand velocity are in Section 5.2.1. The ICs of NGC 5474 correspond to theconfiguration of Appendix A taken after .
98 Gyr (see also Tables 1 andA1).Models’ name PBwithDM M0.5 d20 PBwithDM M1 d20 M PB [ M (cid:12) ] 0.5 1 M dm , PB [ M (cid:12) ] 0.98 1.95 N PB N dm , PB l (cid:63), PB [ kpc ] 0.042 0.033 l dm , PB [ kpc ] 0.16 0.16 the systems have reached r ∼ − from the centre since, atshorter distances, the stellar components of the PBs develop non-equilibrium features (tidal tails, elongated structures) apparentlyinconsistent with observations, so we do not include them in anyfollowing analysis.The panels of Fig. 12 show the total stellar (PB anddisc) density map from three configurations from model PB-withDM M0.5 d20, projected assuming an inclination i = 21 ◦ .The panels, from the left to the right, are projections taken after t int = 3 .
43 Gyr , .
75 Gyr and .
28 Gyr , respectively after oneand a half, two and a half and three full excursions above and be- low the equatorial plane. The gradient of contours is as in Fig. 8 andthe small insets show the corresponding position of the PB along itsorbit. When the PB crosses the galaxy’s equatorial plane it causes,due to the gravitational perturbation, the development of a long-living and distinct spiral pattern in the stellar disc made up by twosymmetric arms that dig the stellar disc up to and extend outto − . It is worth noticing that the spiral pattern does notform when the NGC 5474 galaxy model is evolved in isolation (seeAppendix A, right panels of Fig. A3).Similar spiral arms form also in the H I disc. Although muchmore structured and extended (we recall that h gas /h (cid:63) (cid:39) ), thepattern of the H I follows the one of the stellar component. As anexample, Fig. 13 shows the H I projected density map correspond-ing to the middle panel of Fig. 12. As a reference, we have superim-posed with white contours the projected stellar density map of themiddle panel of Fig. 12. A spiral pattern forms also in model PB-withDM M1 d20. In both models the arms develop after the satel-lite has crossed the discs plane at least at ∼ . This happenssooner in model PBwithDM M1 d20, but it lasts for less due to thesmaller dynamical friction time. At this distance, at least for modelPBwithDM M0.5 d20, the crossings of the equatorial plane do notperturb sensitively the kinematics of the H I disc, whose rotationcurve (inset in Fig. 13) still looks very similar to the the initial oneand to the measured rotation curve of NGC 5474. When compar-ing with the observed morphology of NGC 5474, it is importantto bear in mind that our models do not include star formation. Forexample, in the real galaxy, star forming and H II regions trace a dif-ferent spiral pattern with respect to old-intermediate age stars. Oursimulations can approximately trace the latter but not the former.Figure 14 shows the PB stellar projected density profile ob-tained from the middle panel of Fig. 12. The profile has been com-puted as in the previous section by binning with 50 spherical annuli,equally spaced in R , out to .
13 kpc (corresponding to . ,as in B20). The background has been evaluated from a wider anddistant annulus, and subtracted to the main profile. For compari-son, the yellow curve shows the PB S´ersic model as in the analytic MNRAS , 1–21 (2020) R. Pascale et al. − . − . . . . η [kpc] − − − ξ [ k p c ] gas distribution R [kpc]304050 v [ k m s − ] Figure 13.
Same as the middle panel of Fig. 12, but showing the H I discsurface density map. The density decreases from yellow to blue. The whitecontours show the stellar spatial distribution of the configuration of the mid-dle panel of Fig. 12, where, for clarity, we have removed the lowest iso-density contour. The blue dot shows the kinematic centre of the H I disc.Small inset: H I circular speed as a function of the galactocentric distance(blue points with error bars) from the same snapshot as in the main panel,compared to the galaxy’s deprojected rotation velocity computed in Sec-tion 2.1.1 (black points with error bars). The orange curve shows the circu-lar speed of the analytic model of NGC 5474. R [kpc] Σ [ M (cid:12) k p c − ] . r c s ec = . k p c Real PBsoftening N -body model Figure 14.
