Dynamics of Galaxies GIUSEPPE BERTIN Scuola Normale Superiore, Pisa CAMBRIDGE UNIVERSITY PRESSContents Preface Acknowledgments PART ONE: BASIC PHENOMENOLOGY 1 SCALES References ad further wading Problems 2 OBSERVATIONAL WINDOWS: 21 Radio 22 Millimerie 23 Infaced 24 Optical 25 Uhrsviolet 26 Xeavs 27 Other sources References and further rang Problems 3. CLASSIFICATIONS 21 Hubble classification 32. Momiology of elliptical galaxies 83. Momiology of spiraf galaxies References and farther reading Problems 4 PHOTOMETRY, KINEMATICS, AND DARK MATTER 41 Luminosity profiles 42. Doppler Tine shifts and Tinewidths 43° Global scaling laws 44 Dark riaiter and cosmology References and further reading Problems 5 BASIC QUESTIONS, SEMIEMPTRICAL APPROACH, AND THE DYNAMICAL WINDOW 1 Suucture 32. Formation and evolution page xiii xv wow w i 4 4 16 6 a 18 2 B 25 28 % 36 41 45 4” 34 7 soviii Contents 53. Modeling and some fundamental questions 54. Relation to ether branches of astrophysics PART TWO: PHYSICAL MODELS 6 0 i SELF-GRAVITY AND RELATION TO PLASMA PHYSICS 6.1 Gravity and aoll-geavity 62. Collective behavior 63. A concise dictiousy References and farther reading. Problems RELAXATION TIMES, ABSENCE OF THERMODYNAMICAL EQUILIBRIUM 74. Two-siar relaxation times 4.2 Colisioeality in inbomogensous systems 7.3. Collisionless relaxation processes References and further reading Pratleras MODELS 81 Systems of many particles 2 Continuum: limit and stellar dynamics 63 FluicJimit and Fuid models 4 Virial equations AS Asymptoties versus toy models References and farther reading. Problems EQUILIBRIUM AND STABILITY: SYMMETRY ANI SYMMETRY BREAKING 9.) The Jeans theorem 92. Syrnmetry beeaking 9.3. Motivation foe perturbation analyses 9.4. Spiral galaxies versus elliptical galaxies Refeceness and further reading Problems CLASSICAL ELLIPSOIDS: 10.1 Ellipsoidel figures of eguilibrivan 102 Rigidly rotating equilbciums elipsoide 103 The mechanism of bifarsavion References and fuser reading Prodlems INTRODUCTION TO DiSPERSIVE WAVES 11.4 Hyperbolic waves 112. Dispersive waves 13 Water waves References sand futher reveling. Problems 60 a gasea dh ” 5 n 80 82 geceseae 192 106 107 108 109 no 15 i 11 123 124 135 127 131 13s 136Contents 12 JEANS INSTABILITY 121 Homogeneous fi model 122 Homogeneous kivetic model and Landim damping 128 Kinetie analysts of a plane-parallel ab References acid forther raging Problems PART THREE: SPIRAL GALAXIES 13 14 15 16 ORBITS 19.4 Action and angle vatiables 122. Epicyelic orbits 123 Rotating frame 134 Trapping atthe Lagyangien pointy 1533 Equations forthe guiding centers 13.6 Exact orbits and the problem of solf-consistency References and further reading Pratlems THE BASIC STATE: VERTICAL AND HORIZONTAL EQUILIBRIUM OF THE DISK 14 Vertical equibibrium 152, Quasi-MaxweTlian distribotioa functions for a cool disk 143 Bxact models 144A reference bosie state for a one-component model of the disk References acd further reading Preiblems DENSITY WAVES: 154 Dispersion relaions 152° Marginal stability 153. Wave branches is Turning points and resonances 18.5 Dywatuical mechanisins 156 The equations of the homogencows shearing sheet Roferences and further reading Problems ROLES OF GAS 161 Waves and effective stability in 2 two-compament disk, 162. The modeling process fr « dik of stars and gas 163. Seif-regulation fora disk of stars ard gas 164 Difleremt behaviors of gas and stars at resonances 165 Role of near-infrared observations References und furiber reading, Problems GLOBAL SPIRAL MODES V7 Bact equations for Hear density permacbations im w fluid disk model (12 Reduction to an ordinary differential equation for tightly wound pesterations 173. Atwo-curning point problem 137 137 11 143 4s 148 147 149 as 183 1s 156 180 164 167 168 170 an ns "7 19 153 185 186 139 194 197 201 202 207 2a 214 215 220, ma mm 2 231 233 236 27 240X Contents 174 Quantuun condition and discrete spectrum of spirel mades 243 175. From Hineae mndes to Spica xruciores in galonies 2s References and further reading 10 Problem 251 18 SPIRAL STRUCTURE IN GALAXIES 253 18.1 Quasi-sttlonary spiral structure and thove levele of persistence 284 182 Dynamical classification of spiral moxphalogies 256 83. Ioterpreting specific observed features 261 1&4 Evolution 262 References and farther seating, 268 Problem: 264 19 BENDING WAVES 266 19,1 Bending waves in simple slab model 269 192 Disk-hale interaction and celated two-stream institity 2 19.3 Inhamngeneces disks and the zere-thickness lit 72 124 Bending waves on o current shes and the solar acctors 24 Rotetences and further zeading 2 Problems 280 20 DARK MATTER IN SPIRAL GALAXIES 281 20... Dark matter in our Galeny 282 202 Usternal galaxies oe 203 Maximum disk 287 24 Dynamical argamens 288 205 Modified Nowromian dynamics? 288 References and further seading 2a1 Problems 293 PART FOUR: ELLIPTICAL GALAXIES 298 21 ORBITS 297 21.1 Spherical potentials 209 212 Classification of potentials in relation to the Feans theorem 302 212 Nonspherics! potentials with isolating integrals 304 214 More general potentials 309 References amd further seeding a0 Problems 3 22 STELLAR DYNAMICAL MODELS 33 22.1 Four approaches 318 222. Spherical isouopic models 319 22.3 Anisoiropic foo models 331 [References and further sending, 345 Problems 39 23 STABILITY 352 23.1 Basic equations for linear moves 353 232 The radial orbit instabitiry 359Conients xi References and farther reading 363 Problem 365 24 DARK MATTER IN ELLIPTICAL GALAXIES 366 241 Kinematical tracers 305 242 Stellar dynamical diagrosties 369 243 Dark mtr io groups and clusters 3m Refereuces and further reading 316 Problem 7% PART FIVE: IN PERSPECTIVE 379 25 SELECTED ASPECTS OF FORMATION AND EVOLUTION 381 28.1 Collisionless collapse 382, 252, Dissipative collapse and accretion processes 387 253, Global ebaracterstics of galaxies and their evolution 395 References and further reading 309 Bibliography 403 Index 40712 Jeans instability In this chapter we derive the basic equations for the Jeans instability! in the siruplest formala- tion. This is meant to be the firstexample in the astrophysical context of the concepts introduced in Chap. 11. In reality, the study of stability brings us one step beyond the concept of dispersive waves. Most of what has been described in Chapter 11 is applicable in cases in which the fre- quency in the relevant dispersion rélation is allowed to be complex, w = tog -+ iar, provided that |a2;| < Jag]. But we should be aware that in the opposite limit we are no longer talking ahout waves in the usual sense. In the simplest case, gravity and pressure forces are the main ingredients in the Jeans instability mechanism. Below, quite different resutts are derived from the sume ingredients, depending on the mode] considered, In particular, although the fluid model is characterized by a dispersion relation for which neutral waves (cor =0) are allowed, a kinetic analysis leads to a dispersion relation in which stable perturbations are generally damped. At the end of the chapter, we briefly examine the linearized equations for perturbations of a plane-parallel slab, demonstrating how, in general, the collisionless Boltzmann equation is solved by integration along the unperturbed characteristics, The analysis also points out some analogies to the plasma context and thus justifies some of the points made in Chap. 6, We do not deal here with the variety of applications that the Jeans instability has in astraphysies, as this is widely availabie in the literature, The more specific applications to the large-scale dynamics of galuxy disks are postponed until Part Three. 12.1 Homogeneous fluid model We consider a formal equilibrium configuration of a homogeneous self-gravitating fluid ini- tially atrest. This basic state is taken as the limit of more realistic inhomogeneous basic states for which all the seale lengths characterizing the density and pressure distributions are much longer than the scales of the perturbations that are consideresl. This qualification should thus be taken into account at the interpretation level of the derived dispersion relation. Let uy. pi. ®1, and pi represent perturbations of the fluid velocity, density, gravitational potential, and pressure, respectively. Then the linearized equations for momentum and mass conservation, the equation of state, and the Poisson equation become au oe =-Vpi — mV, (2.138 Jeans instability a + mVu =0, 412.2) P=GpL 12.3) VO, = 42Gp, (12.4) Here po and ¢? are constants that characterize the homogeneous basic state. Taking the diver- ‘gence of Eq. (12.1) and eliminating m1, ®,, and py by means of Eqs. (12.2)-(12.4), we are left with a single partial differential equation for o, that resembles Eq, (11.12): ba EV ar — wip = 0, (12.5) with oF = 4 Goo. (12.6) ‘The corresponding dispersion relation is thus cok? — apo. 12.7) Tn the limit G -> 0 we recover sound waves, whereas in the cold case c? + 0 we find the free-fall kinematic limit, In general, the frequency of the perinebation is determined by a balance between the stabilizing pressure contribution and the destabilizing self-gravity term For sufficiently long waves, |k| < iy, with (12.8) 12.1.1 Relation to the virial theorem ‘The closest realistic basic state for which we might wy to apply the above concepts could be a small region of a nonrotating spherical system of finite mass M, total kinetic energy K, and gravitational energy W. In this case, we may check whet would be the relevant parameter regime for the Jeans dispersion relation, given the fact that the virial theorem for such a system imposes a condition between K and W. Aa average thermal speed c, may be defined from the relation 2K = Me?, and a scale length 8 for the size of the system may be introduced by means of the relation W = —GM?