This page intentionally left blank
RE L ATIVISTIC FIGURES OF EQUI LI BRI UM Ever since Newton introduced his theory o...
49 downloads
1356 Views
3MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
This page intentionally left blank
RE L ATIVISTIC FIGURES OF EQUI LI BRI UM Ever since Newton introduced his theory of gravity, many famous physicists and mathematicians have worked on the problem of determining the properties of rotating bodies in equilibrium, such as planets and stars. In recent years, neutron stars and black holes have become increasingly important, and observations by astronomers and modelling by astrophysicists have reached the stage where rigorous mathematical analysis needs to be applied in order to understand their basic physics. This book treats the classical problem of gravitational physics within Einstein’s theory of general relativity. It begins by presenting basic principles and equations needed to describe rotating ﬂuid bodies, as well as black holes in equilibrium. It then goes on to deal with a number of analytically tractable limiting cases, placing particular emphasis on the rigidly rotating disc of dust. The book concludes by considering the general case, using powerful numerical methods that are applied to various models, including the classical example of equilibrium ﬁgures of constant density. Researchers in general relativity, mathematical physics and astrophysics will ﬁnd this a valuable reference book on the topic. A related website containing codes for calculating various ﬁgures of equilibrium is available at www.cambridge. org/9780521863834. R EI N H A RD M E INE L is a Professor of Theoretical Physics at the TheoretischPhysikalisches Institut, FriedrichSchillerUniversität, Jena, Germany. His research is in the ﬁeld of gravitational theory, focusing on astrophysical applications. M A RCU S A N S OR G is a Researcher at the MaxPlanckInstitut für Gravitationsphysik, Potsdam, Germany, where his research focuses on the application of spectral methods for producing highly accurate solutions to Einstein’s ﬁeld equations. A N D REA S K LEINW ÄC HT E R is a Researcher at the TheoretischPhysikalisches Institut, FriedrichSchillerUniversität. His current research is on analytical and numerical methods for solving the axisymmetric and stationary equations of general relativity. G ERN OT N EU GE BAUE R is a Professor Emeritus at the TheoretischPhysikalisches Institut, FriedrichSchillerUniversität. His research deals with Einstein’s theory of gravitation, soliton theory and thermodynamics. D AV I D P ETRO FF is a Researcher at the TheoretischPhysikalisches Institut, FriedrichSchillerUniversität. His research is on stationary black holes and neutron stars, making use of analytical approximations and numerical methods.
REL ATI V I S TI C FIGUR E S OF E Q UI LI BR IUM REINHARD MEINEL FriedrichSchillerUniversität, Jena
MARCUS ANSORG MaxPlanckInstitut für Gravitationsphysik, Potsdam
ANDREAS KLEINWÄCHTER FriedrichSchillerUniversität, Jena
GERNOT NEUGE BAUER FriedrichSchillerUniversität, Jena
DAVID PETROFF FriedrichSchillerUniversität, Jena
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521863834 © R. Meinel, M. Ansorg, A. Kleinwächter, G. Neugebauer and D. Petroff 2008 This publication is in copyright. Subject to statutory exception and to the provision of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published in print format 2008
ISBN13 9780511413773
eBook (EBL)
ISBN13
hardback
9780521863834
Cambridge University Press has no responsibility for the persistence or accuracy of urls for external or thirdparty internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.
Contents
Preface Notation 1 Rotating ﬂuid bodies in equilibrium: fundamental notions and equations 1.1 The concept of an isolated body 1.2 Fluid bodies in equilibrium 1.3 The metric of an axisymmetric perfect ﬂuid body in stationary rotation 1.4 Einstein’s ﬁeld equations inside and outside the body 1.5 Equations of state 1.6 Physical properties 1.7 Limiting cases 1.8 Transition to black holes 2 Analytical treatment of limiting cases 2.1 Maclaurin spheroids 2.2 Schwarzschild spheres 2.3 The rigidly rotating disc of dust 2.4 The Kerr metric as the solution to a boundary value problem 3 Numerical treatment of the general case 3.1 A multidomain spectral method 3.2 Coordinate mappings 3.3 Equilibrium conﬁgurations of homogeneous ﬂuids 3.4 Conﬁgurations with other equations of state 3.5 Fluid rings with a central black hole 4 Remarks on stability and astrophysical relevance v
page vii ix 1 1 3 3 5 10 13 16 26 34 34 38 40 108 114 115 128 137 153 166 177
vi
Appendix 1 Appendix 2 Appendix 3 Appendix 4 References Index
Contents
A detailed look at the massshedding limit Theta functions: deﬁnitions and relations Multipole moments of the rotating disc of dust The disc solution as a Bäcklund limit
181 187 193 203 208 216
Preface
The theory of ﬁgures of equilibrium of rotating, selfgravitating ﬂuids was developed in the context of questions concerning the shape of the Earth and celestial bodies. Many famous physicists and mathematicians such as Newton, Maclaurin, Jacobi, Liouville, Dirichlet, Dedekind, Riemann, Roche, Poincaré, H. Cartan, Lichtenstein and Chandrasekhar made important contributions. Within Newton’s theory of gravitation, the shape of the body can be inferred from the requirement that the force arising from pressure, the gravitational force and the centrifugal force (in the corotating frame) be in equilibrium. Basic references are the books by Lichtenstein (1933) and Chandrasekhar (1969). Our intention with the present book is to treat the general relativistic theory of equilibrium conﬁgurations of rotating ﬂuids. This ﬁeld of research is also motivated by astrophysics: neutron stars are so compact that Einstein’s theory of gravitation must be used for calculating the shapes and other physical properties of these objects. However, as in the books mentioned above, which inspired this book to a large extent, we want to present the basic theoretical framework and will not go into astrophysical detail. We place emphasis on the rigorous treatment of simple models instead of trying to describe real objects with their many complex facets, which by necessity would lead to ephemeral and inaccurate models. The basic equations and properties of equilibrium conﬁgurations of rotating ﬂuids within general relativity are described in Chapter 1. We start with a discussion of the concept of an isolated body, which allows for the treatment of a single body without the need for dealing with the ‘rest of the universe’. In fact, the assumption that the distant external world is isotropic, makes it possible to justify the condition of ‘asymptotic ﬂatness’in the body’s far ﬁeld region. Rotation ‘with respect to inﬁnity’ then means nothing more than rotation with respect to the distant environment (the ‘ﬁxed stars’) – very much in the spirit of Mach’s principle. The main part of Chapter 1 provides a consistent mathematical formulation of the rotating ﬂuid body problem within general relativity including its thermodynamic aspects. Conditions vii
viii
Preface
for parametric (quasistationary) transitions from rotating ﬂuid bodies to black holes are also discussed. Chapter 2 is devoted to the careful analytical treatment of limiting cases: (i) the Maclaurin spheroids, a wellknown sequence of axisymmetric equilibrium conﬁgurations of homogeneous ﬂuids in the Newtonian limit; (ii) the Schwarzschild spheres, representing nonrotating, relativistic conﬁgurations with constant massenergy density; and (iii) the relativistic solution for a uniformly rotating disc of dust. The exact solution to the disc problem is rather involved and a detailed derivation of it will be provided here, which includes a discussion of aspects that have not been dealt with elsewhere. The solution is derived by applying the ‘inverse method’– ﬁrst used to solve the Korteweg–de Vries equation in the context of soliton theory – to Einstein’s equations. The mathematical and physical properties of the disc solution including its black hole limit (extreme Kerr metric) are discussed in some detail. At the end of Chapter 2, we show that the inverse method also allows one to derive the general Kerr metric as the unique solution to the Einstein vacuum equations for welldeﬁned boundary conditions on the horizon of the black hole. In Chapter 3, we demonstrate how one can solve general ﬂuid body problems by means of numerical methods. We apply them to give an overview of relativistic, rotating, equilibrium conﬁgurations of constant massenergy density. Conﬁgurations with other selected equations of state as well as ringlike bodies with a central black hole are treated summarily. A related website provides the reader with, amongst other things, a computer code based on a highly accurate spectral method for calculating various equilibrium ﬁgures. Finally, we discuss some aspects of stability of equilibrium conﬁgurations and their astrophysical relevance. We hope that our book – with its presentation of analytical and numerical methods – will be of value to students and researchers in general relativity, mathematical physics and astrophysics. Acknowledgments Many thanks to Cambridge University Press for all its help during the preparation and production of this book. Support from the Dentsche Forschungsgemeinschaft through the Transregional Collaborative Research Centre ‘Gravitational Wave Astronomy’ is also gratefully acknowledged.