Projected density distribution of the PB as a function of thedistance R from the PB centre (blue circles with errorbars) superimposedto the analytic S´ersic model (yellow dashed line) from which the ICs havebeen sampled (i.e. with the same m and R e as the observed PB, B20, butwith a total initial mass M PB = 0 . × M (cid:12) ). The PB corresponds tothe configuration taken as in the middle panel of Fig. 12 and Fig. 13.. model from which the ICs have been sampled (corresponding to theS´ersic model of B20, with a total stellar mass M PB = 10 M (cid:12) ).The two profiles differ the most in the central parts, even though,as mentioned and discussed in the previous Section, we find thevery same difference also when the system is evolved in isolation.As a second, interesting feature, we find that the satellite − Y [kpc] − Z [ k p c ] t int = 1 .
86 Gyr − ξ [kpc] − − − η [ k p c ] - - - - R [kpc] N H I [ c m − ] ICs N -body modelObservations-5 0 5 Y [kpc] -808 Z [ k p c ] Figure 15.
Top panel: H I projected density distribution when thegalaxy is viewed edge-on. The configuration is taken from model PB-withDM M1 d20 after t int = 1 .
86 Gyr from the beginning of the sim-ulation. The shades of colours, from yellow to black show regions of de-creasing density, the dashed curve shows the ( X, Y ) -plane while the solidcurve is ◦ inclined, as the warped H I disc. Middle panel: H I velocity fieldmap from the same snapshot as in the top panel. The system is projectedwith an inclination of i = 21 ◦ and the spatial and velocity resolutions areas in Fig.s 8 and 10. The small inset shows with a blue dot the correspond-ing position of the PB along its orbit (black circle) and with a red dot thegalaxy’s centre. The two orange lines mark the galaxy’s possible inclina-tions. Bottom panel: projected HI column density profile computed fromthe same snapshot as in the top and middle panels (blue dot with errorbars)superimposed to the observed profile (black squares with errorbars), as de-rived in Section 3, and to the analytic models from which the ICs have beensampled (dashed red curve). mildly warps the H I disc when it crosses the galaxy’s equatorialplane. The top panel of Fig. 15 shows the projected density dis-tribution of the H I disc from model PBwithDM M1 d20, whenthe galaxy is viewed edge-on, after the system has evolved for t int = 1 .
86 Gyr and the PB has completed a full oscillation inthe vertical direction, crossing the equatorial plane twice. We se-lected a snapshot in which the PB is also off-centred fromthe discs’ centre, similarly to Fig. 12 (see the small inset in themiddle panel of Fig. 15). The H I bends by approximately ◦ andthe warp lives for another complete full vertical oscillation of thePB around NGC 5474. However, the H I velocity field strongly con- MNRAS000
86 Gyr and the PB has completed a full oscillation inthe vertical direction, crossing the equatorial plane twice. We se-lected a snapshot in which the PB is also off-centred fromthe discs’ centre, similarly to Fig. 12 (see the small inset in themiddle panel of Fig. 15). The H I bends by approximately ◦ andthe warp lives for another complete full vertical oscillation of thePB around NGC 5474. However, the H I velocity field strongly con- MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 strains the minimum distance that a system with such dynamicalmass can reach. The middle panel shows the H I velocity field mapfrom the same configuration of the top panel, but when the galaxyis seen inclined by i = 21 ◦ (the resolution of the map is as inFig.s 8 and 10). The iso-velocities contours are regular enough tobe consistent with those of R04, even though the approaching andreceding arms are not symmetric over the full map: the approach-ing arm reaches a −