/R; a typical density would thus be 2 = 3M/(4z R°), ‘Therefore the virial constraint would requize that 2 GM Ng ayn 129) = Ee carcpr. (129) so that Rave. 2.10) 3 312.1 Homogeneous fluid modet 139 ‘The quantity @, is constructed as in Bq, (12.6) in terzns of p. We have thus proved that the sysiem is not large enough to accommodate Jeans unstable perturhations. Ta other words, it appears that a finite spherical, nonrotating, self-gravitating system has its size determined in such a way that the Jeans instability is suppressed. This is, of course, an indication of a dimensional analysis argument only. it suggests that the actual stability of a finite-mass spherical system should be checked by a different analysis for witich the inhomogeneity of the system is fully recognized. Some physical ingredients, not present in the simplest Jeans stability analysis given above, may open the way to instability (see Chap. 23). 12.1.2 Electrostatic plasma waves In the plasma context, an analysis similar to that leading to Bq. (12.5) can be carried out for electrostatic waves in an ionized gas.° Here a homogeneous basic state is more natural, because we can start from a neutrality condition ensured by the presence of ions and electrons. In the simplest derivation we imagine that the ions are fixed and only electrons are mobile, (A kinetic analysis, in line with what will be given below in Section 12.2, is required for defining the limitations and the merits of this simple Quid model? in particular, the fluid model used here would be justifled only for sufficiently high frequency waves.) The different behavior of electrostatic forces with respect to charges of the same sign produces a difference in the Poisson equation, which now becomes Wo, =—-damq, (2p where 1, is the number density perturbation for the electron component and ¢ = —e is the electron charge. The final dispersion relation for electrostatic plasma waves* is ee ea (2.12) with 2 we Anne (12.13) Me Asa result, plasma waves are stable because Coulomb forces help prevent the development of charge clumps. The format anatog to the Jeans wave number is 2.14) ‘The associated length scale, called the Dehye length (expressed in terms of temperature T and particle number density n, the standard definition is 4 = JET 74zne?, where & is here the Boltzmann constant), sets the size beyond which low-frequency electrostatic forces are screened collectively by the plasma. Ifa point charge is imposed, charges rearrange themselves inside a Debye sphere to neutralize the field outside the sphere.”140. Jeans instability 12.13 Simple generalizations We casily generalize the Jeans stability analysis to other simple basic configurations,* for exaruple to the case in which the system is inidially in solid-body rotation, with angular velocity 2. In the rotating frame of reference, Eq. (12.1) is replaced with am, pore = Wr — wo 1 + 2p x D (2.15) with the appearance of a Coriolis correction. If we introduce the angle ¢ defined by the two vectors k and 9, the relevant dispersion relation is found to be (of + of) o* +oFag = 0, 12.16) with at + oy = 4? + CP? — Ae Goo, 2.17) wpe} = 407 (c2k? — 4 Gpn) cos? 9. (2.18) Theretiore, for cos# = 0, the Jeans criterion for stability, k| > ky, holds unmodified, whereas for k perpendicular to the rotation axis the dispersion relation is oF = 40 + 02k" — Aa Gpo. 2.19) In the presence of differential rotation (see Chaps, 13 and 15), Eq, (12.19) bas 4? replaced swith «? (« is the epicyclic trequency). Note how rotation, by means af Coriolis forces, is able ‘to suppress Jeans instability (if 4? > 4: Gpp) on the plane perpendicular to the rotation axis. If we refer to typical values for po and c, that are applicable to the solar neighborhood, we find a Jeans length close to 1 kpe. This indeed compares well with the thickness of the disk; along the vertical direction the rotation of our Galaxy has no effect (see also Chap. 14). If we are interested in generalizing the analysis to a thin-sheet geometry (such as a homo- geneaus slab of the type discussed below in Section 12.3, in the limit in which the thickness becomes vanishingly small), the Poisson equation for even density perturbations , = 04(2) {to replace Eq, (12.4)] becomes, 2, dz where & refers to a Fourier analysis in the plane of the sheet. This is solved under the condition of evanescence ut |z| — 00, leading to the Gauss jump condition ~ PO, =42Go5t2), (12.20) “lk = 2nGey. 221) The resulting replacement of ay with {kjoo/2 in the final dispersion relation is a signature of surface waves, which we have already met in the context of water waves in the previous chapter. These simple generalizations are the starting point for the study of density waves in galaxy disks, as will be discussed in Chap. 15.12.2 Homogeneous kinetic model and Landau damping l4t 122 Homogeneous kinetic model and Landau damping ‘We now consider the Jeans instability analysis in a homogeneous kinetic model.’ We refer to an isotropic equilibrium described by a Maxwellian distribution So = pol 2c) exp(—v*/2c"), (12.22) Equations (12.1}-(12.3) are replaced with the linearized collisionless Boltzmann equation for fiz DA _ afi “Dp Tag TM = PUT A, (12.23) which is a simple linear hyperbolic equation (see Section 11.1), Thus in this linearized analysis, Jf. is taken to evolve along the unperturbed orbits, that is, along the characteristics given by sttaight lines at velocity v. The Poisson equation in terms of fi is vo, aan ff fv. (12.24) In this simple case we Fourier analyze in space and time; in particular we (ake ©; = , exp (—iwt + ikx). The above equations are then reduced ta kv (kW, = fob, (02.25) é by awd f fide. (12.26) Po The formal dispersion relution obtained when ©; is eliminated is , et r x 2. = exp(—x")dx, (12.27 7 Fthax—e oP with kya 2, (12.28) (12.29) © alkie’ The integral in Bq. (12.27) is singular at x = ¢, and therefore the result demands a physical interpretation. The Landau prescription® consists of enforcing » causality requirement by defining the imegral by analytic continuation starting from the physically justified case of growing modes. Thus the dispersion relation is, 2 Ea1teze, 41230) Ky where Z is the plasma dispersion function ?142 Jeans instability Re(S) Fig, 121, The plasma dispersion function for Imig) = 0, 12.2.1 The plasma dispersion function ‘The plasma dispersion function (see Fig. 12.1) can be written as!? Z)— newer f exp(—A}dt = iy exp(—o7Vf1 + erfliz)]. It satisfies the differential equation 24 UZ+2=0, 20) sive or 2042625 4 2Z = 0. Sometimes we refer to the associated function [suggested by Eq. (12.30) Zz Wea -U+iZ20= 5. 2.31) (12.32) (12.33) 2.34)12.3 Kinetic analysis of a plane-parallel slab 143 ‘The plasma dispersion function has two interesting asymptotic fimits, one far | & 1, 26) = ifr expl—7) — 2f + = + OCC}, (32.35) and one regular limit for \f | 3 1, toa 6 3 Ze) = ~ at oe”), (12.36) 122.2 Marginal stability and the fluid limit ‘We derive the marginal stability condition by considering the dispersion relation in the limit of the asymptotic expansion given by Eq. (12.35). Thus we find that @ pov i hive, 1237) a7 so that k?/3 <1 gives ¢; > 0 (growing in time) and k?/43 > 1 gives fr <0 (damped in time), ‘This behavior is markedly different from that of the fluid mode} [see Eq, (12.7)]. where at the threshold of Jeans instability we move from instability to neutral oscillations. This propery characteristic of the coitisioniess Boltzmann equation is called Landau damping. In the stellar dynamics of galaxy disks (see Sections 16.4 and 19.2) we will find interesting phenomena associated with it, We obtain the fluid limit of the dispersion relation by considering the opposite regime, in which |g] 2% 1, because, under this condition, the details of the distribution Functions are lost and the Maxwellisn is seen as a fiuid beam,'” Then the relevant dispersion relation becomes (12.38) (12.49) which should he cumpared with the result of the fluid model given in Eq. (12.7). 123 Kinetic analysis of a plane-parallel stab ‘We now consider an equilibrium configuration made of a plane-parallel slab, homagengous in the horizontal directions and inhomogeneous in z, so that the relevant integrals of the motion are ¥, and £, =(1/2)u}-+ (2). Thus we ean take fo = fa(v., E.). For the following discussion no more is written about the equilibrium; we just assume that the vertical orbits ate bound and periodic, with bounce frequency wp = w»(E;). A detailed study of a fully self-consistent ‘asic state of this type will be given in Section 14.1. Only some key steps of the related kinetic stability analysis are outlined here)?144 Jeans instability The linearized collisionless Boltzmann equation for a perturbation f; in terms of the per- turbed potential ; is (see Eg. (12.23)] Df _ af Df, 9@o df, a Ue, pe — SO = (HOV, Br ap TUNA a wae Gea = ON Ve) (12.40) ‘We can rearrange the various terms and Fourier analyze in time and space, except for the vertical coordinate, so that Di ao, fo Spe Tn PUM Sal eT a (124y Do, ae, a = EK, fobs + (ee a v.ve.0} 2 12.42) +i (038 +kiv, n) oy (12.43) In Bq. (12.41) we have made use of the identity 3fo/8v, = veiify/PE:. In Eq. (12.43) we have introduced the natural Doppler-shifted frequency & = « — kv. This equation can be integrated along the unperturbed characteristics (see Section 11.1) ta obtain fz ine fi (038 + hiv. f) f_atowres, my (12.44) ‘The first teem on the right-hand side of Eq. (12.44) is traditionally called the adiabatic response, although it is only a part of exe actual response in the adiabatic zero-frequency limit (see discussion of this issue in Subsection 22.2.3). The integral appearing in the second term is meant to be calculated with integrals of the motion fixed at vy and E,. The structure of the response fi to the perturbed potential ©; identified here is very general and