Notation
Units: G = c = 1 (G: Newton’s gravitational constant, c: speed of light) Complex conjugation: a + ib = a − ib
(a, b real)
Greek indices (α, β, . . . ): run from 1 to 3 Latin indices (a, b, . . . ): run from 1 to 4 Minkowski space: ds2 = ηab dxa dxb = dx2 + dy2 + dz 2 − dt 2 (x1 = x, x2 = y, x3 = z, x4 = t) Metric of a rotating ﬂuid body in equilibrium: ds2 = e−2U e2k (d2 + dζ 2 ) + W 2 dϕ 2 − e2U (dt + a dϕ)2 = e2α (d2 + dζ 2 ) + W 2 e−2ν (dϕ − ω dt)2 − e2ν dt 2 Killing vectors: ξ = ∂/∂t and η = ∂/∂ϕ Fourvelocity of the ﬂuid: ui = e−V (ξ i + ηi ),
= constant
Energymomentum tensor: Tik = ( + p) ui uk + p gik Equation of state: = (p)
ix
1 Rotating ﬂuid bodies in equilibrium: fundamental notions and equations
1.1 The concept of an isolated body An important and successful approach to solving problems throughout physics is to split the world into a system to be considered, its ‘surroundings’ and the ‘rest of the universe’, where the inﬂuence of the latter on the system being considered is neglected. The applicability of this concept to general relativity is not a trivial matter, since the spacetime structure at every point depends on the overall energymomentum distribution. Our aim is to ﬁnd a description of a single ﬂuid body (modelling a celestial body, e.g. a neutron star) under the inﬂuence of its own gravitational ﬁeld. Fortunately, one often encounters such a body surrounded by a vacuum, where the closest other bodies are so far away that an intermediate region with a weak gravitational ﬁeld exists. In such a situation (see Fig. 1.1) one can discuss the far ﬁeld of the body. If the distant outside world (the ‘rest of the universe’) is isotropic, which it is according to astronomical observations and the standard cosmological models, then the line element corresponding to the far ﬁeld of an arbitrary stationary body can be written as follows (see Stephani 2004): ds2 = gab dxa dxb = gαβ dxα dxβ + 2gα4 dxα dt + g44 dt 2 , with gαβ = (1 + 2M /r)ηαβ + O(r −2 ), gα4 = 2r −3 αβγ xβ J γ + O(r −3 ), g44 = −(1 − 2M /r) + O(r −2 ),
1
(1.1)
2
Rotating ﬂuid bodies in equilibrium
Fig. 1.1. The far ﬁeld of an isolated body (adapted from Stephani 2004).
where r 2 = ηαβ xα xβ = x2 +y2 +z 2 . For r → ∞ the metric acquires the Minkowski form, i.e. the spacetime is ‘asymptotically ﬂat’. We stress that the condition of asymptotic ﬂatness as discussed here is a consequence of the assumption of an isotropic outside world.1 M is the gravitational mass of the body and J α its angular momentum. The gα4 term represents the famous Lense–Thirring effect of a rotating source on the gravitational ﬁeld, also called the ‘gravitomagnetic’ effect – in analogy to the magnetic ﬁeld generated by a rotating electric charge distribution in Maxwell’s electrodynamics. In the next section, we shall provide arguments suggesting that the metric of a rotating ﬂuid body in equilibrium is axially symmetric. Therefore, throughout this book, we shall deal with stationary and axisymmetric spacetimes. Under these conditions, the exterior (vacuum) Einstein equations can be reduced to the socalled Ernst equation, which can be attacked by analytic solution methods from soliton theory. However, the full rotating body problem requires the simultaneous solution of the inner equations, including the correct matching conditions. Note that the shape of the body’s surface is not known in advance! The ﬁnal result must be a globally regular and asymptotically ﬂat solution to the Einstein equations, which can only be found by numerical methods in general (see Chapter 3). But, fortunately, there are a few interesting limiting cases that can be solved completely analytically (see Chapter 2). 1 For an anisotropic outside world, it would be necessary to add a series with increasing powers of r to (1.1). The
expressions (1.1), without these extra terms, would nevertheless be a good approximation to the body’s far ﬁeld as long as r is not too large (‘local inertial system’ on cosmic scales). However, for an isotropic outside world, the notion of a body’s rotation with respect to the local inertial system coincides with the notion of rotation with respect to the external environment (the ‘ﬁxed stars’). Later, we shall simply speak of a rotation ‘with respect to inﬁnity’.
1.3 The metric of an axisymmetric perfect ﬂuid body
3
1.2 Fluid bodies in equilibrium We want to consider conﬁgurations that are strictly stationary, thus implying thermodynamic equilibrium and the absence of gravitational radiation. This leads us, more or less stringently, to the conditions of (i) zero temperature, (ii) rigid rotation, and (iii) axial symmetry.
Thermodynamic equilibrium would also permit a nonzero constant temperature.2 However, as discussed for example in Landau and Lifshitz (1980), such conﬁgurations are unrealistic. Normal stars are hot, but not in global thermal equilibrium: their central temperature is much higher than their surface temperature and they emit a signiﬁcant amount of electromagnetic radiation. Fortunately, neutron stars – the most interesting stars from the general relativistic point of view – can indeed be considered to be ‘cold matter’ objects, since their temperature is much lower than the Fermi temperature. Hence, our idealized assumption of zero temperature ﬁts very well for neutron stars. Provided that some (arbitrarily small) viscosity is present, any deviation from rigid rotation will vanish in an equilibrium state of a rotating star. For the calculation of the rigidly rotating equilibrium state itself, we may then adopt the model of a perfect ﬂuid, since viscosity has no effect in the absence of any shear or expansion. It will, however, affect stability properties. Moreover, within general relativity, any deviation of a uniformly rotating star from axial symmetry will result in gravitational radiation, which is also incompatible with a strict equilibrium state. For a more indepth discussion of points (ii) and (iii), see Lindblom (1992). Therefore, in the next sections, we shall treat stationary and axisymmetric, uniformly rotating, cold, perfect ﬂuid bodies. 1.3 The metric of an axisymmetric perfect ﬂuid body in stationary rotation In accordance with our assumptions of axisymmetry and stationarity, we shall use coordinates t (time) and ϕ (azimuthal angle) adapted to the corresponding Killing vectors: ξ=
∂ , ∂t
η=
∂ , ∂ϕ
(1.2)
2 Note that in general relativity, the equilibrium condition of constant temperature T is replaced by the Tolman condition T (−gtt )1/2 = constant (Tolman 1934), where the prime denotes a corotating frame of reference.
4
Rotating ﬂuid bodies in equilibrium
where ξ is normalized according to ξ i ξi → −1
at spatial inﬁnity
(1.3)
and the orbits of the spacelike Killing vector η are closed, with periodicity 2π . The symmetry axis is characterized by η=0
along the symmetry axis.
(1.4)
It can be shown that the metric of an axisymmetric perfect ﬂuid body in stationary rotation is orthogonally transitive, i.e. it admits 2spaces orthogonal to the Killing vectors ξ and η (Kundt and Trümper 1966). This allows us to write the metric in the following form (Lewis 1932, Papapetrou 1966): ds2 = e−2U e2k (d 2 + dζ 2 ) + W 2 dϕ 2 − e2U (dt + a dϕ)2 , (1.5) or, equivalently, ds2 = e2α (d 2 + dζ 2 ) + W 2 e−2ν (dϕ − ω dt)2 − e2ν dt 2 ,
(1.6)
where the functions U , a, k and W as well as ν, ω and α depend only on the coordinates and ζ . It can easily be veriﬁed that these functions are interrelated according to α = k − U,
−1 W −1 e2ν ± ω = W e−2U ∓ a .
(1.7)
We also note that U , a (or ν, ω) and W can be related to the scalar products of the Killing vectors, thus providing a coordinate independent characterization: ξ i ξi = − e2U = −e2ν + ω2 W 2 e−2ν ,
(1.8a)
ηi ηi = W 2 e−2U − a2 e2U = W 2 e−2ν ,
(1.8b)
ξ i ηi = − ae2U = −ωW 2 e−2ν .
(1.8c)
We call U the ‘generalized Newtonian potential’ and a the ‘gravitomagnetic potential’. Without loss of generality, the symmetry axis can be identiﬁed with the ζ axis, i.e. it is characterized by = 0 and we have 0 ≤ < ∞,
−∞ < ζ < ∞.
(1.9)
On the axis, the following conditions hold, see Stephani et al. (2003):
→0:
a → 0, W → 0, W /( ek ) → 1.
(1.10)
1.4 Einstein’s ﬁeld equations inside and outside the body
5
At spatial inﬁnity, i.e. for 2 +ζ 2 → ∞, the line element approaches the Minkowski metric in cylindrical coordinates , ζ and ϕ: ds2 = d 2 + dζ 2 + 2 dϕ 2 − dt 2 ,
(1.11)
which means that U → 0, a → 0, k → 0, W →
as 2 + ζ 2 → ∞
(1.12)
2 + ζ 2 → ∞.
(1.13)
as well as ν → 0, ω → 0, α → 0
as
Sometimes we shall use a ‘corotating coordinate system’ characterized by
= ,
ζ = ζ,
ϕ = ϕ − t,
t = t,
(1.14)
where is the constant angular velocity of the ﬂuid body with respect to inﬁnity. It can easily be veriﬁed that the line element retains its form (1.5) or (1.6) with
e2U = e2U [(1 + a)2 − 2 W 2 e−4U ],
(1.15a)
(1 − a )e2U = (1 + a)e2U ,
(1.15b)
k − U = k − U ,
(1.15c)
W = W
and ν = ν,
ω = ω − ,
α = α.
(1.16)
Note that ∂ = ξ + η, ∂t
∂ = η. ∂ϕ
(1.17)
We shall call the primed quantities U , a , etc. ‘corotating potentials’. 1.4 Einstein’s ﬁeld equations inside and outside the body The stationary and rigid rotation of the ﬂuid is characterized by the 4velocity ﬁeld ui = e−V (ξ i + ηi ),
= constant,
(1.18)
where = dϕ/dt = uϕ /ut is the constant angular velocity with respect to inﬁnity. Using ui ui = −1, the factor e−V = ut is given by (ξ i + ηi )(ξi + ηi ) = −e2V .