24 km s − amplitude at − , while itpeaks at −
15 km s − at the same distance in the opposite direc-tion. The maximum velocity difference between the approachingand receding arms reported by R04 is only − − . The con-figuration is the last still compatible and in good agreement withthe observations: for longer times, further interactions between thetwo systems erase any sign of regularity and differential rotationfrom the H I velocity field. The PB of model PBwithDM M0.5 d20does not warp or distort the H I disc when it crosses through it,keeping the H I velocity field regular, thanks to its lower dynamicalmass and to the fact that most of the crossings happen for distanceslarger than . We recall that in model PBwithDM M1 d20, thePB dynamical mass is considerably high, more or less comparableto the one of NGC 5474. So, it sounds plausible that a system witha lower mass (in between the two models) could, at the same time,warp the H I without significantly distorting the H I velocity fieldmap.The bottom panel of Fig. 15 shows the H I column density dis-tribution corresponding to the middle panel, compared with the H I column density we derived in Section 3 from observations. Thedensity profile of the H I changes and bends at ∼ − , whichcorresponds approximately to the distance of the latest crossing ofthe PB, but the overall shape is consistent with the observed one,apart from the centre, where the two profiles differed the most al-ready at the beginning of the simulation (see Appendix A, Fig. A2). As a member of the M 101 Group, and appearing in projection soclose to its giant central galaxy, a tumultuous past has always beeninvoked as the main driver of all of the peculiarities of NGC 5474.However, while the hypotheses of a gravitational interaction withM 101 may explain, for instance, the warped H I distribution (R04),it does not look like an explanation for its off-centred bulge. Off-set bars are observed in Magellanic spirals (Odewahn 1989), eventhough the mechanism that can induce the misplacement is still un-known. NGC 5474 is, however, not a Magellanic spiral: i) it doesnot possess a bar, but rather a very round and regular stellar compo-nent; ii) it has two spiral arms and not one; iii) off-centred bars inMagellanic spirals are observed mostly in binary systems. To com-plicate things, the only available H I observations of NGC 5474 dateback to the early 90s, and only trace the large scale structure of thegalaxy, while the more recent H α observations, tracing the innerkinematics, seem to be hardly reconcilable with the H I data (E08;B20), though the galaxy’s low inclination does not allow to drawrobust conclusions on the disc’s kinematics.Following the work of B20, who renewed the interest in thisgalaxy accomplishing a detailed study of its stellar populations, wehave produced state-of-the-art hydrodynamical N -body models ofNGC 5474, aimed to investigate the nature of the galaxy’s centraland compact stellar component, usually interpreted as an off-setbulge. Using analytic models we have argued that, if the PB reallylies within the galaxy’s disc plane, it is implausible that its dynam-ical mass is more than M (cid:12) , because such a system would: i) shift the entire galaxy’s gravitational potential minimum, makingthe kinematic centre of the discs coincide with the centre of thebulge; ii) induce strong distortions in the H I velocity field map,inconsistent with observations.Through hydrodynamical N -body simulations, we tried to re-produce configurations where the PB appears off-centred as a resultof orbital decay due to dynamical friction, when it moves withinthe galaxy’s discs plane. We explored PBs of different massesand sizes but, in none of the considered scenarios we were ableto reproduce the observations: a system with a large initial size( R e ≥