will be discussed again in Part Three (see especially Subsections 5.1.1 and 16.4.1) and in Part Four (see Chapter 23, especially Section 23.1). The chapter concludes with the following short digression on the origin of resonances in stellar dynamics, 12.3.1 Bounce orbit expansion The integration along the unperturbed characteristics is trivial in the horizontat variables, because we just have to take [see also Eq. (11.69) x, =k 4-2, 2.45) As u result the relevant integral becomes , a f dt O80), "') = exp(-iet + #KLKL) f adrdy(cexp(—iar). (12.46) Because the z motion is periodie, with bounce frequency an = w»(E.), ©: (2) is periodic 100, and we can expand it as 42 G2) = > &, explinwye), 2.47) teProblems 145 $0 that the integral becomes f af uxie).t) ten ° = exp(—for + ikix,) os 8 f dr expli (ia, — &)r), (12.48) thus giving rise to an infinite number of resonances. These should be dealt with following the Landau prescription introduced in Section 12.2. References and further reading 1 Jeans, JL. 1902, Philos. Trans. R, Soc. London, 199, 1; 1929, Astronomy and Cosmogony, Camisridge University Press, Cambridge, UK; see also Chap. Kil of the monograph by Chandrasekhar, S. 1961, Hydrodynamic and Hydrmagnetic Stability, Oxford University Press, Oxford, UK. 2 These are often referred to as Lanymwir waves; see Stix, TH. 1962, The Theory af Plasma Waves, McGraw-Hill, New York; Krall, N.A., Trivelpiece, A.W, 1973, Principles af Plasma Physics, McGraw-Hill, New York, 3 See Chapter 4 and Chapter 8 of Krall, N.A., Trivelpiece, A.W, 1973, op. cit. 4 Tonks, L., Langmuir, I, 1929, Phys. Rew, 33, 195; Bohm, D., Gross, EP. 1949, Phys. Rew, 75, 1851 5 In reality, the shielding mechanism is generally shared by both electrons and ions, because in the zero-frequency Limit both components necessarily participate in the process. 6 For the generalization to a differentially rotating fluid, see Schatzman, B,, Bel, N. 1955, Comptes rendus des séances de t’Academie des Sciences, 241, 20, for the generalization to the magnetobydrodynamical context, see Chandrasekhar, §., Fermi, E, 1953, Astrophys. J. 118, 116. 7 See the excellent discussion by Lynden-Bell, D, 1967, in Vol. 9 of Lectures in Applied Mathematics Series, American Mathematical Society, Providence, RL. 8 Landau, L.D. 1946, 4. Phys. USSR, 10, 25, 9 Fried, B.P., Conte, $, 1961, The Plasma Dispersion Function, Academic, New York. 10 See Abramowitz, M., Stegun, I. 1970, Handbook of Mathematical Functions, Dover, New York IL See Krall, N.A., Trivelpiece, A.W, 1973, op. cit. 12 A thorough analysis has been given by Mark, J.W.-K. 1971, Astrophys. J., 169, 455. Problems 1. From the equations of the fluid model given in Chap. 8, prove that the frequency of waves in a diffecentially rotating disk should enter the dispersion relation in terms of the so-called Doppler-sbifted value w—ms2, where me is the integer number of the azimuthal Fourier analysis, and Q is the unperturbed angular velocity of the disk.146 Jeans instabitity 2. Derive the asymptotic behavior of the plasma dispersion function Z(¢) for small and large arguments, Note that for large arguments the asymptotic expansion is subtle and has to be qualified in relation to the region of the complex plane considered; the expression given in the text (in Subsection £2.2.1) describes the regular limit, which is the one to be used in the derivation of the fluid limit of approximation (12.39), 3. In the standard cosmological model, elementary linear density perturbations with wave number & are governed by a simple differential equation (see Peebles, PILE. 1993, op. cit.): Bb. + 2H = (Gp — Me} /a*yp,. Here H = a@/a is the Hubble expansion rate and a is the cosmological expansion parameter. The standard Jeans dispersion relation is recovered by setting H = 0 and a = 1. Consider the cold Binstein-de Sitter limit (c, > 0; 2g = 0, Qp = 0, 80 that H? = 87 Gp/3 and ao £7", see also Section 4.4). Find the two power-law solutions that describe linear perturbations in this case. (The solution growing in time describes the mode of Jeans gravitational collapse.) Why is the standard exp(—iart) Fourier analysis not useful in this context?Part Three Spiral galaxies13. Orbits The theory of orbits is iraportant in the study of the dynamics of galaxies, Because the relaxation times for the relevant processes are very long (see Chap. 7), much of the interest lies in the study of orbits in a mean field. What is learned from these studies applies beyond the case of single particles, for the evolution operator in a continuum description (either in a uid or ina steltar dynamical description; see Chap. 8) can be essentially identified with the Hamiltonian that governs single-particle orbits. Galaxy disks are relatively cool systems, that is, most star (or gas cloud) orbits are very close to being circular, so that a typical star (or gus cloud) velocity at a given location is very close to the average velocity of rotation of the disk. A proper description of a collection of orbiting particles should be given in terms of a distribution function (set Chap. 14). However, the physically intuitive properties of suc coo! disks can be easily traced back to the characteristics ‘of quasi-circular orbits of individual particles, which are outlined below in this chapter, The deviations from circular orbits are called cpicycles, To lowest order they correspond to a harmonic oscillator in the radial direction characterized by the epicyclic frequency «, In cool disks the typical radial velocity of a star ¢ is much smaller than the local average velocity u. Thus the epicycle is small because the typical radial excursion c/x is much smaller than + Note that the typical restoring force associated with the radia) oscillations (which is basically determined by the conservation of angular momentum) is cx. which tums out to be of the ordez of the vertical force that is due to the self-gravity of the disk 2 Go, This point will be further discussed in relation to the problem of density waves in Chap. 15. The epieyclic motion is analogous to the Larmor oscillation of a charged particle in a mag- netic field (see Fig. 13.1). The analogy is the basis for the development of common tools of investigation, as described at the end of this chapter, These studies separate the orbit into two parts, the orbit of a guiding center, which reacts to perturbing forces perpendicular to the gyration axis in terms of drifts rather than accelerations, and the rapid oscillations (Larmor or epicyclic) around it. One useful concept of classical dynamies that has found wide applications in the description of the motion of eliayged particles in plasma physics is that of the adiabatic invariant," A well-known example is that of the ratio of the energy associated with the Larmor oscillation and the cyclotron frequency. Under appropriate conditions, this quantity is approximately conserved when the particle moves in an inhomogeneous and/or time-dependent field, which is the key to a simple interpretation of a number of interesting phenomena, such as the trapping of particles in a magnetic mirror. This has a simple counterpart in the study of star orbits in a cool disk.Fig 13.1, Decomposition of a quasi-cireulur orbit into circular and epieyctie motion (upper figures; after Bertin, G., Lin, C.C. 1996, Spirat Srructure in Galavies: A Density Wave Theory, MIT Press, Cambridge, MA, p, 70: © 1996 Massachusetts Institute of Technology), Sketch of a Larmor gyration of a charge moving in a magnetic field and the phenomenon of magnetic trapping in a mirror field (lower figures, ‘tose figures are taken irom Schmidt, G. 1979, Physics of High Temperature Plasmas, 2ad.ed., Academic, New York, pp. 19, 22: © 1979 Academic Press Ine.) From she empirical point of view, the best chance we have for a direct measurement of the properties of individual stellar orbits in a galaxy disk is provided by the solar neighborhood, However, the systematic motions occurring in the disk of our Galaxy are not easily derived from the observations, simply because the Sun is part of such general flow, A classical study by Oort® provided the empirical demonstration that our Galaxy is in a state of differential rotation, as proposed earlier by B. Lindblad. The basis for that study was the idea of expanding the avetage velocity field (flow field) in a Taylor series around the location of the Sun, from which we could easily express the properties of the average motion relative to the Sun expected in the various directions in the sky, along the galactic plane, and at various distances from the Sun. In formulas, Up ~ RIK + Ccost2i) + Asin(@)], (3) U, ~ RUB + Acos(2i) — C sin(2!)], (13.2) where & and ! denote distance from the Sun's location and the galactic tongitude. For a generic13.1 Action and angle variables Ist flow field 0, the four constants that result from the Taylor expansion can be written as As te ae rel, (33) = pe -F.- (13.4) ce [feted L (13.5) z (ae 1, (13.6) Because the disk is cool, by observing the motion of groups of slars in various directions we can easily check that the flow is basically a pure differential rotation; that js, we find K * 0 and C * 0. Furthermore, the parameters A and B, called the Oort constants, obtained by fitting the data, give the local properties of the galactic rotation curve, Note that for a pure differential rotation u, = 0 and up ~ rS%(r), so that , 2 a-(54) . «(3% =| . (13.7) ° war |, ‘These and other kinematical studies provide a rather detailed picture of the orbits of stars in the solar neighborhood,? where itis found that @ < « < ©, (52 * 26 kav(s kpe) so that the rotation period is 240 Myr; based on the Oort limit for the local mass density, described in Chap. 14, the vertical period of oscillation is 70 Myr], The vertical frequency of oscillation about the galuctic plane is thus the highest. 