(1.19)
6
Rotating ﬂuid bodies in equilibrium
Note that V is equal to the corotating potential U , V ≡ U ,
(1.20)
as deﬁned in (1.15a). The energymomentum tensor of a perfect ﬂuid is Tik = ( + p) ui uk + p gik ,
(1.21)
where the massenergy density and the pressure p, according to our assumptions as discussed in Section 1.2, are related by a ‘cold’ equation of state = ( p) following from = (µB ),
p = p(µB )
(1.22)
at zero temperature, with the baryonic massdensity µB . Examples will be given in Section 1.5. The speciﬁc enthalpy3 h=
+p µB
(1.23)
can be calculated from ( p) via the thermodynamic relation dh =
1 dp µB
(zero temperature)
(1.24)
leading to dh dp = h +p
⇒
p h( p) = h(0) exp 0
dp . ( p ) + p
(1.25)
Note that h(0) = 1 for ordinary baryonic matter.4 From T ik ;k = 0 (a semicolon denotes the covariant derivative), we obtain, as a ﬁrst integral of the equations inside the body, h( p) eV = h(0) eV0 = constant.
(1.26)
This means that surfaces of constant p coincide with surfaces of constant V . The boundary of the ﬂuid body is deﬁned by p = 0, hence V = V0
along the boundary of the ﬂuid.
(1.27)
3 Note that = µ + u , where u denotes the internal energydensity. Hence h = 1 + h with h being the B N N int int
speciﬁc enthalpy as it is usually deﬁned in the nonrelativistic (Newtonian) theory.
4 An exception is strange quark matter as described by the MIT bag model, see Section 1.5.
1.4 Einstein’s ﬁeld equations inside and outside the body
7
The constant V0 is related to the relative redshift z of zero angular momentum photons5 emitted from the surface of the ﬂuid and received at inﬁnity via z = e−V0 − 1.
(1.28)
Equilibrium models, for a given equation of state, are ﬁxed by two parameters, for example and V0 . The full set of equations that follows from Einstein’s ﬁeld equations Rik − 12 Rgik = 8πTik for the metric in the form (1.6), with (1.18) and (1.21), can be written in the following way, see e.g. Bardeen and Wagoner (1971):
1 + v2 1 + 2p , (1.29a) ∇ · (B∇ν) − 2 B3 e−4ν (∇ω)2 = 4πe2α B ( + p) 2 1 − v2 v , (1.29b) ∇ · ( 2 B3 e−4ν ∇ω) = −16π B2 e2α−2ν ( + p) 1 − v2 ∇ · ( ∇B) = 16π Be2α p
(1.29c)
with B := W /
and v := Be−2ν ( − ω),
(1.29d)
together with two equations, which provide the possibility of determining α via a line integral if the other three functions ν, ω and B are considered as given, 1
−1 (α + ν), + B−1 [B, (α + ν), − B,ζ (α + ν),ζ ] − −2 B−1 ( 2 B, ),
2 1 1 + B−1 B,ζ ζ − (ν, )2 + (ν,ζ )2 + 2 B2 e−4ν [(ω, )2 − (ω,ζ )2 ] = 0, (1.30a) 2 4 1
−1 (α + ν),ζ + B−1 [B, (α + ν),ζ + B,ζ (α + ν), ] − −2 B−1 ( 2 B,ζ ),
2 1 2 2 −4ν 1 −1 (1.30b) − B B, ζ − 2ν, ν,ζ + B e ω, ω,ζ = 0, 2 2 and (1.26), which allows us to express p and , via (1.25) and the equation of state, in terms of eV ≡ eU = eν 1 − v 2 . (1.31) 5 Zero angular momentum means η pi = 0 ( pi : 4momentum of the photon), i.e. the (conserved) component of i
the orbital angular momentum with respect to the symmetry axis vanishes. In particular, this is satisﬁed for all photons emitted from the poles of a body of spheroidal topology, since η vanishes on the axis of symmetry. For other points on the surface, the condition ηi pi = 0 places a restriction on the directions of emission.
8
Rotating ﬂuid bodies in equilibrium
In (1.29), the operator ∇ has the same meaning as in a Euclidean 3space in which
, ζ and ϕ are cylindrical coordinates. Note that v as deﬁned in (1.29d) is the linear velocity of rotation with respect to ‘locally nonrotating observers’.6 Its invariant deﬁnition is given by v ηi ui = . √ 1 − v2 ηk ηk
(1.32)
In (1.30), we have made use of the comma notation for partial derivatives, e.g. ∂ν/∂ = ν, . Note that instead of (1.30), the second order equation for α α,
+ α,ζ ζ −
1 1 ν, + ∇ν ∇ν − B−1 ∇B − 2 B2 e−4ν (∇ω)2
4
(1.33)
= −4πe ( + p), 2α
which follows from (1.29), (1.30) and (1.26), see Trümper (1967), can be used. For the metric in the form (1.5), the equations take a simpler form if one uses the corotating potentials U , a , k and W . With W = W , see (1.15c), they read
∇U · ∇W e4U (∇a )2 1 + = 4π( + 3p)e2k −2U , ∇ U − U, +
W 2W 2 2
(W −1 e4U a,
), + (W −1 e4U a,ζ ),ζ = 0,
(1.34a) (1.34b)
W,
+ W,ζ ζ = 16π pW e2k −2U
(1.34c)
together with 1 − W,ζ k,ζ = (W,
− W,ζ ζ ) + W [(U, )2 − (U,ζ )2 ] W, k,
2
+
e4U 2 [(a,ζ )2 − (a,
) ], 4W
W,ζ k,
+ W, k,ζ = W, ζ + 2WU, U,ζ −
(1.35a)
e4U a a 2W , ,ζ
(1.35b)
and (1.26). 6 Locally nonrotating observers (also called ‘zero angular momentum observers’) have a 4velocity ﬁeld i uzamo = e−ν (ξ i + ωηi ). They rotate with the angular velocity ω with respect to inﬁnity, but their angular i momentum ηi uzamo vanishes, see Bardeen et al. (1972). This provides a nice interpretation for the metric
functions ω and ν.
1.4 Einstein’s ﬁeld equations inside and outside the body
9
The vacuum case: the Ernst equation Outside the body, the source terms on the right hand sides of Equations (1.34) vanish. Equation (1.34c) becomes a twodimensional Laplace equation: W,
+ W,ζ ζ = 0.
(1.36)
By means of a conformal transformation in ζ space, it is always possible to choose W ≡ .
(1.37)
In these ‘canonical Weyl coordinates’ the remaining ﬁeld equations, written down for the functions U , a and k, are7 ∇ 2U = −
e4U (∇a)2 , 2 2
(1.38)
( −1 e4U a, ), + ( −1 e4U a,ζ ),ζ = 0
(1.39)
together with the two equations k, = [(U, )2 − (U,ζ )2 ] + k,ζ = 2 U, U,ζ −
e4U [(a,ζ )2 − (a, )2 ], 4
e4U a, a,ζ , 2
(1.40a) (1.40b)
which allow us to calculate k via a pathindependent8 line integral. Equation (1.39) implies that a function b can be introduced according to a, = e−4U b,ζ ,
a,ζ = − e−4U b, ,
(1.41)
satisfying the equation ( e−4U b, ), + ( e−4U b,ζ ),ζ = 0.
(1.42)
It can easily be veriﬁed that the two Equations (1.38) and (1.42) can be combined into the Ernst equation (Ernst 1968, Kramer and Neugebauer 1968).
f ∇ 2 f = (∇f )2
(1.43)
7 As a consequence of the form invariance of the line element (1.5) under a coordinate transformation (1.14), the vacuum equations for U , a, k and W are the same as those for U , a , k and W (= W ), and can be read off
from Equations (1.34) and (1.35) for = p = 0.
8 The integrability condition is satisﬁed by virtue of (1.38) and (1.39).
10
Rotating ﬂuid bodies in equilibrium
for the complex ‘Ernst potential’ f := e2U + ib.
(1.44)
The Ernst equation (1.43), together with (1.41), (1.40) and (1.37), is equivalent to the vacuum Einstein equations in the stationary and axisymmetric case. As already mentioned, the vacuum equations for the corotating potentials U and a have the same form as those for U and a. Therefore, the Ernst potential can also be introduced in the corotating system and the Ernst equation retains its form as well. This remarkable fact will be used later. The global problem For genuine ﬂuid body problems, we shall not make use of canonical Weyl coordinates and the Ernst formalism in the exterior region. It is of greater advantage to have a global coordinate system , ζ in which all metric functions and their ﬁrst derivatives are continuous at the surface of the body. In particular, this requirement leads to a unique solution W ( , ζ ), which differs from W ≡ in the vacuum region.9 The global problem consists in ﬁnding a regular, asymptotically ﬂat solution to Equations (1.29) and (1.30) with source terms inside the ﬂuid and without source terms in the vacuum region. We stress that the shape of the surface, characterized by p = 0, is not known from the outset.
1.5 Equations of state In this section, we shall provide some examples of equations of state = ( p), which will be used in this book. The relation to the baryonic massdensity µB , consistent with Equations (1.23) and (1.25), will also be given. Note that in our units (with c = 1), there is no difference between energydensity and (total) massdensity µ, i.e. = µ = µB + uint , where uint is the internal energydensity. Homogeneous ﬂuids This simple model is characterized by the equation of state (EOS) = constant.