320 kpc ), while evolving into the strong tidal force field ofNGC 5474, develops massive tidal tails and gets flattened and elon-gated (in some cases destroyed) after less than . . The veryshort time needed to show non-equilibrium features makes very un-likely that a PB with these characteristics can either come from thegalaxy’s outer regions or be an off-centred pseudo-bulge. A com-pact system reaches the required distance from the centre but,on the basis of structural analysis, it remains either too compact orgets too flattened to look similar to the observed one.Through N -body simulations, Levine & Sparke (1998)showed that an off-set between a stellar disc and the gravitationalpotential minimum of its host galaxy can stand for sufficiently longtime if the stellar disc spins in a sense retrograde to its orbit aboutthe halo centre. According to the authors, we should then interpretthe discs as off-centred with respect to the bulge, and not the otherway around. We believe that this is not the case of NGC 5474 inwhich the off-set stands in the gas kinematics as well (Levine &Sparke 1998 considered collisionless simulations with no gas). Ifwe imagine the gas to behave similarly to the stellar counterpart,according to Levine & Sparke (1998), the H I velocity field shouldbe clearly and strongly asymmetric, which is not the case for NGC5474.As different authors proposed (R04; Mihos et al. 2012b, B20),we have explored the hypothesis that the off-set is produced by pro-jection effects, once the PB is orbiting around NGC 5474. Due tothe structural homology between the PB and a dE galaxy, we havecoated it with dark-matter halo. We have shown reference cases ofpolar orbits where, in projection, the PB looks off-centred of ,just as observed. We exploit the gravitational interaction betweenthe satellite PB and NGC 5474 to show that it may: i) explain theformation of the galaxy loose spiral pattern, formed by two sym-metric arms, together with a very similar structure in the H I dis-tribution; ii) partially account for the formation of the warped H I disc, at least in cases of sufficiently massive PB.Of course, the large parameter space, the lack of tight observa-tional constraints and the degeneracy induced by projection wouldallow hundreds of orbits to reproduce the observed, present-dayconfiguration. As such, we do not expect to have solved all the mys-teries behind NGC 5474, but rather to have shown in a quantitativemanner that its PB is probably not the bulge or the pseudo-bulgeof NGC 5474, and we have also presented a possible, intriguing al-ternative scenario where the PB is a satellite galaxy of NGC 5474,moving on a polar orbit, that has the advantage of explaining someof the other peculiarities of NGC 5474.While our study is not sufficient to ascertain the real nature ofthe PB, it provides for the first time a sound way out to the mainpuzzle of the structure of NGC 5474: the odd off-centred ‘bulge ’ islikely not a bulge at all, but a satellite dwarf galaxy projected nearthe center of a M33-like bulge-less spiral (B¨oker et al. 2002; Daset al. 2012; Grossi et al. 2018). MNRAS , 1–21 (2020) R. Pascale et al.