13.1 Action and angle variables For the dynamics of disk galaxies the basic situation to be considered is when 1 mean-field potential © is stationary and axisymmetric around the z axis, Note that such » potential is very different from the Keplerian potential generated by a point mass located at r = 0. In the equatorial plane defined by z = 0, the culculation of orbits is reduced to a one-dimensional problem by the introduction of an effective potential e Gar, = ON + 32 13.8) so that the energy integral can be written as 1 Ba Spt Oe. 3.9 ‘Thus the radial momentum (in our case this is identified with the radial velocity) can be expressed as a function of r and the integrals of the motion £ and J. where J is the specific angular momentum, For'a karge class of potentials, the function em exhibits one minimum at152 Orbits zB Ee Fig. 13.2, Sketch of the effective potential for equatorial orbits in an axisymmetric field, = ro (see Fig. 13.2), which identifies the radius of circular orbits with angular momentum. J Tewe take J > 0 and define ldo r= ldo 3.10) ride the guiding center radius is related to the specific angular momentum by F§Q0r0) = J, 3.11) which is generally 2 one-to-one relationship; for ©, to exhibit a minimum at ro, the function J = Fire) defined by Eq. (13.11) must be monotonically increasing. Typically, for a given value of J, bound orbits are associated with energies in the range Ey < E < 0, with Ea= 5A) + Ory) (13,12) the minimum energy, which corresponds to the circular orbit. In the radial coordinate the motion is periodic and takes place between two turning poimts, /in(E,.2) < fo < fou(E, J)-A. radial action variable can thus be set, As = fran E, djdr, (13.13) with the property dE = dd + O,db. (13.14) Here the radial frequency is defined as @, = 2m/t,, with the bounce time given by dr ub xen (13.15) In turn, the angular frequency is defined by J 1 i dr a -(3)- ohtam za (516)13.2 Epicyclic orbits 153 Orbits are closed (in the inertial frame of reference) if the ratio between the two frequencies is rational. 132 Epieyclic orbits In the limit of quasi-ciroular orbits, which can be quantified as [F® 1, 13.17 hb |< 4 2 we have 2,(E, J) «(rp), 26(E, J) + Qlvy), and the radial periodic motion can be approx- imated by a harmonic oscillator with the epicyclic frequency defined by 2 ddiney) 1d se at (+4 cae) Bg Fo. (13.18) {We also have J, ~[E — Eo(ra}]/e(ra), which is analogous tov] /B for a gyrating charged particle in a magnetic field,} Thus if we separate ont the motion of the guiding center by writing r(t)=7— + r(¢) and A(t} — Qiro) +8: (2), we find that the lineasized equation for the conservation of angular momentum leads to ro lt) = 2200) 3.9) thus the epieycles are ellipses characterized by aspect ratio 20/1 (therefore they are usually elongated in the direction of the motion), with the star running in the opposite direction with respect to the guiding center (ie., the motion ia the epicycle is clockwise if the motion on the circular orbit at rp is counterclockwise), Note that from Bq, (13.18) the condition for the stability of circular orbits (¢? > 0) for- mally coincides with the classical Rayleigh’s criterion for the stability of a rotating fluid. A few important special cases should be noted. A pure harmonic potential (i... the mean-field potential qysociated with a homogeneous sphere} implies solid-body rotation in the sense that SP = 4nGp/3 = constant; in this case we have « = 2&2, and orbits are closed in the form of ellipses centered at = 0. A point mass generates a Keplerian potential: ftom the third law of planetary motion we see that in this case « = ©, and thus orbits are closed in the form of ellipses with one focus at r = 0. For galaxy disks, because they are often characterized by a flat rotation curve, the typical relation should be x © 29? and orbits are generally not closed. Some simple cases of orbits with and « in rational ratio are shown in Fig, 13.3. In Chap. 14 itis shown that the velocity distribution for a relatively cool disk, because of the epicyclic constraints, has an anisotropic pressure tensor for which the radial pressure exceeds the tangential pressure by the ratio 492 /x?, For some purposes (¢.g., for some detailed stellar dynamical studies of density waves in which an integration along the unperturbed orbits is performed), itis of interest to have a full description of the epicyclic expansion beyoud the lowest-order harmonic oscillator obtained by approximation of the potential gy with a parabola in ro. We obtain such a systematic expatision’ by introducing an appropriate phase variable 7. To do this, we first consider the154 Orbits QC OO Fig, 133. Quasi-circular orbits when the ratio of angular to radial frequency is rational (3/2, upper left; 2/3 lower lett; ¢, upper right; 1/4, lower right). (Ina frame rotating with angular velocity @, the relevant angular frequency is the frequency in the inertial frame reduced by the value of 2.1 transformation (2, J)—> (a, ro), where the dimensionless epicyclic energy a is given by 2 AB Foro) rake rp) 13.20) ‘Thus the radial momentum can be expressed as a function: Poth ra) = {a7 ro yrg + ASeulro, J) — Seer, DIY”. (3.21) ‘Now we introduce the phase variable 2 by replacing the radial velocity coordinate p, with Ptr, d, Yo) = Yewi7p)a sin A. (13.22) The complete epicyclic expansion is thus obtained by Taylor expansion of Eg. (13.21) around 7 = ry, which, when inserted into Eg. (13.22), gives Pade Lda cod. (13.23) From here we obtain the expression for dr/d? as well, Then from ar dr dh & = arostrnysinn = 2S, (13.24) dd _ F327) de dd a PF aad we obtain the desired expressions for A = A(t) and 9 = 6(¢), which completes the derivation. ‘We can summarize the first terms of the full expansion by noting that (13.25) 2, = eons(t + 3B? + ~). (13.26) 3 Qy = 2%} 4 5 (As + Byla? +13.3 Rotating frame 155 with By = —As + 2A — 1 and Lf, 2dine Ae =n(i42 ‘ « 3(°+35in Lana) 2 diary (13.28) Here we can easily check that for the harmonic oscillator Az — 1/2and As = B; = 0, whereas for the Keplerian case A; = 0, Ay = 0, und B; = —1. Note that for the isochrone potential (see Chap. 21) A; = 1 andalll the other A, vanish, Many of these results find application in the study of the dynamics of galaxies. They are also of interest in some simple problems of celestial mechanies, in which the potential © is often close to Keplerian. For example, the potential of the Earth in space (r > rp), because of its flattening at the poles, is approximately given by 2 Or ¥)~ ~ oh - sath rasoos 13.29) where we have retained only the quadrupole term in the general solution to the Laplace equation (here at r = rp the quantity x/2 — represents the geographical Latitude); for the Barth we have Jzo & 10-7, We recall the expression for the Legendre polynamial Pzo(x) = (Gx* —1)/2. The epicyclic theory easily allows us to shidy the precession of the perigee of a satellite on the equatorial plane, where 9 = 2/2. The preceysion rate iy proportional to the difference between « and © 13.3 Rotating frame It should be emphasized that an orbit can appeur open in one frame of reference and closed in another. In fact, suppose we move (0 a rotating frame for which the polar coordinates are ir, ¢), with ¢ = 4 — Q,; here Q, is the angular velocity of the rotating frame. Then orbits are described by the new Elanailtonian (Jacobi integral) Uf a, Pe Has Petey + Or) — pe, (13.30) with pg = J, so that H = E— Jy. In the rotating frame, the important ratio 29/x be- comes 22 — &p}/x, which then may be rational or not, depending on our choice of 2p In the dynantics of galaxies there are sometimes physical reasons that identify a specific value of the angular velocity of the rotating frame. The three important possible conditions of AQ, —2)/« = 1, 0, +1 are often called conditions of inner Lindblad resonance, corotation, and outer Lindblad resonance, respectively (see Fig. 13.4). Thus, in the rotating frame, at the Lindblad resonances enbits appear closed inte ellipses centered at x = 0, This feature and the fact that 8 — «/2 can be approximately constant on « wide radial range led Lindblad to con- Jecture that a two-armed spiral structure could persist as a kinematical wave in a differentially rotating disk (see Fig. 13.5), The shear flow pattern associated with the differential rotation in an axisymmetric disk, with the flow reversal at the corotation circle, is somewhat reminiscent of certain magnetic surface156 Orbies 15 z5 3 v<0 Tt v>0 Tt ve 4p tea Tour Fig. 13.4. Location of corotation and outer Lindblad resonance for a differentially rotating disk with respect to an assigned frequency ©,,. The corotatiom circle divides the disk into two regions, one rotating faster (v < 0), the other rotating slower (e = 0) than 2, (afier Bertin, G., Lin, C.C. 1996, Spiral Siracture in Galaxies: A Density Wave Theory. MIT Press, Cambridge, MA, p, 79; © 1996 Massachusetts Institute of Technology). configurations noted in magnetically confined toroidal plasmas, in which a suitable projection of the magnetic field changes sign on a neutral surface. 