(1.45)
Assuming h(0) = 1, we obtain from (1.23) and (1.25) that = µ = µB , i.e. the internal energy density is zero. 9 An important exception is given by the disc limit, where it turns out that W ≡ holds globally, see Subsection
1.7.3. Another application of the Ernst formalism will be the derivation of the Kerr metric in Section 2.4.
1.5 Equations of state
11
Relativistic polytropes This model is deﬁned by 1 (n > 0), (1.46) n see Tooper (1965). Here K is called the ‘polytropic constant’, γ the ‘polytropic exponent’, and n the ‘polytropic index’. With h(0) = 1, we obtain from (1.23) and (1.25) the relation p = KµB γ ,
γ =1+
= µB + np,
(1.47)
i.e. p = (γ − 1)uint and the EOS reads = ( p/K)1/γ + p/(γ − 1). Note that the homogeneous case = constant is contained as the limit n → 0. It should, however, be noted that this EOS – if applied in the dynamic case – guarantees a speed of sound less than the speed of light only for n ≥ 1. Completely degenerate, ideal gas of neutrons The general EOS for a completely degenerate, ideal Fermi gas (a genuine zerotemperature EOS!) was derived by Stoner (1932) in the framework of specialrelativistic Fermi–Dirac statistics. The two limiting cases, the nonrelativistic and ultrarelativistic limit, lead to polytropic relations (1.46) with exponents γ = 5/3 and γ = 4/3 respectively. This EOS, applied to an electron gas, plays a crucial role in the theory of white dwarfs, see Chandrasekhar (1939). Here we have in mind the application to a neutron gas, ﬁrst considered by Landau (1932), and used in the famous work by Oppenheimer and Volkoff (1939) to calculate models of neutron stars. Pressure, energydensity and baryonic massdensity are related as follows: m4n f (x), 24π 2 3 m4n = µB + g(x), 24π 2 3 m4 µB = 2n 3 x3 , 3π p=
(1.48a) (1.48b) (1.48c)
where f (x) = x(2x2 − 3)(x2 + 1)1/2 + 3 arcsinh x, g(x) = 8x3 (x2 + 1)1/2 − 1 − f (x), see, for example, Kippenhahn and Weigert (1990). Here mn is the mass of a neutron and is the reduced Planck constant (remember that we use units with c = 1). It can easily be veriﬁed that h(0) = 1 and that (1.25) is satisﬁed.
12
Rotating ﬂuid bodies in equilibrium
Strange quark matter as described by the MIT bag model This model, in its simplest version, leads to 4/3
= A µB + B,
1 4/3 p = A µB − B, 3
(1.49)
with welldeﬁned constants A and B, see, for example, Gourgoulhon et al. (1999) and references therein (B is called the ‘MIT bag constant’). The resulting EOS = ( p) is very simple: = 3p + 4B. Again, the thermodynamic relation (1.25) is satisﬁed. Note that here
h(0) = = 4B−1/4 (A/3)3/4 < 1, µB p=0
(1.50)
(1.51)
corresponding to the assumption that strange quark matter (also called strange matter) represents the absolute ground state of matter at zero pressure and temperature. The dust limit The dust model is characterized by p = 0,
(1.52)
i.e. the energymomentum tensor reduces to Tik = ui uk .
(1.53)
In this case, the equations T ik ;k = 0 imply geodesic motion of the ﬂuid elements, ui ;k uk = 0,
(1.54)
(ui );i = 0,
(1.55)
and the local conservation law
which allows us to identify the massenergy density with the baryonic massdensity, = µB ,
i.e.
h( p) = h(0) = 1.
(1.56)
1.6 Physical properties
13
1.6 Physical properties 1.6.1 Mass and angular momentum The gravitational mass M and the total angular momentum J (strictly speaking, the component with respect to the axis of symmetry) in an asymptotically ﬂat, stationary and axisymmetric spacetime are given by i k 1 j (1.57) M = 2 (Tik − 2 T j gik )n ξ dV , J = − Tik ni ηk dV ,
where is a spacelike hypersurface (t = constant) with the volume element dV = (3) g d3 x and the future pointing unit normal ni , see for example Wald (1984). Note that ηi ni = 0. The baryonic mass M0 , corresponding to the local conservation law (µB ui );i = 0, is given by the expression M0 = − µB ui ni dV . (1.58)
Nearby equilibrium conﬁgurations with the same equation of state are related by δM = δJ + µc δM0 ,
µc = h(0)eV0 .
(1.59)
This follows from a variational principle (Hartle and Sharp 1967, see also Bardeen 1970 and Neugebauer 1988). The factor µc = h(0)eV0 thus plays the role of the equilibrium value of the body’s chemical potential (in appropriate units):
∂M
. (1.60) µc = ∂M0 J = constant Note that h( p) eV = µc = constant,
(1.61)
see (1.26), is indeed the Tolman equilibrium relation for the chemical potential h.10 The gravitational mass M and the total angular momentum J can also be read off from the asymptotic behaviour of ξ i ξi and ξ i ηi /ηk ηk as r = 2 + ζ 2 → ∞: ξ i ξi = −1 +
2M + O(r −2 ), r
ξ i ηi 2J = − + O(r −4 ), r3 ηk ηk
(1.62)
(1.63)
10 In the zero temperature case, there is no difference between enthalpy and free enthalpy (Gibbs free energy).
14
Rotating ﬂuid bodies in equilibrium
which, for the metric as given in (1.5) or (1.6), means11 U =−
M + O(r −2 ), r
a=
2J 2 + O(r −2 ) r3
(1.64)
M + O(r −2 ), r
ω=
2J + O(r −4 ). r3
(1.65)
or, equivalently, ν=−
In asymptotically Cartesian coordinates x1 = cos ϕ, x2 = sin ϕ, x3 = ζ , this corresponds exactly to (1.1) with J 1 = J 2 = 0, J 3 = J , where one has to take into account that α = −ν + O(r −2 ), which follows from (1.30). 1.6.2 Ergospheres Regions in which the Killing vector ξ i , which is normally timelike (ξ i ξi < 0), becomes spacelike (ξ i ξi > 0), are called ergospheres. The boundary of an ergosphere, also called an ergosurface, is characterized by ξ i ξi = 0. Ergospheres appear when a rotating source becomes sufﬁciently relativistic, i.e. far away from the Newtonian limit. Despite the fact that the Killing vector ξ i , corresponding to stationarity, is spacelike within the ergosphere, the spacetime can still be considered to be locally stationary, provided there exists a timelike linear combination of ξ i and ηi . We can assume that the spacetime of rotating ﬂuid bodies in equilibrium is locally stationary everywhere. It should be noted, however, that this condition is violated inside, and on the event horizon of, black holes. Next we discuss some mathematical and physical aspects of the presence of ergospheres. Mathematical aspects For the metric in the form ds2 = e2α (d 2 + dζ 2 ) + W 2 e−2ν (dϕ − ω dt)2 − e2ν dt 2 ,
(1.66)
nothing special happens with α, W , ν and ω at an ergosurface12 or inside the ergosphere. The local stationarity of the spacetime is guaranteed precisely when e2ν > 0, i.e. the function ν remains real inside the ergosphere. However, if we write the metric in the equivalent form ds2 = e−2U e2k (d 2 + dζ 2 ) + W 2 dϕ 2 − e2U (dt + a dϕ)2 , (1.67) 11 Note that W = + O(r −1 ) as r → ∞. 12 Special attention has to be paid to points at which the ergosurface reaches the symmetry axis. This happens at
the poles of black hole horizons. For rotating ﬂuid bodies, the ergosphere has a toroidal shape in general, and does not touch the rotation axis except for the (singular) limiting case of inﬁnite central pressure.
1.6 Physical properties
15
we have to note that the function e2U = −ξ i ξi changes its sign. Inside the ergosphere, e2U < 0 holds, i.e. the function U is no longer real. The ergosurface is characterized by e2U = 0. The behaviour of e2k and a compensates for the ‘dangerous’ effects of e2U ≤ 0 such that all metric coefﬁcients behave perfectly well at the ergosurface and inside the ergosphere.13 In the vacuum case, one can introduce canonical Weyl coordinates (characterized by W ≡ ) and the Ernst potential f ≡ e2U + ib, see Section 1.4. It is important to note that f behaves perfectly well at the ergosurface, too. This means that the function b, in contrast to a, behaves well. Vice versa, analytic solutions of the Ernst equation (i.e. solutions for which e2U and b can both be Taylorexpanded in the two real variables and ζ ) with zerolevel sets of e2U lead to smooth ergosurfaces in spacetime, see Chrus´ciel et al. (2006). Note that we continue to use the notation e2U for f independent of its sign. Physical aspects The ergosurface is sometimes called the ‘limiting surface of stationarity’ or simply the ‘stationary limit’ since no timelike world lines with a tangent vector (4velocity) proportional to ξ i , which would represent observers that are stationary with respect to inﬁnity, can exist any longer. Inside the ergosphere, the only term in the above line element ds2 , which may become negative, is the term 2gϕt dϕdt. Therefore, a timelike world line (ds2 < 0) requires dϕ/dt = 0, which means that observers must rotate about the axis of symmetry. The direction of this rotation is dictated by the sign of gϕt = −ωW 2 e−2ν : ω
dϕ >0 dt
for timelike world lines within the ergosphere.
(1.68)
For a uniformly rotating source, the sign of the function ω always coincides with the sign of the angular velocity of the source, i.e. any observer must rotate in the same direction as the source inside the ergosphere. It is interesting to note that the Killing vector ∂/∂t = ξ + η of the corotating system, see Section 1.3, becomes spacelike far away from the rotation axis:
e2U = −(ξ i + ηi )(ξi + ηi ) → −2 2
as
→ ∞.
(1.69)
This corresponds to the wellknown fact that no observers that are too distant from the axis can be stationary with respect to the corotating system as this would require superluminal motion. 13 At the ergosurface, e2k vanishes and a diverges. Inside the ergosphere, e2k < 0.