ACKNOWLEDGMENTS
We thank the anonymous referee for his/her comments and sug-gestions that considerably improved the quality of this work. Weacknowledge the use of computational resources from the parallelcomputing cluster of the Open Physics Hub ( https://site.unibo.it/openphysicshub/en ) at the Physics and Astron-omy Department in Bologna. We acknowledge funding from theINAF Main Stream program SSH 1.05.01.86.28. FM is supportedby the Program ‘Rita Levi Montalcini’ of the Italian MIUR. Wethank F. Fraternali and G. Iorio for very helpful discussions. RPacknowledges G. Sabatini for useful suggestions and comments.
DATA AVAILABILITY
The rotation curves and the H I column density map of NGC5474 are available at https://dx.doi.org/10.1086/117185. The rota-tion curve and the H I density distribution rederived in this articlewill be shared on request to the corresponding author. REFERENCES
Barnes J., Hut P., 1986, Nature, 324, 446Bellazzini M., Annibali F., Tosi M., Mucciarelli A., Cignoni M., BeccariG., Nipoti C., Pascale R., 2020, Astronomy & Astrophysics, 634, A124Binney J., Tremaine S., 2008, Galactic Dynamics: Second Edition. Prince-ton University PressB¨oker T., Laine S., van der Marel R. P., Sarzi M., Rix H.-W., Ho L. C.,Shields J. C., 2002, AJ, 123, 1389Calzetti D., et al., 2015, AJ, 149, 51Ciotti L., 1999, ApJ, 520, 574Costa T., Pakmor R., Springel V., 2020, arXiv e-prints, p. arXiv:2006.05997Das M., Ramya S., Sengupta C., Mishra K., 2012, in Astronomical Societyof India Conference Series. p. 153Dehnen W., Read J. I., 2011, European Physical Journal Plus, 126, 55Epinat B., et al., 2008, MNRAS, 388, 500Fisher D. B., Drory N., 2008, AJ, 136, 773Fisher D. B., Drory N., 2010, The Astrophysical Journal, 716, 942–969Forbes D. A., Sinpetru L., Savorgnan G., Romanowsky A. J., Usher C.,Brodie J., 2017, MNRAS, 464, 4611Grand R. J. J., et al., 2017, MNRAS, 467, 179Grossi M., Fernandes C. A. C., Sobral D., Afonso J., Telles E., BizzocchiL., Paulino-Afonso A., Matute I., 2018, MNRAS, 475, 735Hastings W. K., 1970, j-BIOMETRIKA, 57, 97Hernquist L., 1990, ApJ, 356, 359Huchtmeier W. K., Witzel A., 1979, A&A, 74, 138Iorio G., Nipoti C., Battaglia G., Sollima A., 2019, MNRAS, p. 1366Kormendy J., Kennicutt Robert C. J., 2004, ARA&A, 42, 603Kornreich D. A., Haynes M. P., Lovelace R. V. E., 1998, AJ, 116, 2154Kornreich D. A., Haynes M. P., Lovelace R. V. E., van Zee L., 2000, TheAstronomical Journal, 120, 139–164Kregel M., van der Kruit P. C., de Grijs R., 2002, MNRAS, 334, 646Lau E. T., Nagai D., Kravtsov A. V., Vikhlinin A., Zentner A. R., 2012, ApJ,755, 116Levine S. E., Sparke L. S., 1998, ApJ, 496, L13Marinacci F., et al., 2018, MNRAS, 480, 5113McConnachie A. W., 2012, AJ, 144, 4Metropolis A. W., Rosenbluth M. N., Teller A. H., Teller E., 1953, Journalof Chemical Physics, 21, 1087Mihos J. C., Keating K. M., Holley-Bockelmann K., Pisano D. J., KassimN. E., 2012a, ApJ, 761, 186Mihos J. C., Harding P., Spengler C. E., Rudick C. S., Feldmeier J. J.,2012b, The Astrophysical Journal, 762, 82Mu˜noz-Cuartas J. C., Macci`o A. V., Gottl¨ober S., Dutton A. A., 2011, MN-RAS, 411, 584 parameter DM halo stellar disc H I disc( M dm , M (cid:63) , M gas ) = (3 . , . , . × M (cid:12) ( N halo , N (cid:63) , N gas ) = (708750 , , ( l dm , l (cid:63) , l gas ) = (0 . , . , . Table A1.