134 Trapping at the Lagrangian points We now consider a weakly nonaxisymmetric potential of a form that js seen as stationary in a suitable rotating frame (rotating at angular velocity 2p): b= olr) + Oud). (13.31) For simplicity (and in view of a number of simple applications) we take 1,07, 6) = AW eos[—md + Wir], (13.32) where A.m > O; m is an integer. If the amplitude A(r) varies slowly and the phase wer) is instead rapidly varying, the potential ©, may describe an m-armed spiral pattern with local pitch angle / determined by ini = (13.33) A where we have set k = dy/dr. With respect to the azimuthal coordinate, at a given radial location the perturbing potential presents m maxima and m minima. ‘The perturbing potential ¢, is taken to be small, that is, we consider the case in which the nonaxisymmetric forces are small: mA = «1 «13.34 aay < > &13.4 Trapping at the Lagrangian points 157 Fig. 13.5, lltusteation of Lindblad’s kinematic density waves (aller Beatin, G., Lin, C.C. 1996, Spirat Structure in Galaxies: A Density Wave Theory. MIT Press, Cambridge. MA, p.72; © 1996 Massachusetts Institute of Technology). Here (2 is defined in terms of Bq, (13.10) based on @). We choose our frame of reference s0 that ¥(ro9) = 0, where re is the corotation radius. Given the stationarity of ©, the Jacobi integral is conserved, whereas in the inertial frame the energy £ and the angular momentum J vary along the orbits, with F = 2,4. In general the orbits are very complicated. However, in the rotating frame we may easily identify 2m points that are stationary, that is, points where a particte released at rest cam in principle stay in equilibrium. These are a generalization of the Lagrangian points that are own in the contest of celestial mechanics. The equations of motion show that these points ‘occur at the stationary points of cb, with respect to the azimuthal coordinate (mg — ¥) = br, with 6 = 0,1.2,...,(2m — 1), so that the tangential force vanishes. The vanishing of the radial force requires that the folLowing relation be satisfied: dA (i ~ BH) —~ CY = 13:35)158 Orbits where b is even or odd depending on whether ©, is a maximum or a minimum (with respect to the azimuthal coordinate). Thus we see that the Lagrangian points occur in the vicinity of the corotation circle. A curious behavior is found when the stability of these equilibrium points is checked. For simplicity, we focus on a case for which A = constant, so that the stationary points are at 1 = rq and @ = hx/m. Then, if we make the linearized stability analysis for the motion in the neighborhood of r = reo, @ = 0 (which under the above assumptions comesponds to a ‘maxinmm of the perturbing potential along the corotation circle), we find the following relation for the eigenfrequencies @ of the normal modes of oscillations; ote (- where & = «9/2, and all the r-dependent quantities are evaluated at the corotation circle. ‘Thus we see that under the normal condition of monotonically decreasing S2(r) the point of maximum of the perturbing potential is stable. For small yalues of €4, one of the two frequencies derived from Eq. (13.36) corresponds to-a modification of the (fast) radial epicyclic oscillation, whereas the new solution, which corresponds to the symmetry breaking and modifies the neutral Gisplacements along the corotation citcle at o» = 0, is a slow libration frequency, given by 22 Of ~ Imes Si) (3) (13.37) 2 ee +e 7 dinr + mea a? + ames a) =0, (13.36) a dine J Note the scaling dy, ~ e4.In contrast, a simitar analysis carried out at the perturbing potential minima at corotation would show that they are generally unstable. ‘The above study of the stationary points for a nonaxisymmetric rigidly rotating potential is analogous to the classical restricted three-body problem in celestial mechanics.° A two-star or a star~planet (in circular orbit) configuration gives rise to five Lagrangian points. Two points are at 60° with respect 10 the two masses on opposite sides on the plane of the orbit. The other three points are on the axis passing through the two masses. If the mass ratio € is small (such as for the Sun—fupiter case, for which the mass ratio is 4/10), then all five points are close to the cororation circle, that is. the circle of the orbit of the lighter mass, Then two of the three axial Lagrangian points are ata distance O(e")r.. from the lighter mass, and the third is on the opposite side with respect to the heavier mass. If € is sufficiently stall (below a threshold value, ie. 0.04), then the wo Lagrangian points off axis are stable, as is well known from the trapping of two separate families of asteroids in the solar system. The libration period scales as €~"”? also in this case, Another application is found in the study of geostationary satellites, many of which are used fortelecommunication purposes. In this case, the modification to the Barth's basically Keplerian potential occurs hy means of a quadrupole term that fits in the description of Eq. (13.32) with W=0 and A(r)~ 1/7? (us can also be checked from the potential theory of the classical ellipsoids; see Chup. 10). This is because the mass distribution of the Earth is not perfectly axisymmetric, The rotating frame for which such nonaxisymmetric perturbations are stationary is obviously the one that makes one tun in 24 h. The weakness of the nonaxisymmetric field13-4 Trapping at the Lagrangian points 189 is quantified by the fact that the libration period about the two stable potential maxima is #200 days. The trapping of orbits close to a resonant circle is a rather general process, which would be best described by use of action and angle variables.” For example, at a Lindblad resonance for an m = 2 perturbation, a trapping phenomenon occurs that refers to the major axis of the elliptical orbits that are closed in the axisysnmetric case but that are not closed because of the presence of ,. Thus for an m = 2 stationary perturbation the major axis slowly librates and is trapped around an appropriate direction, while it precesses away from a direction at 94° from that. 134.1 Trapping of a charged particle on the top of a hill ‘The above discussion should serve as a reminder that the presence of rotation may bring in some counterintuitive features in the properties af orbits, A simple problem in mechanics, in which the presence of a magnetic fleld replaces the role of rotation, presents some analogies with the above case. It should also be stressed that the concept of stability investigated here refers to cases in which no dissipation Is present. From these examples it could also be appreciated wity, in some circumstances, dissipation may have a destabilizing role* (see also Chap. 10). Consider a charged particle (charge q and mass m) constrained to move on.a hill described by an axisymmetric surface z = z(r) in the presence of a constant vertical magnetic field B = Bye, and constant gravity g = —ge,. Then in polar cylindrical coordinates the energy is Em Sut ae a ceed mgs, (13.38) which, because of the conservation of canonical momentum associated with the axial symmetry, pom e Lo, (13.39) can be written as m Baste yw t ba (13.40) with ¢ 2 eal Pe (o» - Eo) + mgr). a4 Suppose that xr) = Cr" 03.42) with 1 < @ < 2and C > 0. Then the orbit of the particle stays close to the top of the hill, which behaves as a stable equilibrium point even though it corresponds fo the maximum of the potential energy.160 Orbits Mation in frame where Ey%0 Motien in frame where Ey=0 jon oot US Mognetic feld, evr of tepove 28s Magnetic field out of the age 8 2 28. Elec Feld vericelly upward \ | 5 \ wecirers a | CALLA i I Srronger megnetic tiela, smatier redivs of curveture Magnetic field cut of page BES, Weoker magnetic field lorger roduus of curvature i By srodient of Fig. 13.6. Qualitative representation of the Ex B drift (above) and magnetic gradiem: drift (below) for -cbuarged particles (from Krall, N.A., Trivelpiece, A.W. 1973, Principles of Plasma Physies, McGraw-Hill, New York, pp. 612, 625; with kind permission of the authors). 135 Equations for the guiding centers 13.5.1 Larmor oscillations and drifts We briefly recall the basic idea about the guiding center description of particle orbits? (see Fig. 13.6). Consider a charged! particle (charge g and mass m2) moving ina constant magnetic fiekd B in the presence of a constant force F: a vg ae tlt Bt (13.43) ‘We can separate the equations in the parallel and perpendicular directions (with respect to the13.5. Equations for the guiding centers 161 magnetic field) to obtain mitt =, (13.44) “se (vy, « B)+ Fy; (13.45) then we separate the motion of the guiding center vq from the Larmor oscillation hy means of vy=vtw, (13.46) (13.47), ‘The resulting guiding center motion is given by the relation vo = CheXB 13.48) from which it is evident that a porpendiculur force produces a velocity and not an acceleration. The more general ithomogencous and/or time-dependent case in which B and F are not constant can stil) be worked out with a similar strategy if the resulting drifts turn out to be slow and the basic gyration frequency is sufficiently high. For example, a gradient in B can be ubsorbed in F, in Eq, (13,45). Some effects, such as those that are due to the curvature of the field, may require some detailed analysis, which is not discussed here.'® When the curvature is negligible, the parallel motion is basically reduced to the solution of the one-dimensional problem B= Fup +B + U —comtam, (13.49) where U is the potential associated with Fy and p. = mv3/28 is the adiabatic invariant associated with the Larmor gyration. Essentially the basic condition for this type of description is that the variations in the magnetic field encountered by the particle on its orbit must be slow with respect to the relevant cyclotron frequency. 