16
Rotating ﬂuid bodies in equilibrium
1.7 Limiting cases The few analytical solutions that can be found for ﬁgures of equilibrium rely on the fact that the problem simpliﬁes signiﬁcantly in certain limiting cases: (i) the Newtonian limit, where one has only a single gravitational potential satisfying the simple Poisson equation; (ii) the nonrotating limit, where the spherical symmetry implies a simple system of ordinary differential equations; and (iii) the disc limit, where a boundary value problem to the vacuum equations can be formulated. In this section we shall derive the relevant equations. The ﬁrst two limits are well known, which enables us to be brief. The disc limit will be treated in much greater detail, since it is less well known and plays an important role in Chapter 2. In addition, a limiting case of a different nature, namely the massshedding limit, is also discussed. This limit poses particular challenges to the numerical methods to be presented in Chapter 3.
1.7.1 The Newtonian limit The Newtonian limit, in our context, is approached when the following conditions are satisﬁed: (i) The metric deviates only slightly from the Minkowski metric. (ii) The linear velocity of rotation v, as deﬁned in (1.32), is small as compared with the velocity of light: v c, i.e. v 1 in our units. (iii) The pressure is small as compared with the massenergy density: p = µc2 , i.e. p µ in our units.14
It turns out that these conditions are all satisﬁed for rotating ﬂuid bodies in equilibrium whenever the absolute value of the parameter V0 becomes sufﬁciently small: V0  1.
(1.70)
The metric function U becomes the Newtonian potential15 satisfying the Poisson equation ∇ 2 U = 4πµ,
(1.71)
which reduces to the Laplace equation in the vacuum region: ∇ 2U = 0
outside the body.
(1.72)
14 This condition implies via (1.23) and (1.25) that µ ≈ h(0)µ , i.e. for all equations of state with h(0) = 1, the B massdensity µ can be identiﬁed with the baryonic massdensity µB in the Newtonian limit. 15 Note that ξ i ξ = g = −e2U ≈ −(1 + 2U ) ≈ −(1 + 2ν) in the Newtonian limit. tt i
1.7 Limiting cases
17
The leading order terms of (1.25) and (1.26) give the Newtonian relation p dp V = V0 − , (1.73) 0 µ(p ) which is nothing other than the integrated form of the Euler equation ∇V = −
∇p µ
(1.74)
with 1 V = U − 2 2 . 2
(1.75)
The latter relation follows from (1.15a) and (1.20) to leading order. In a sense, V can be considered to be a ‘corotating potential’ in Newtonian theory as well (it includes the ‘centrifugal potential’ −2 2 /2). Note, however, that V satisﬁes the equation ∇ 2 V = 4πµ − 22 ,
(1.76)
which does not reduce to the Laplace equation (1.72) outside the body. This is in remarkable contrast to the fact, discussed in Section 1.4, that the Ernst equation retains its form in the corotating system (a nice justiﬁcation for calling Einstein’s theory ‘general relativity’). From (1.73) and p = 0, we obtain the Newtonian surface condition V = V0
along the surface,
(1.77)
just as in general relativity. 1.7.2 The nonrotating limit If one considers static, nonrotating ﬂuid conﬁgurations, the ﬁeld equations take a particularly simple form. Besides the mathematical simpliﬁcation, this assumption is often justiﬁed on physical grounds, since most celestial bodies possess rather small rotation rates and hence a static model is a good approximation. Since the spacetime continuum of a static perfect ﬂuid body is spherically symmetric (see MasoodulAlam 2007), a corresponding form of the line element is appropriate. The ﬁeld equations presented in Section 1.4 become ordinary differential equations with respect to the radial coordinate r that can be introduced alongside ϑ by = r sin ϑ, ζ = r cos ϑ. The line element reads ds2 = e2α (dr 2 + r 2 dϑ 2 + r 2 sin2 ϑ dϕ 2 ) − e2ν dt 2 ,
(1.78)
18
Rotating ﬂuid bodies in equilibrium
i.e. ω = 0 and W = eα+ν , cf. (1.6). These two conditions correspond to staticity and spherical symmetry respectively. However, in order to achieve a particularly concise form, one usually considers the ﬁeld equations in standard Schwarzschild coordinates (˜r , ϑ, ϕ, t): ds2 = e2α˜ d˜r 2 + r˜ 2 (dϑ 2 + sin2 ϑ dϕ 2 ) − e2ν dt 2 .
(1.79)
These coordinates are obtained through the simple, purely radial transformation r˜ = reα . Since the matter is at rest in this coordinate system, we may write
(1.80)
ui = (0, 0, 0, e−V )
(1.81)
V = ν.
(1.82)
with Taking the integrated relativistic Euler equation (1.26) into account, one may derive the Tolman–Oppenheimer–Volkoff equation (Tolman 1939, Oppenheimer and Volkoff 1939): dp ( + p)(m + 4πp˜r 3 ) , (1.83) = r˜ (2m − r˜ ) d˜r through which a ‘mass function’ m(˜r ), dm = 4π r˜ 2 , d˜r
m(0) = 0 ,
(1.84)
is determined. Equation (1.84) provides a relation between this function and the thermodynamic quantities , µB , p and h [see also Equations (1.22)–(1.25)], and by virtue of (1.26) and (1.82), the metric potential ν is also given. Moreover, the function α˜ is obtained through 2m . (1.85) r˜ If for a static perfect ﬂuid model, the equation of state and a physical parameter (e.g. the central pressure pc ) are speciﬁed, then the complete interior solution can be determined through the above equations. The spatial location r˜ = r˜0 of the body’s surface is then given by the condition of vanishing pressure. The metric in the exterior of the body is, of course, given by the wellknown Schwarzschild vacuum solution, with the gravitational mass M = m(˜r0 ). Note that the constant of integration in (1.26) is then also ﬁxed upon demanding continuity of the metric. An important consequence, which can be derived under the reasonable assumption that the energydensity does not increase outwards, is the socalled Buchdahl limit (Buchdahl 1959): A spherically symmetric star can only exist in a e−2α˜ = 1 −
1.7 Limiting cases
19
state of equilibrium (can only compensate its own gravitational attraction with a ﬁnite pressure) if the ratio of its mass M to its radius r˜0 satisﬁes the inequality M 4 < . r˜0 9
(1.86)
Here the stellar radius r˜0 assumes a coordinate invariant meaning by virtue of the relation S = 4π r˜02 ,
(1.87)
where S is the surface area of the star. The inequality (1.86) shows that a spherical star in equilibrium always has a (coordinate) radius greater than 9/8 times the Schwarzschild radius 2M of a black hole of the same mass. Beyond this limit, the star must inevitably collapse.
1.7.3 The disc limit As a rule, a perfect ﬂuid ball set in rotational motion takes on an oblate shape and we may expect that there are extremely ﬂattened ﬂuid conﬁgurations represented by an inﬁnitely thin circular disc rotating about an axis of symmetry (in our context denoted by the ζ axis). Here we shall construct a corresponding mathematical model. Later on, we shall show that the ﬁeld equations are rigorously solvable in this limiting case (see Section 2.3). Exact solutions like this help to achieve deeper insight into the geometrical structure of the gravitational ﬁeld of rotating bodies, facilitate a reliable discussion of physical effects and provide us with the interrelationship between characteristic parameters such as angular velocity, mass and angular momentum. Moreover, the study of the disc limit has astrophysical relevance: Discs play an important role as galaxy models or intermediate states in collapse processes. It should be mentioned that an approximate solution to the disc problem was found (Bardeen and Wagoner 1971) by solving a postNewtonian expansion to high order numerically. The exact solution (Neugebauer and Meinel 1995) conﬁrmed many of the predictions made in this notable paper. The idea of the subsequent analysis is to describe the disc limit of perfect ﬂuids by a boundary value problem of Einstein’s vacuum equations with boundary data derived from the ﬁeld equations inside the body, which degenerates to a circular disc with the coordinate radius 0 covering the domain 0 ≤ ≤ 0 of a threedimensional slice through spacetime – the 3surface ζ = 0, see Fig. 2.3.16 This domain can be considered to be the world tube of the surface elements of the 16 Unessential coordinates t and ϕ are omitted.
20
Rotating ﬂuid bodies in equilibrium
twodimensional surface 2 , 2 :
ζ =0
(0 ≤ ≤ 0 ),
t = constant,
in which energy and momentum of the ﬂuid source are distributed as ﬁnite quantities per unit area in complete analogy to surface charge and surface current in electrodynamics. Hence, we can follow the treatment of surface layers in electrodynamics and derive junction conditions across the twodimensional surface by integrating the ﬁeld equations (1.34) over a ‘pill box’that is centred on the surface and applying Gauss’theorem. These junction conditions combined with the obvious reﬂectional symmetry of the metric coefﬁcients in (1.5), U ( , ζ ) = U ( , −ζ ),
a( , ζ ) = a( , −ζ ),
k( , ζ ) = k( , −ζ ),
W ( , ζ ) = W ( , −ζ ),
(1.88)
or, alternatively, in (1.6), will form the desired boundary conditions on the disc. In the limit ζ → 0, Equations (1.88) ensure the continuity of the metric across ζ = 0 (including the surface layer), {U , a, k, W }ζ =0+ = {U , a, k, W }ζ =0− = {U ( , 0), a( , 0), k( , 0), W ( , 0)} , (1.89) and imply a jump in its normal derivative, U,ζ , a,ζ , k,ζ , W,ζ ζ =0+ = − U,ζ , a,ζ , k,ζ , W,ζ ζ =0− ,
(1.90)
where ζ = 0± means ‘ζ → 0 from above (ζ > 0)’ and ‘ζ → 0 from below (ζ < 0)’. Note that Equations (1.88), (1.89) and (1.90) hold for the corotating potentials {U , a , k , W } too. Before inspecting the ﬁeld equations (1.34), we have to be aware of the behaviour of p and at the disclike surface layer. We assume ﬁnite on 2 p= (1.91) = σ0 ( ) δ(ζ ), 0 outside 2 ; where δ(ζ ) is the Dirac delta distribution. To motivate the ﬁrst assumption, let us consider a (geometrical) transition from an oblate spheroidal ﬂuid body (e.g. a Maclaurin ellipsoid) to a disclike surface layer (‘Maclaurin disc’). During all steps of the ﬂattening process, the central (maximum) pressure should remain ﬁnite and the pressure retain its value, zero, on the body’s surface. Consequently, any volume integral over p has to vanish in the disc limit (‘set of measure zero’). Obviously, the volume energydensity becomes δinﬁnite when related to surface elements of 2 .