Main input parameters of the hydrodynamical N -body simula-tion of NGC 5474 in isolation. M dm , M (cid:63) , M gas : respectively total dark-matter, stellar and H I masses. N halo , N (cid:63) , N gas : number of particles usedfor the dark-matter halo, the stellar disc and the H I disc. l dm , l (cid:63) , l gas : re-spectively the softening used in all the simulations for the halo, the stellarand the H I discs. The softenings are computed requiring that the maximumforce between the component’s particles should not be larger than the com-ponent’s mean-field strength (Power et al. 2003; Dehnen & Read 2011). Allparticles have the same mass m part = 5000 M (cid:12) .Naiman J. P., et al., 2018, MNRAS, 477, 1206Navarro J. F., Frenk C. S., White S. D. M., 1996, ApJ, 462, 563Nelson D., et al., 2018, MNRAS, 475, 624Odewahn S. C., 1989, PhD thesis, Texas Univ., Austin.Oh S.-H., de Blok W. J. G., Walter F., Brinks E., Kennicutt Robert C. J.,2008, AJ, 136, 2761Oh S.-H., et al., 2015, AJ, 149, 180Pillepich A., et al., 2018, MNRAS, 475, 648Power C., Navarro J. F., Jenkins A., Frenk C. S., White S. D. M., SpringelV., Stadel J., Quinn T., 2003, MNRAS, 338, 14Read J. I., Iorio G., Agertz O., Fraternali F., 2017, MNRAS, 467, 2019R¨ock B., Vazdekis A., Peletier R. F., Knapen J. H., Falc´on-Barroso J., 2015,MNRAS, 449, 2853Rownd B. K., Dickey J. M., Helou G., 1994, AJ, 108, 1638Skibba R. A., et al., 2011, ApJ, 738, 89Smith R. J., Glover S. C. O., Clark P. C., Klessen R. S., Springel V., 2014,Monthly Notices of the Royal Astronomical Society, 441, 1628Springel V., 2010, MNRAS, 401, 791Springel V., Di Matteo T., Hernquist L., 2005, Monthly Notices of the RoyalAstronomical Society, 361, 776–794Springel V., et al., 2018, MNRAS, 475, 676S´ersic J. L., 1968, Atlas de galaxias australesTikhonov N. A., Lebedev V. S., Galazutdinova O. A., 2015, AstronomyLetters, 41, 239Tully R. B., et al., 2013, AJ, 146, 86Weinberger R., Springel V., Pakmor R., 2020, ApJS, 248, 32van de Voort F., Pakmor R., Grand R. J. J., Springel V., G´omez F. A.,Marinacci F., 2020, MNRAS, 494, 4867van der Hulst J. M., Huchtmeier W. K., 1979, A&A, 78, 82 APPENDIX A: THE INITIAL CONDITIONS OF NGC 5474
The ICs of the NGC 5474-like galaxy model, as sampled in Sec-tion 4.2.3, are evolved in isolation for t max = 10 Gyr to checkhow the components respond to the presence of each other, and tolet the system, which is built in quasi-equilibrium, shift towardsequilibrium. We use an adaptive timestep refinement, with typicaltimesteps of . . Table A1 lists some of the main input param-eters used to generate the ICs of the galaxy model of NGC 5474 andrun the simulation (masses, number of particles and softenings).In Fig. A1 we show the dark-matter density (top panels) andvelocity dispersion tensor elements σ r , σ θ and σ φ (bottom panels)and compare them with the corresponding quantities from the dark-matter analytic model. The profiles are computed after the galaxyhas evolved for t int = 0 .
98 Gyr (left column) and t int = 4 .
89 Gyr (right column). The halo density increases by approximately afactor of two while the velocity dispersion components σ i (with MNRAS000
89 Gyr (right column). The halo density increases by approximately afactor of two while the velocity dispersion components σ i (with MNRAS000 , 1–21 (2020) ydrodynamical N -body models of NGC 5474 ρ [ M (cid:12) k p c − ] t int = 0 .
98 Gyr modelsofteningicssnapshot t int = 4 .
89 Gyr r [kpc] σ i [ k m s − ] r [kpc] σ r σ θ σ φ Figure A1.
Top panels: density distribution of the halo (black points witherrorbars) from two reference snapshots of the NGC 5474 simulation ofAppendix A, corresponding to t int = 0 .