135.2. Star drifts and stellar hydrodynamics As in the case of Larmor gyrations of charged particles in a magnetic field, the epicyclic theory can be extended to the case of weakly nonaxisyrametric and weakly time-dependent fields. This tool allows us to gain a physical perception of the general properties of orbits without resorting to aumerical surveys and to move well beyond the simple small oscillations that are consitlesed in the standard stabitity analysis of the Lagrangian points, as was briefly given in the previous section. The specific theory in the context of stellar dynamics,'! formally developed for cool disks in the vicinity of the relevant corotation radius, is based on the generalization of the definitions of angular and epicyclie frequency as 2 13 lang sor (143328), (13.50)162 Orbits where’now ® is the suin of an axisymmetric potential and a weakly nonaxisymmetric and weakly time-dependent perturbation. Thus the guiding cemter orbits [ro(¢), do(#)] can be derived from an effective Jacobi integral Ho 23 Eghra. dyn 2) — 162070, dp, 1p» a3.sy, where Ey generalizes the concept of energy associated with the circular orbits Eq. (13.12): 1 Eo = 57520, do, 1) + Bor dost. (13.52) ‘The relevant equations of the metion are 22 Hy 79 387 = eh ae 13.53) rox? Dy 8? aE (1353) 20 BH =e 5 to Ore (BS) ‘The second term on the right-hand side of Eq. (13.53) comesponds to the well-known polatiza- tion drift of plasma physics, It should be stressed that in the presence of rotation the perturbation forces induce drift velocities in the motion of the guiding centers and not accelerations. We then obtain the complete properties of the orbits by combining the information om the motion of the guiding centers with the fact that the adiabatic invariant jc is essentially constant: = Ha Holro. bot) 3.55) «70, bos!) where H is the star Jacobi integral, which depends on the physical coordinates (r, ) and the conjugate momenta. When the perturbing potential is independent of time, the guiding center orbits are then simply obtained as contours of the function Hoéry, ¢y)- The Lagrangian points of the star orbit analysis described atyove are then recovered as stationary points for Ho; their stability properties are easily reconstructed by inspection of the relevamt Hessian. If we take as a perwurbation two-armed spiral potential of the form given in Eq. (13.32), we find thut the contours of constant Hy identify two islands at corotation connected by a separatrix passing through the astable potential minima (see Fig. 13.7). Outside the islands, moving away from corotation cither inside or outside the corotation circle, the contours form distorted circles, This new shear flow configuration should be compared with the unperturbed shear flow. The trapped orbits at the stable Lagrangian points define a kind of cat's eyes." There is a clear analogy with the structure of magnetic islands that originate in plasma configurations by means of magnetic reconnection.!* The concepts introduced in this section find many applications in the context of magnetically coufined plasmas, in particular in the description of tapped and circulating particles in toroidal plasma configurations." These sources of analogies are not discussed here any further. In clos- ing, another development, which is conceptually very interesting. is briefly touched on. For a collisionless system it is possible to construct Guid equations from the moments of the colli- sionless Boltzmann equation (see Chap. 8). In the absence of collisions, a well-known problem13.5 Equations for the guiding centers 163 Fig, 13.7, Vorticity islands marked by the guiding center orbits (in the corotating frame) in an axisym- metric disk in the presence ofa two-armed rigidly rotating spiral feld (from Berman, R.M., Mark, W.-K. 1979, Astrophys. J, 231, 388), The structares created are reminiscent of magnetic islands that may fesult from reconnection in sheared magnetic configurations o: cat's eyes zhot may originute from a shear flow. The lower frame shows full orbits, which can be seen as « result of the superposition of the slow zibration of the guiding center and the rapid epicyclic oscillation (see also Contopeules, G. 1973, Asirphys. J 181, 657).164 Orbits is how to clase the fluid equations into a finite set or, in more plyysical terms, how to define an appropriate equation of state. This question has found a solution in plasma physics in terms of the double adiabatic theary,'° which makes use of the conservation of the adiabatic invariant to set a constraint equivalent to that of an equation of state. This leads to the justification of “MHD-like equations for a collisionless plasma, with the peculiarity that pressure is considered anisotropic. It should be stressed that the closure is obtained under a set of assumptions that make the double adiahatic theory applicable to only a rather limited class of perturbations. Still the procedure is very interesting, especially from the physical point of view. A similar theory has been developed in the context of the stellar dynamics of galaxy disks.'* 136 Exact orbits and the problem of self-consistency Many interesting aspects of the theory of orbits relevant to the dynamics of spiral galaxies are not treated in this book, In this section at least some key issues that are involved are mentioned and then the subtle step of moving from the one-particle description to collective bebavior is discussed. 12.6.1 Three~limensional orbits and polar-ring galaxies ‘One important study addresses the properties of orbits outside the galactic plane. Here the basic reference case is thal of an axisymmetiic potential, with the equatorial plane defined by the galaxy disk. The field is in general significantly nonspherical, mainly because of the presence of the disk, but also as a result of the fiattening of the bulge-halo component. Thus, even for the quasi-planar case, we expect out-of-disk orbits te be subject to a wobble, that is, 4 precession, In general, a description of fully three-dimensional orbits requires the. use of numerical integrations. Some well-known methods of classical mechanics, such as the inspection of the relevant surfaces of section, help to point out the integrability properties of the assumed potential These investigations have found at least two important applications. In one study, which addresses mainly quasi-planar orbits, the relation between vertical and equatorial frequencies opens the way 10 a discussion of bending waves (see Chap. 19). much as the case of epieyclic oscillations properly introduces the analysis of density waves in the disk. Here one point should be stressed because it defines a major difference between bending and density waves. ‘The vertical restoring force on a star just off the galactic plane is dueto twocomtributions. One is associated with the inhomogeneity of the disk and with the spheroidal bulge-halo companent; even for a spherical field, at « radial location r on the disk the equatorial force és given by Wr, which translates, by projection, into a restoring force just off the plane Fi" = 272. However, the local disk self-gravity contributes (even in the absence of such a term; sec also Chap. 14) a force Ff?! dor Goz. The fact that 82, in the solar neighborhood exceeds & by a factor of approximately TO shows that at least in the middle of the optical disk, the local. self-gravity term F°? should generally dominate. However, this strong restoring force is not the one relevant for the problem of bending waves, because these waves displace the disk as aLG Exact ackits and the problen of selfcansistency 165 oe e Fig. X28. Optical images of SO galucies: with “polar rings”: from ieft to right, starting om the tap com UGE F576, A OF WNBOL, AM 2026-5050, ESO 603-G21, NGC 2685, ESO 188-1Gi2, ESO 474-525, Zee FY icant Schweizer, Fy Whinnore, B.C, Rubia, VC. 1983, Asuna J. 88, 999), There we galues such as NGC 3998 (see Knapp, G.R., van Drie, Wi van Woerdin, A, 1985, Astron. Astroptiys,. 142, 1) where the polar ring is observed ony in AL whole, so that tit local self-gravity af a patch of the disk dows not act on itself when the pateh is displaced from the equatorial plane. The subtlety involved in this point can be rephrased ia the following way. With respect to the limit of zero thickness, a bending wave is intrinsically noplinvar. Ts any case it may be misleading tw assign directdy 4 term Q as te kinewnaticn! ingredient to ihe dispersion relation of bending waves, fn contuast, there is ac doubt that «? does play such a role for density waves, as the pioncering work of B. Lindblad first pointed out. These points will be discussed furtber ju Chap, 19 In a second application, the shservation af kinematical properties of either indivi Jeets (such as globular clusteys) or of gaxcous rings significantly off the galactic plane ean be used to puta constraint on the underlying potential well of the gulagy.!? Indeed, there is ‘a interesting class of galaxies with prominent rings (the so-called polatring galaxies; see Fig, 13.8) for which we have a.chance of additional dynasnical constsaints beyorid thase jsr0- vided by the standard rotation curve associated with the motion of the matenul in the disk, A general working hypothesis in smmy models of gaseous sings is that they are idenGfiod with closed ocbits (for sinyle particles) associated with the underlying potential. In seality. the dynamics of an observed ting is related co the settling of the gas into such an orbix,!* which may involve some sontrivial collective processes. Thus the nelated problem of di- agnostics, that is, the problem of reconstructing the underlying field from observed orbits, is166. Orbits generally marked by severe ambiguities. The astrophysical problem has an additional source of uncertainty, which is the fact that we generally have incomplete empirical knowledge of the orbit under consideration, 13.62. Exact orbits in nonaxisymmetic disks Within the context of orbits in the galactic plane, the axisymmetric case is trivial; the case of nonaxisymmetric potentials opens instead a number of interesting possibilities even in the case in which the perturbation is stationary in a suitable rotating frame. (A nonaxisyrametric potential stationary in the inertial frame is probably of secondary interest from the physical point of view.) In general, the orbits that violate the conditions for the guiding center descrip- tion outlined above must be calculated numerically. Indeed large surveys of orbits have been performed, both in the presence of barred or ovally distorted potentials and in the presence of spiral fields.!? The orbits have thus been classified systematically.” 