1.7 Limiting cases
21
Taking into account Equation (1.91), the ‘pillbox integration’ of the ﬁeld equation (1.34c) results in the junction condition
W,ζ ζ =0+ − W,ζ ζ = 0− = 0. (1.92) Using Equation (1.90) we arrive at the boundary condition
2 : W,ζ ζ = 0± = 0
(1.93)
for the vacuum equation (1.36) W,
+ W,ζ ζ = 0. Consider its reformulation (W − ),
+ (W − ),ζ ζ = 0
(1.94)
under the boundary conditions (1.93), (1.10) and (1.12), i.e.
(W − ),ζ ζ = 0± = 0 at 2 , (W − ) → 0
as → 0
and also as
2 + ζ 2 → ∞.
(1.95)
Obviously, the only regular solution W − = W ( , ζ )− to this ‘mixed’ boundary value problem of the twodimensional Laplace equation (1.94) is W − =0
(1.96)
for all values of ≥ 0 and ζ , i.e. we are automatically led to ‘canonical Weyl coordinates’, see Equation (1.37). From now on we set W = . Since Equation (1.34b) has the form of a vanishing divergence, ∇ · ( −2 e4U ∇a ) = 0, we obtain, after a ‘pillbox integration’ and using (1.89) and (1.90),
(1.97) ζ = 0 : a,ζ ζ = 0 = 0 in all points of the disc ‘plane’ ζ = 0 including the disclike surface layer. In analogy to (1.41), we may introduce a function b via
a , = e−4U b ,ζ ,
a ,ζ = − e−4U b ,
(1.98)
satisfying
( e−4U b , ), + ( e−4U b ,ζ ),ζ = 0
(1.99)
22
Rotating ﬂuid bodies in equilibrium
inside and outside the disc. Using (1.98) we get from (1.97)
ζ = 0 : b, = 0 ⇒ b ( , 0) = constant, ζ =0
(1.100)
where the arbitrary constant of integration can be put equal to zero, b ( , 0) = 0.
(1.101)
This condition holds at all points of the disc ‘plane’ ζ = 0, t = constant inside and outside the disc layer. Finally, we obtain from (1.34a)
1
1 2 : σ ≡ σ0 e2k−2U = U,ζ ζ =0+ − U,ζ ζ =0− = U,ζ ζ =0+ (1.102) 4π 2π as a result of the ‘pillbox integration’, exploiting properties (1.91) of the matter distribution and the symmetry relations (1.89) and (1.90). Note that we have used k − U = k − U , see (1.15c). The surface energydensity σ0 , introduced in (1.91), as well as σ as deﬁned in (1.102), depend on the choice of coordinates. An invariant (‘proper’) surface energydensity σp can be deﬁned by
1 1 i U,i N + = U,ζ ζ =0+ eU −k , (1.103) 2 : σp = ζ =0 2π 2π where N i = (ζ,k ζ ,k )−1/2 ζ,i is the unit normal vector of the timelike hypersurface ζ = 0. Note that N i = eU −k δζi in the coordinates used here. Thus we can rewrite the volume energydensity in (1.91) to read = σp eU −k δ(ζ ).
(1.104)
To complete the boundary conditions, we recall that U = V (1.20) has to be constant on the surface of every ﬂuid body, V = V0 , see (1.27). Therefore, we have to prescribe 2 :
U = V0 ,
(1.105)
i.e. U must be a constant in the disc layer. Because of the speciﬁc form of the boundary values (1.101) and (1.105), we choose the Ernst form of the vacuum equations (1.43) and describe the disc limit of perfect ﬂuids by the boundary value problem
f ∇ 2 f = (∇f )2
(1.106)
with ζ = 0,
≤ 0 :
f ( , 0) = e2V0 .
(1.107)
1.7 Limiting cases
23
Clearly, a regular solution of the boundary value problem has to satisfy the conditions (1.12) at spatial inﬁnity. In terms of the Ernst potential f , they take the simple form f → 1 as 2 + ζ 2 → ∞.
(1.108)
For an illustration of the boundary values see Fig. 2.3. The mixture of primed and unprimed boundary values will lead to our using the ‘corotating’ Ernst equation
f ∇ 2 f = (∇f )2
(1.109)
in the analysis of the boundary value problem as well. To complete the formulae for the metric coefﬁcients in (1.5), we go back to Section 1.4. Combining the vacuum relations (1.39) and (1.40) we have a, = e−4U b,ζ , a,ζ = − e−4U b, , 1 −4U 2 2 2 2 [(b, ) − (b,ζ ) ] , k, = (U, ) − (U,ζ ) + e 4 1 −4U b, b,ζ . k,ζ = 2 U, U,ζ + e 4
(1.110)
Thus we can compute a( , ζ ) as well as k( , ζ ) from the Ernst function f ( , ζ ) via a pathindependent line integration through the vacuum region, starting, say, at the axis of symmetry with the values a = 0 and k = 0 and ending at any point ≥ 0, ζ including the points of the disc. One expects that the reﬂectional symmetry will simplify the discussion of the boundary value problem. In terms of the Ernst functions f = e2U + ib and f = e2U + ib , we have f ( , −ζ ) = f ( , ζ ),
f ( , −ζ ) = f ( , ζ ),
(1.111)
whence ζ =0:
f¯,ζ ζ =0+ = − f,ζ ζ =0− ,
f¯ ,ζ ζ =0+ = − f,ζ ζ =0− .
(1.112)
These relations can easily be derived from the corresponding relations (1.88)–(1.90) and the ‘deﬁnition’ of b and b in (1.41) and (1.98) respectively. Note that the derivation implies a suitable choice of the integration constants, see (1.100), (1.101). The previous analysis of the boundary value problem has shown that the metric of the disc solution must be continuous everywhere (even across the disc). However, the jumps in the ﬁrst derivatives of the metric coefﬁcients U , a, k across the disc, see (1.90), require a careful ‘interpretation’ of the state variables of the disc. We
24
Rotating ﬂuid bodies in equilibrium
have already discussed the interrelation between the jump in the derivative U,ζ of the corotating ‘generalized Newtonian potential’ U = V [which is a function of U and a, see (1.15a)], the δlike distribution of the energydensity and the ‘smooth’ behaviour of the pressure p, see Equations (1.91), (1.102)–(1.104). As a consequence, we may calculate the gravitational mass M and the total angular momentum J of the disc via (1.57) from the energymomentum tensor Tik = σp eU −k δ(ζ )ui uk ,
(1.113)
where we have omitted the pressure term of (1.21), since a ﬁnite pressure cannot contribute to volume integrals over a surface layer. One has to keep this in mind when denoting (1.113) as the ‘energymomentum tensor of the disc’. The 4velocity ui of the disc matter is well deﬁned by (1.18), since its components ui ξi and ui ηi can be expressed in terms of the continuous metric coefﬁcients U and a, see Equations (1.8), (1.20) and (1.15a). The deﬁnition of the 4acceleration is based on the existence of the Christoffel symbols (ﬁrst derivatives of the metric), which are not deﬁned across the disc. A more general approach could start with the expression v i := e−V (ξ i + ηi ),
v i vi = −1,
= constant,
(1.114)
which is well deﬁned outside the disc. Interpreting (1.114) as the 4velocity ﬁeld of a cloud of particles (‘observers’), we get for the 4acceleration v˙i ≡ vi;k v k = V,i .
(1.115)
Obviously, v i and ui coincide along the surface layer. The same holds for the
components of the accelerations v˙ and u˙ , which, according to (1.105), vanish along the surface layer, 2 :
u˙ = v˙ = V, = 0.
(1.116)
Interpreting the ζ component of the 4acceleration u˙ ζ as the ‘mean value’ of the components v˙ζ ‘from above’ and ‘from below’, 2 :
and making use of
U,ζ ζ = 0+
1
v˙ζ ζ =0+ + v˙ζ ζ =0− 2
= − U,ζ , we may assert −
u˙ ζ :=
(1.117)
ζ=0
2 :
u˙ i = 0,
(1.118)
1.7 Limiting cases
25
i.e. the motion of the surface energy elements of the layer is geodesic. Since geodesic motion is a characteristic property of dust, we identify, in a ﬁnal modelforming step, the massenergy density with the baryonic massdensity µB , µB = = σp eU −k δ(ζ ),
(1.119)
i.e. we interpret the disc limit as a dust limit (1.56), thus arriving at a disc of dust model, formally characterized by an energymomentum tensor (1.113), (1.18) and the local energymomentum balance T ik ;k = 0, implying, again in a formal way, geodesic motion (1.54) and local baryonic mass conservation (1.55). Despite µB = , the baryonic mass M0 as calculated from (1.58) differs from the energymass M in Equation (1.57), of course. 1.7.4 The massshedding limit Our intuition tells us that if a star is rotating too quickly, its gravitational pull will no longer sufﬁce to hold it together. If it is on the verge of losing mass, it is said to be rotating at the massshedding limit. The shedding of mass ﬁrst sets in at the equator17 and in Newtonian theory, this limit can be described by stipulating that the pressure gradient (more precisely, ∇p/µ) vanish there, thus implying that the gravitational force balances the centrifugal force in a corotating reference frame. In Einsteinian theory, ∇p/( +p) → 0 implies that a ﬂuid element follows a geodesic. Focusing for the moment on Newtonian theory, we can consider the function 1 V = U − 2 2 , 2 which is constant along the surface of the body: V = V0
along the boundary,
(1.120)
(1.121)
see Subsection 1.7.1. We describe the surface by the parameterization ζ = ζb ( ), restricting ourselves to the halfspace ζ ≥ 0. Taking the derivative of (1.120) with respect to along the surface of the body yields 0=
∂U ∂U dζb + − 2 . ∂
∂ζ d
(1.122)
Reﬂectional symmetry implies that
∂U
= 0, ∂ζ ζ = 0
(1.123)
17 In Newtonian theory, it can be proved that ﬁgures of equilibrium are always reﬂectionally symmetric with
respect to the plane ζ = 0 (Lichtenstein 1933). In Einsteinian theory, the same symmetry is to be expected, see Lindblom (1992), and we assume it to exist for this discussion.