96 Gyr (left column of panels).and t int = 4 .
89 Gyr (right column of panels). Bottom panels: velocitydispersion tensor elements σ i , with i = r (blue points with errobars), θ (grey points with errobars), and φ (green points with errobars). The left-hand panels also show the corresponding quantities as computed from theICs (orange points with error bars), while the analytic density distribution(top panels) and the velocity dispersion of the isotropic analytic model (bot-tom panels) from which the halo ICs have been sampled is shown witha dashed-red curve. The radial bins in the velocity dispersion profiles havebeen computed so that they contain the same number of particles N halo / .The light blue bands in all panels extend out to l dm , with l dm the adoptedsoftening (see Table A1). The yellow and blue dashed vertical lines show,respectively, the region marked by the stellar ( h (cid:63) ) and H I ( h gas ) discs scalelengths. i = r, θ, φ ) increase by − − , in a region confined mostlywithin h (cid:63) , where the discs contribute the most. The increase in thevelocity dispersions is of a factor 1.3, and the overall profiles areapproximately the same one would get with an isotropic Hernquistmodel with M tot = M dm + M (cid:63) + M gas . For a spherical system,the parameter β = 1 − σ θ + σ φ σ r = 1 − σ θ σ r (A1)measures the models’ anisotropy distribution . An isotropic modelcorresponds to β = 0 , while β < and < β ≤ indicate,respectively, tangential and radial velocity distributions. We alsonote that the halo changes its velocity distribution from isotropic( β = 0 ) to radially biased ( β > ) due to the weak collapse inducedby the deeper potential well.In Fig. A2 we show some of the main structural and kinematicproperties of the stellar and H I discs corresponding to the sameconfigurations as in Fig. A1. The top rows show the stellar pro-jected density distribution, the middle row the H I projected densitydistribution, while the H I streaming velocity profile is shown inthe bottom panels. As for the halo, the discs adjust in the very be-ginning of the simulation: they respond to the contraction of thehalo increasing their central densities, in a region extending out We recall that spherical symmetry implies σ θ = σ φ . Σ ? [ M (cid:12) k p c − ] t int = 0 .
98 Gyr t int = 4 .
89 Gyr Σ ga s [ M (cid:12) k p c − ] . . . . R [kpc] v φ [ k m s − ] . . . . R [kpc] Stellar disc modelICsSnapshot
Figure A2.
Main structural and kinematic properties of the stellar and H I discs from two reference snapshots of the NGC 5474 simulation of Ap-pendix A, corresponding to t int = 0 .
98 Gyr (left column of panels) and t int = 4 .
89 Gyr (right column of panels). Top panels: stellar disc surfacedensity profile (black points with error bars); middle panels: H I projecteddensity distribution (black points with error bars); bottom panels: H I az-imuthal velocity curve (black points with error bars). The system has beenprojected assuming the symmetry axis as line of sight. In the left columnwe also show the corresponding quantities as derived from the ICs (orangepoints with error bars). The dashed red and blue curves show, respectively,the stellar and H I surface density as from the analytic model from which theICs have been sampled, while the dashed green curve in the bottom panelsshow the analytic model circular speed. R (cid:39) h (cid:63) , until a new equilibrium configuration is reached. As a re-sult of the deeper potential well caused by the collapse of the haloin the galaxy’s central parts, mostly the stellar disc develops radialdensity waves that are still in place after , although in a lowdensity region. Figure A3 shows the discs density maps, face-onand edge-on, computed from the same configurations as in Figs A1and A2. Both discs are generated with non negligible thickness.We recall that the stellar disc vertical profile follows from equa-tion (22), with z (cid:63) = 0 . h (cid:63) , while the thickness of the gaseousdisc is determined by means of the vertical hydrostatic equilibrium(23). The radial density wave is clearly visible as the outer ring inthe stellar surface density map of the left panel of Fig. A3. Also,we notice that the stellar disc does not develop any spiral structurealong the whole simulation. MNRAS , 1–21 (2020) R. Pascale et al. − Z [ k p c ] stellar disc HI disc − Y [kpc] − . − . . . . X [ k p c ] − Y [kpc] t int = 0 .
98 Gyr − Z [ k p c ] stellar disc HI disc − Y [kpc] − . − . . . . X [ k p c ] − Y [kpc] t int = 4 .
89 Gyr
Figure A3.
Stellar and H I discs main structural and kinematic properties. The two left-hand columns of panels show the face-on (bottom panels) and edge-on(top panels) stellar disc and H I surface density maps, projected along the galaxy’s symmetry axis. The configuration is taken after the system has evolved for t int = 0 .
98 Gyr . The two right-hand columns of panels show the same as in the left panels but after the system has evolved for t int = 4 .
89 Gyr . The yellowand blue circles show, respectively, the region marked by the stellar and H I discs scale lengths. The snapshot are taken from the simulation of NGC 5474 ofAppendix A. MNRAS000