13.6.3. Orbital response and self-consistency The simple interpretation of spiral arms as being due to 4 collective crowding of orbits (as Lindblad initiatly suggested) has the major merit of focusing the discussion on the problem of quasi-stationary structures as being due to waves,?) with remarkable attention being paid to two-armed structures. However, such kinematical waves bave the obyiaus shortcomings that they do not take into account the velocity dispersion of stars and, more importantly, that, when the spiral arms are set up, the gravitational field is no longer axisymmetric. We will study in Chap. 15 how those difficulties can be overcome. This point naturally introduces the problem of moving from a single-particle description to the relevant collective behavior. In practice, a quite severe problem arises immediately when we address the issue of non- axisymmetric fields, The study of orbits, when thought of in terms of arelatively large collection of orbits, gives essentially an indication of the response of a specific component (such as a set of stars) to an imposed field, The trapping of stars at the Lagrangian points associated with the potential maxima is rather disconcerting, as potential theory tells us that the potential maxima should roughly coincide with the density minima. ‘Then the indication is that the orbits would act opposite to (out of phase with) the assumed spiral field; in other words, there is litte chance that these nonaxisymmetric fields would be naturally supported, We can partially understand this puzzling behavior by noting the general property that stars with guiding cemters close to the separatrix (Le,, those that are barely trapped) spend a long time away from the trapping point, staying indeed close to the region where their presence would be needed to support the assumed field. This is just a suggestion, which shows that the problem of demonstrating how a field can be supported by the orbits that it imposes is nontrivial. Nevertheless, as is recognized in a full analysis of density waves, the physical conditions for the establishment of density waves involve a number of issues (especially for nonbarred spiral modes, in particular the dissipation in the gas) that go well beyond the single-particle orbits.References and further reading 167 Tn any case a systematic program to study whether stationary nonaxisymmetric fields can receive adequate orbital support has been undertaken and there are a mumber of interesting results." To some extent the issue addressed by these investigations is the same as the issue of finding triaxial self-consistent models by direct superposition of orbits, which has led to some notable results in the context of elliptical galaxies.”? References and further reading 1 See Kulsmd, RM, 1957, Pays. Rev, 106, 205; Kruskal, M. 1962, 4. Math. Phys., 3, 806; Northrop, T.G. 1963, The Adiabutic Motion of Charged Particles, Wiley Interscience, New York. 2 Oont, 1. 1927, Bull, Astron, Inst, Neth., 3 (120), 275. 3 The previse numbers are subject to several sources of uncertainties; sce Kerr, R.J.. Lynden- Bell, D. 1986, Mon, Not. R. Astnm, Soe., 221, 1073. 4 See, for example, Chandrasekhar, 5. 1961, Hydroeynamic and Hydromagnetic Stability, Oxford University Press, Oxford, UK; reprinted by Dover, New York: 5 Shu, EH. 1969, Astrophys. J., 158, 505; Mark, J. W.-K. 1976, Astrophys. J., 203, 81. Mark's analysis removes an undesired secular term present in the original derivation, 6 Contopoulos, G, 1973, Astrophys, J, 181, 657, 7 Lynden-Bell, D, 1973, in Dynamical Structure and Evolution of Stellar Systems, eds L. Martinet, M. Mayor, Geneva Observatory, Switzerland, 8 M.D. Kruskal, as quoted in Coppi, B. 1966, in Non-Equilibrium Thermodynamics, Vari- ational Techniques, and Stability, eds. RJ. Donnelly, R. Herman, 1. Prigogine, Chicago University Press, Chicago, p. 259. 9 For example, see Schmidt, G. 1979, Physics of High Temperature Plasmas, 2nd ed,, Aca- demic, New York. 10 See the book by Schmidt mentioned above; see also Krall, N.A., Trivelpiece, A.W. 1973, Principles of Plasma Physics, McGraw-Hill, New York. 11 Berman, R.H. 1975, Ph.D. Dissertation, Massachusetts Institute of Technology, Cam- bridge, MA; Berman, R.H., Mark, W.-K. 1977, Astrophys. L, 246, 257; Berman, R.H., Mark, J.W.-K. 1979, Astrophys. J, 231, 388, 12 Thomson, W. 1880, Nasure (London), 23, 45 13 See, e.g., White, R. 1983, Handbook of Plasma Physics, 1, eds. A.A. Galeev, R.N. Sudan, North-Holland, Dondrecht, The Netherlands. 14 Benin, G., Coppi, B., Taroni, A, 1977, Astrophys, J., 218, 92. 15 Chew, G.F, Goldberger, M.L., Low, RE. 1956, Proc. 8. Sov, London A, 236, 112. 16 See Berman, R.H.; Berman, R.H., Mark, 7.W.-K., op. cit. 17 Sce, eg., Sackett, PD., Rix, H.-W,, Jarvis, B.1., Freeman, K.C. 1994, Astrophys. J., 436, 629. 18 See, e.g., Christodoulou, D.M., Katz, N., Rix, H-W., Habe, A. 1992, Astrophys. J, 395, 113.168 Orbits 19 See, c.g., Contopoulos, G. 1973, Astraphys. J., 181,657; Contopaulos, G., Papayannopou- los, T. 1980, Astron. Astropitys., 92, 33. 20 See Athanassoula, E., Bienaymé, O., Martinet, L., Pfenniger, D, 1983, Astron, Astrophys., 127, 349. 21 A more detailed description of this important point can be found in Bertin, G., Lin, CC. 1996, Spiral Structure in Galaxies: A Density Wave Theory, MIT Press, Cambridge, MA. 22 See Patsis, PA., Hiotelis, N., Contopoulos, G., Grasbgl, P. 1994, Astron. Astrophys. 286, 46 and references therein; see also Lynden-Bell, D. 1979, Mon. Noi, &. Astron, Sac., 187, 101. 23 Schwarzschild, M. 1979, Astrophys. J.. 232, 236; Schwarzschild, M. 1982, Astopirys. J, 263, 599. Problems 1, For a disk with a perfectly flat rotation curve, V = constant, let corotation be al ¢ = feo. Recall that the inner Lindblad resonance and the outer Lindblad resonance locations are defined by the conditions m[2r,.) — Q1/« = —1 and +1. Where ate the inner Lindblad and the outer Lindblad resonances located for m = 2? What happens for m = 1 and for m = 3? Compare with the case of a Keplerian rotation curve, 2, Following the epicyclic expansion procedure outlined in Section 13.2, prove that indeed the quantity Ag has the expression revorded in Eq. (13.28). 3. Define the orbit [r.(r), 6,(7)] along the characteristics (see Chap. 11) in an axisymmetric potential, the conditions r,(+1.) = r,0,(+t,) = £6,. Here the quantities 2t, = t, and 28, are functions of E and J and denote the radial period of oscillation of the stars in the equilibrium potential and the azimuthal angle traversed in such a period. Prove that, to two onlers in the epieyclic expansion for the radial coordinate and to one order for the azimuthal coondinate, we can write Qo) AR, real Ri Re), Bes -afe+ with s = x(ro)e, 0. = 72Lr0)/eiro), te = aw /e(ve), and Ry = nll +coss) + E sins, Rp = [1 — Ase RP — (3 — 2A. Here § = asin sq and 7 = cos s are dimensionless radial and azimuthal epicyclic velocities associated with the orbit at 7 = r, [see Eqs. (13.20) and (13.223]. From the identity 13.Q(r0) = 26, att = Ty, check also that, to first order, 7 = r(1 — 9). (Hint: Replace the r and 9 variables in Eqs. (13.22)-(13.25) with (7) and @(r), integrate the equations, and then impose the desired boundary conditions. This is one key step toward the integration along the unperturbed characteristics of the stellar dynamical equations leading to the dispersion relation for tightly wound density waves; see Chap. 15 and the article by Shu, FH. (1970) cited there.}Problems 169 4, Consider the problem of the stability of geostationary satellites on the equatorial plane ia view of the presence of a small departure from axisymmetry of the Barth's mass distribution and compare this problem with that of the trapping of star orbits at corotation in the presence of a rigidly rotating two-armed spiral field. 5, For the classical restricted three-body problem, with © = 1n/At < J, find the approximate location of the two Lagrangian points close to m {note the singular character of the perturbation analysis involved in the identification of the distance from the smaller mass, O(€"/?)rgo; see Section 8.5 for comments on singular perturbations]. 6. For a system of two stars of equal mass orbiting around each other at a fixed distance, find the Location of the associated five Lagrangian points and discuss the linear stability of orhits in their vieinity (within the framework af the restricted three-body problem).14 The basic state: vertical and horizontal equilibrium of the disk ‘As described in Chap. 9, the concept of the basic state is a key step in the modeling process. As in other contexts, for galaxy disks the identification of an appropriate basic state is made for several purposes. One interesting aspect is that, by studying the internal structure of the basic state, we can clarify the overall constraints imposed by self-consistency; in this respect, the two examples given below, deafing with the vertical and the horizontal equilibrium of a rotation-supported disk, are rather simple (in collisionless stellar dynamics, self-consistency imposes much stronger and less intuitive constraints in pressure-supported systems) but very instructive. The models that are set up can then be used to ft the: data and thus derive information on the structure of galaxies. Furthermore, a broad choice of equilibrium models can form the ‘basis for systematic stability analyses. In this case the basic state is meant to be a tool to study evolution, but it should not be assumed to represent necessarily the actual conditions of the galaxy in the distant past. From the point of view of stellar dynamics, the case of axisymmetric galaxy disks is relatively simple, because