26
Rotating ﬂuid bodies in equilibrium
from which it follows that at the equator ∂U = 2
∂
if
dζb
d < ∞.
(1.124)
This equation tells us that the force due to gravity is equal in magnitude and opposite in direction to the centrifugal force. The inequality in (1.124) tells us that the surface does not meet the equatorial plane at a right angle. In other words, a cusp in the equatorial plane necessarily implies that the star is rotating at the massshedding limit. It can easily be veriﬁed that the same is true in Einsteinian theory. Moreover, numerical results suggest both in Newtonian and Einsteinian gravity that a cusp is a necessary and sufﬁcient condition for the existence of a massshedding limit. The potentials and surface function describing a massshedding star are not analytic, which makes a highly accurate description of them particularly challenging. For homogeneous Newtonian stars, it turns out that these functions are not even C 2 . The proof of this as well as a more general discussion of differentiability can be found in Appendix A1.1. Despite the aforementioned challenges, the extremely simple Roche model for massshedding stars is very accurate in certain cases as is shown in Appendix A1.2. 1.8 Transition to black holes 1.8.1 Horizons The event horizon of a stationary and axisymmetric black hole is given by a hypersurface H whose normal vector χ i is a linear combination of the two Killing vectors ξ i and ηi , χ i ≡ ξ i + h η i ,
h = constant,
(1.125)
and becomes null (lightlike) on that hypersurface: H:
χ i χi = 0,
(1.126)
i.e. the horizon is a null hypersurface to which a Killing vector ﬁeld is normal (a Killing horizon). h is called the ‘angular velocity of the horizon’, see Hawking and Ellis (1973) and Carter (1973). For a recent review of the status of the rigorous mathematical theory of stationary black holes, we refer the reader to Beig and Chrus´ciel (2006). Because of the symmetries of the spacetime (and the horizon), each of the Killing vectors ξ i and ηi must be tangential to the horizon, and therefore H:
χ i ξi = 0,
χ i ηi = 0.
(1.127)
1.8 Transition to black holes
27
For the metric in the form ds2 = e2α (d 2 + dζ 2 ) + W 2 e−2ν (dϕ − ω dt)2 − e2ν dt 2 ,
(1.128)
the relations (1.125)–(1.127) lead to the following boundary conditions on the horizon, see Bardeen (1973a): H:
W = 0,
e2ν = 0,
ω = h .
(1.129)
[Note that W 2 = (ξ i ηi )2 − ξ i ξi ηk ηk = (χ i ηi )2 − χ i χi ηk ηk .] An immediate consequence of the condition W = 0 is the fact that in canonical Weyl coordinates, where W ≡ , the horizon is a part of the ζ axis. This part must not be confused with the other parts of the ζ axis, where W = 0 holds because the Killing vector η vanishes, deﬁning the axis of symmetry. 1.8.2 Kerr black holes Kerr black holes are the only stationary and axisymmetric, isolated black holes surrounded by a vacuum. This follows from the black hole uniqueness theorems, see Robinson (1975), Heusler (1996) and references therein, and Section 2.4, where the Kerr solution will be constructed as the unique solution to the black hole boundary value problem. The Kerr metric (Kerr 1963) in Boyer–Lindquist coordinates (Boyer and Lindquist 1967) is given by 2 dr 2 2 ds = + dϑ + e−2ν sin2 ϑ (dϕ − ω dt)2 − e2ν dt 2 (1.130) with = r 2 + (J /M )2 cos2 ϑ,
= r 2 − 2Mr + (J /M )2 ,
2 r 2 + (J /M )2 − (J /M )2 sin2 ϑ
e2ν =
and
ω=
2Jr 2ν e .
(1.131)
(1.132)
The Boyer – Lindquist coordinates r, ϑ are related to the canonical Weyl coordinates
, ζ by
= r 2 − 2Mr + (J /M )2 sin ϑ, ζ = (r − M ) cos ϑ. (1.133) The Kerr solution depends on two parameters, the gravitational mass M and the angular momentum J (in this section we assume J ≥ 0 without loss of generality).
28
Rotating ﬂuid bodies in equilibrium
The horizon of the black hole is given by r = r+ ≡ M + M 2 − (J /M )2 ,
(1.134)
the larger root of the quadratic equation = 0. Note that the Kerr metric describes a black hole only if J ≤ M2
(1.135)
is satisﬁed. The angular velocity of the horizon introduced in the previous subsection is given by h = ω(r+ , ϑ) = constant =
2M 2 M +
J M 2 − (J /M )2
,
(1.136)
cf. (1.129). The boundary of the ergosphere, see Subsection 1.6.2, is characterized by (1.137) r = r0 (ϑ) ≡ M + M 2 − (J /M )2 cos2 ϑ. Within the ergosphere (r+ < r < r0 ), all observers must rotate in the same direction as the black hole (dϕ/dt > 0), cf. (1.68). It is interesting to discuss circular orbits of test particles in the equatorial ‘plane’ ϑ = π/2. Their angular velocity is given by √ M =± (1.138) √ , 3/2 r ± J/ M where the upper sign characterizes direct orbits (corotating with the black hole) and the lower sign holds for retrograde (counterrotating) orbits. The circular orbits exist only for r > rph , with the ‘photon orbit’
J 2 rph = 2M 1 + cos arccos ∓ 2 . (1.139) 3 M The orbits are bound for r > rmb , with the ‘marginally bound orbit’ rmb = 2M ∓ J /M + 2M 1/2 (M ∓ J /M )1/2 .
(1.140)
A particle in an unbound orbit will escape to inﬁnity under the inﬂuence of an inﬁnitesimal outward perturbation. The orbits are stable for r > rms , with the ‘marginally stable orbit’ rms = M 3 + Z2 ∓ [(3 − Z1 )(3 + Z1 + 2Z2 )]1/2 , (1.141)
1.8 Transition to black holes
29
Table 1.1. Photon orbit, marginally bound orbit and marginally stable orbit for the Schwarzschild black hole and for the extreme Kerr black hole. In the latter case one has to distinguish between direct and retrograde orbits.
J =0 J = M 2 (direct) J = M 2 (retrograde)
rph
rmb
rms
3M M 4M
4M √ M (3 + 2 2)M
6M M 9M
where
J2 Z1 = 1 + 1 − 4 M
1/3
J 1+ 2 M
1/3
J + 1− 2 M
1/3 (1.142)
and 1/2 2 J Z2 = 3 4 + Z12 . M
(1.143)
These results on circular orbits of test particles were derived by Bardeen et al. (1972), see also Shapiro and Teukolsky (1983). The two limiting cases of the Kerr black hole are J = 0, the nonrotating (Schwarzschild) black hole, and J = M 2 , the maximally rotating (extreme Kerr) black hole. For J = 0, the horizon is given by r+ = 2M (‘Schwarzschild radius’) and no ergosphere exists. For J = M 2 one has r+ = M and the ergosphere, as with all J > 0, extends up to r0 (π/2) = 2M in the equatorial plane. The values of the characteristic radii rph , rmb and rms discussed above are given in (d) (d) (d) Table 1.1. In the extreme case, rph , rmb and rms , where the (d) indicates direct orbits, all coincide with r+ = M . However, deﬁning proper radial distances (ϑ = constant, ϕ = constant, t = constant) according to r2 δ(r1 , r2 ) =
√ grr dr,
grr =
r1
,
(1.144)
one ﬁnds for ϑ = π/2 (Bardeen et al. 1972) (d)
lim δ(r+ , rph ) =
J →M 2
M ln 3, 2
(d)
lim δ(r+ , rmb ) = M ln(1 +
J →M 2
√ 2)
(1.145)
30
Rotating ﬂuid bodies in equilibrium
and (d) ) = ∞. lim δ(r+ , rms
(1.146)
J →M 2
Moreover, because of the double zero of at r = M for the extreme Kerr metric, one obtains δ(M , r) = ∞
for any r > M and all ϑ.