for realistic models most orbits are quasi circular and the disk is rather thin. The related description can then be decoupled at two different levels: (1) The vertical dynamics can bbe treated separately from the horizontal dynamics; (2) for the horizontal structure, because the angular momentum essentially identifies the radial coordinate, the problem of seif-consistency can be worked out in two separate steps: (2.1) Choose a density-potential pair and (2.2) find a (quasi-Maxwellian) distribution function that supports the assumed density, In contrast, this procedure of going from p to f would generally lead to nonphysical models in the spherical caye. The disk is generally embedded in a bulge-halo component that tends to act dynamically as a fixed external field, The coo! thin disk can easily camry waves and instabilities to which the hotter bulge-halo hardly responds. Stil, there are interesting effects related to the interaction between the bulge-halo and the disk that should be taken iato account, some of which will be described in Chap. 19, However, even asa frozen extemal force field, the presence of the bulge— halo changes the stability characteristics of the disk considerably and should thus be properly included in the galaxy models. Therefore for realistic disk models the density—potential pairs are not to be regarded as fully self-consistent, that is, the gravitational field is partly generated by nondisk material. ‘As described in Part One, galaxy disks are made of several star components (stellar popu- lations with different kinematical characteristics) and several dissipative components {clouds14,1 Vertical equilibrium wi of cold atomic hydrogen and molecular gas, with a wide range of sizes and distributions, and warmer diffuse gas). Typically, the outer disk (r > 2h,) comptises a region for which gas is dynamically important. Furthermore, the central part of the disk generally behaves more like a spheroidal system than a flat layer, simply because its finite thickness becomes appreciable in relation to the radial distance. Actually, for galaxies that possess significant bulges, such a3 M81, the observations are often unable to distinguish whether the disk component continues all the way in to the center or whether itis replaced there with the bulge altogether. This situa- tion is physically very interesting and complex and poses difficult challenges in the modeling process (see also Chap. 8). The main consequence of these empirical remarks is that some elegant models that are constructed and can be studied in great detail (for example, a stellar dynamical model of a self-gravitating disk with flat rotation curve and analytic distribution function) generally lack some important physical ingredients. 14.1 Vertical equilibrium 14.1.1. The isothermal self-gravitating slab Consider an ideal disk model’ that is homogeneous in the hotizontal directions (x and y coordinates) and inhomogeneous in the vertical direction z. Imagine that the density distribution # is even with respect to the equatorial (z = 0) plane, so that o¢z) = p(—z), with a maximum ‘po = (0) and a monotonically decreasing profile away from the equatorial plane. We wish to find a self-consistent equilibrium solution, that is, a pair (a(z). (z}], underthe assumption of isothenmality. The equilibrium that we have in mind is kinetic (i.e., 2 solution of the Vlasov Poisson system), but it is sufficient co start with the hydrostatic condition ide ae 4 4) de - supplemented by the self-consistency relation sz =47Gp (14.2) and by the equation of state p=ep, (4.3) where c? js.a constant. Using dimensionless quantities based on the natural scale-height defined by 2 = c?/2rGpy, we can combine the above relations to give = erp) 4a) where ) = p/p, ¥ = — fc, and ¢ = 2/2. Note that we have taken ¥-(0) = 0, consistent with the condition (0) = 1.’This nonlinear equation for y is easily integrated once to give Lae? _ 3 (2) = 21 — exp. 14.5)172. The basic state: vertical and korizomtedl equitibrium of the disk where the energy integration constant has been set to 2 to satisfy the zero-field condition on the equatorial plane (dy/¢ }p = 0, requited by symmetry. This first-order equation then leads to, the well-known 2 = (1/ cosh¢}* (see also Subsection 9.1.1 in Part Two). Explicitly, returning to the dimensional variables, we have 4 lexp(~z/zo) + exp(+z/za)?” With this choice of gravitational potential, the self-consistent distibution function for the isothermal slab is P= by = prexp(—0/c*). 4.6) f= foexpl-E/e%), (4.7) where fo = po/(/2me) and & = v2/2+ &(z). Note that the constant projected disk density o is related to the scales of the slab, as 20 oo ayy? a=] pdz=2mm | ade =~poto| |} = 2poco. (14.8) = Io dE Jo In the application of Eqs. (14.4) and (14.5) to evaluate the integral we recognize the Gauss theorem, Note also the two interesting, limits, 2c? . om Rls 2aGarel. 2% to 4.9) (consistent with the notion of a constant field outside a plane capacitor), and o~t PO) (0), where w = 4G py is the natural bounce frequency provided by gravity Note that in a two-component slab model (af stars and gas) in which each component is vertically isothermal, the eigeusolution for the putential @ is no longer given by Eq. (14.6). Nonetheless, an interesting global constraint is provided by the Gauss theorem: 2 a a Bg? (+ay = 5Go.t, + xGo," + 41D. poke’, Pca 14. £0) where the subscripts refer to the gas and the star components, and a =, /a,, is the relative density ratio, following the notation that will be adopted in Chap. 16. Here the thicknesses are defined in such a way that the projected column densities are related to the peak (equatorial densities by 7. =2, (Oz, and ay = 2y(0)zz. This relation thus generalizes Eq. (14.8), which can be scen as the Limit where « = 0. Another tivial limit occurs when the two components are identical to each other, with o, =, = o/2anda = 1. 14.12, The inhomogeneous disk and the study of the solar neighborhood The discussion of the simple isothermal self-gravitating slab suggests that by measuring the Jength scale of the vertical density distribution and the vertical velocity dispersion in the solar neighbochood we should be able to obtain a dynamical determination” of the locul mass density A. Of course our disk is inhomogeneous in the equatorial plane and is made of many41 Vertical equilibriun 173 components, so that the treatment performed above is nat directly applicable, However, the basic principle involved can be extended to mote realistic models. The first difficulty, the inhomogeneity of the disk, can be looked at in terms of successive approximations. Basically, fora thin disk the gradients in the vertical direction are stronger than those in the radial direction by the geometric factor zo, Inthis sense, the homogeneous slab is just the lowest-order description of a two-scale analysis, which can be carried through to higher levels of approximations if so desired. For exanaple, for an inhomogeneous axisymmetric basic state the Poisson equation is PO 1a ad oe 1 wy 2 tthe ( ~) 4nGp, 4.12) which includes a second-order correction to Eq, (14.2) in the thickness geometric expansion. Curiously the correction vanishes at z = 0 for a flat rotation curve, We can best express, the thickness parameter in terms of other frequently used parameters of the gero-thickness fluid model by noting that formally zo/r = €0Q°, where co = 7Ga/re? is the self-gravity parameter and Q = ck/nGo. One analysis? uses Eq. (14.1), with the first term determined by the observations of a well- defined (and, it is hoped, dynamically well-mixed, i.c., steady-state} class of stars, such as K giants, to determine the vertical force term and especially its gradient in the vertical direction, ‘Thus we have an estimate of the left-hand side of Eq. (14.12), which leads to a determination of the total volume mass density on the equatorial plane, The class of stars sed for the vertical force determination is best chosen so that their distance is accurately measured and their vertical distribution is sufficiently spread out to give a good sampling in z. In this latter respect K giants have a velocity dispersion (20 km/s) that is approximately twice that of AQ stars, thus extending approximately four times more in terms of vertical height, The vertical range that is best sarapled turns out to be between 300 and 1500 pe. ort noted a serious discrepancy between observed material and dynamically estimated mass and considered three options: (1) that the unseen material, possibly in the form of molecular clouds, is distributed in a thin layer as the cold gas, (2) that the unseen material is distributed as the disk stars (Le., as the K giants he was using), and (3) that the unseem matter is like Population If stars (with high vertical velocity dispersion of ~50 km/s). Based on a number of considerations, his best solution is the one that fotlows option (2), with total mass density ¥107* g/em? & 0.15 Mofpe’ and local mass-to-light ratio M/L * 2, for which ~60% is in the form of observed gas (20%) and visible stars (40%), whereas ~40% is dark matter. He also noted that the numbet for the total density at z = 0 thus determined is the same when AO stars are used instead of K giants. ‘The project has been revived more recently by Bahcall and colleagues. A new feature in the discussion is that of allowing for the presence of a halo density py, which is taken to be constant in z. The local density ratio €, = pn/pua0) in the solar neighborhood is estimated to be in the range 0,03-0,3. In this simple model, the two-component self-consistent problem of an isothermal slab in the halo background can be solved. For small values of €,, the one- ‘component isothermal solution (14.6) is modified as 4 expt —a/an) + expCt2/200 Prine expl—eptz/20)"1, «14.13)