(1.147)
This means that the horizon (together with the whole ‘r = M region’) has an inﬁnite proper radial distance from any point in the ‘exterior’ region r > M . On the other hand, the proper time needed for an infalling particle starting at some r > M to reach the horizon remains ﬁnite. The geometrical situation can be illustrated nicely by embedding the r ≥ r+ part of the ‘plane’ ϑ = π/2, t = constant into threedimensional Euclidean space (Bardeen et al. 1972), see Fig. 1.2. In the limit J = M 2 , an inﬁnitely long ‘throat’ characterized by r = M (circumference: 4πM ) appears. The horizon is situated at the bottom and the direct orbits corresponding to (d) (d) (d) rph , rmb and rms are located at different places along the throat. However, the proper time of an infalling particle needed to pass through this throat is zero. Bardeen and Horowitz (1999) have studied the ‘throat geometry’ (r = M ) by means of the coordinate transformation r˜ =
r−M , λ
ϑ˜ = ϑ,
ϕ˜ = ϕ − h t,
˜t = λt
(1.148)
Fig. 1.2. Embedding diagrams of the r ≥ r+ part of the ‘plane’ ϑ = π/2, t = constant of the Kerr metric as it approaches the limit J = M 2 . This 2surface is represented here as a surface of revolution in threedimensional Euclidean space with the same interior geometry. The positions of the direct orbits rph , rmb , rms and of the boundary r0 (π/2) of the ergosphere are shown. Note that the two gaps in the fourth picture are both inﬁnitely long (adapted from Bardeen et al. 1972).
1.8 Transition to black holes
31
in the limit λ → 0. According to (1.136), h is given here by h =
1 2M
(J = M 2 ).
(1.149)
One obtains ˜ ds = M (1 + cos ϑ) 2
2
2
d˜r 2 r˜ 2 d˜t 2 ˜2 − + d ϑ r˜ 2 4M 4
4M 2 sin2 ϑ˜ + 1 + cos2 ϑ˜
2 r˜ d˜t dϕ˜ + . 2M 2 (1.150)
This represents a completely nonsingular vacuum solution of Einstein’s equations, which is geodesically complete but no longer asymptotically ﬂat, see Bardeen and Horowitz (1999). The area of all surfaces r˜ = constant, ˜t = constant is 8πM 2 and equal to the area of the horizon of the extreme Kerr black hole. In addition to ∂/∂ ˜t and ∂/∂ ϕ, ˜ the metric (1.150) has two more Killing ﬁelds (Wolf 1998): ˜t 2 ∂ ∂ ∂ 1 1 ∂ ∂ − r˜ ˜t − + , ˜t − r˜ . (1.151) 4 2 2 ∂ r˜ 2˜r h ∂ ϕ˜ ∂ r˜ ∂ ˜t 8˜r 2 h ∂ ˜t 1.8.3 From rotating ﬂuids to black holes The condition characterizing a black hole horizon, (ξ i + h ηi )(ξi + h ηi ) = 0,
(1.152)
looks similar to the surface condition of a rotating ﬂuid body, (ξ i + ηi )(ξi + ηi ) = −e2V0 .
(1.153)
An interesting question is whether or not there exist continuous parametric transitions from stationary ﬂuid conﬁgurations to black holes. We now want to show that V0 → −∞
(1.154)
M − 2J → 0
(1.155)
or, equivalently,
is necessary and sufﬁcient for a black hole limit of a ﬂuid body in equilibrium (Meinel 2004, 2006). A remarkable consequence of (1.155) is the fact that such a limit always leads to the extreme Kerr black hole.
32
Rotating ﬂuid bodies in equilibrium
In a ﬁrst step, we prove the equivalence of (1.154) and (1.155). To this end, we combine formulae (1.57) and (1.58) for the gravitational mass M , the angular momentum J and the baryonic mass M0 , leading to + 3p V e dM0 , (1.156) M = 2J + µB where we have used Equations (1.18), (1.21) and the abbreviation dM0 = −µB ui ni dV for the baryonic mass elements. With (1.23) and (1.26), we get + 3p V0 dM0 . (1.157) M = 2J + h(0) e +p Since, for nonnegative and p, 1 ≤ ( +3p)/( +p) ≤ 3 holds, conditions (1.154) and (1.155) are equivalent.18 Next we note that all linear combinations of the two Killing vectors ξ i and ηi that are not proportional to the null (lightlike) vector ξ i + h ηi become spacelike on the horizon. However, the 4velocity (1.18) must always be timelike. Therefore, a black hole limit of a ﬂuid body can only be approached for → h
(1.158)
together with V0 → −∞, i.e. (1.154) is necessary. Finally, we prove that (1.154) is also sufﬁcient for a black hole limit. Because of (1.19) and (1.26), the surface of the ﬂuid is characterized in general by χ i χi = −e2V0 ,
χ i ≡ ξ i + ηi .
(1.159)
The Killing vector χ i is tangential to the hypersurface H generated by the timelike world lines of the ﬂuid elements of the surface of the body with 4velocity ui = e−V0 χ i , see (1.18). Each of the Killing vectors ξ i and ηi must itself be tangential to H because of the symmetries of the spacetime. In the limit V0 → −∞, we approach a situation in which χ i becomes null on H: χ i χi → 0.
(1.160)
Moreover, with the reasonable assumption19 0 ≤ −ξ i ui ≤ 1,
(1.161)
18 We assume 0 < M < ∞ and 0 < h(0) < ∞. 0 19 The condition −ξ i u ≤ 1 ensures that a particle resting on the surface of the ﬂuid is (at least marginally) bound, i i.e. cannot escape to inﬁnity on a geodesic; −ξ i ui ≥ 0 follows from −ξ i ui = −(χ i − ηi )ui = eV0 + ηi ui , since ηi ui will always have the same sign as (ηi ui = 0 on the axis, of course).
1.8 Transition to black holes
33
we ﬁnd that χ i also becomes orthogonal to ξ i (and thus to ηi ) on H in the limit: χ i ξi → 0,
χ i ηi → 0.
(1.162)
Together with the orthogonal transitivity of the spacetime (see Section 1.3), χ i therefore becomes orthogonal to three linearly independent tangent vectors at each point of H, i.e. normal to H. Because of (1.160), we thus approach a situation in which H is a null hypersurface and satisﬁes all deﬁning conditions for a horizon of a stationary and axisymmetric black hole with being the angular velocity of the horizon, see Subsection 1.8.1. According to the black hole uniqueness theorems, we conclude that (outside the horizon) the Kerr metric with J ≤ M 2 results.20 Then, with (1.155), (1.158) and (1.136), we are necessarily led to the case J = M 2 . Therefore, the metric of an extreme Kerr black hole (outside the horizon) results, whenever a sequence of ﬂuid bodies admits a limit V0 → −∞. Later in this book, we shall demonstrate with analytical as well as numerical examples that such parametric (‘quasistationary’) transitions from ﬂuid bodies to black holes are indeed possible. It will turn out that an interesting ‘separation of spacetimes’ with a nonasymptotically ﬂat ‘inner world’ (containing the ﬂuid body and its surroundings) and the r > M part of the extreme Kerr metric as the ‘outer world’ emerges in the limit V0 → −∞. The behaviour of the ‘inner world’ at spatial inﬁnity is given precisely by the ‘throat geometry’ (1.150). 20 Note that the black hole uniqueness proof by construction given in Neugebauer and Meinel (2003) can be extended to the case in which the horizon is degenerate, leading to J = M 2 . This will be shown in Section 2.4.
2 Analytical treatment of limiting cases
A particular difﬁculty of the rotating ﬂuid body problem – in Newton’s as well as in Einstein’s theory – is the free boundary of the ﬂuid, which is not known from the outset. However, in some special cases, the shape of the surface is known (i) by making a lucky guess, (ii) from symmetry considerations leading to spheres in the nonrotating case, and (iii) by considering the limit of extreme ﬂattening leading to inﬁnitesimally thin discs. 2.1 Maclaurin spheroids An important example for rotating ﬁgures of equilibrium within Newton’s theory of gravity – the Maclaurin spheroids – was indeed found by a lucky guess for the surface’s shape. It turns out that homogeneous (constant massdensity) spheroids are solutions to the problem for a certain constant angular velocity, given prescribed values of the mass density and the semiaxes of the spheroid. The Newtonian potential can of course be calculated from the Poisson integral. For ellipsoidal conﬁgurations the choice of elliptic coordinates proves useful. The exterior potential of a homogeneous oblate spheroid of mass M and focal length 0 is given by M 3 1 U =− ξ − ξ2 + arccot ξ (1 − 3η2 ) , arccot ξ + (2.1) 0 4 3 with oblate elliptic coordinates ξ , η (and ϕ) related to the cylindrical coordinates , ζ (and ϕ) by = 0 (1 + ξ 2 )(1 − η2 ), ζ = 0 ξ η (0 ≤ ξ < ∞, −1 ≤ η ≤ 1), (2.2) see Fig. 2.1. U can be written in the form U = A(ξ ) + B(ξ )2 , 34
(2.3)
2.1 Maclaurin spheroids
35
Fig. 2.1. A depiction of the oblate elliptic coordinates of Equation (2.2). Lines of constant ξ are ellipses and lines of constant η are hyperbolas all with the focus (0 , 0).
with
3M 2 A(ξ ) = − (ξ + 1)arccot ξ − ξ 20
and B(ξ ) =
3M (3ξ 2 + 1)arccot ξ − 3ξ 403 (1 + ξ 2 )
.
(2.4)
(2.5)
Let the surface of the spheroid be characterized by ξ = ξ0 . Then the surface condition 1 V ≡ U − 2 2 = V0 = constant (ξ = ξ0 ) (2.6) 2 is satisﬁed for (2.7) = 2B(ξ0 ), V0 = A(ξ0 ). Note that the semiaxes (a = b > c) of the ellipsoid are a = 0 1 + ξ02 , c = 0 ξ0 .
(2.8)
36
Analytical treatment of limiting cases
The mass is related to the constant massdensity µ by M =
4π 2 µa c. 3
(2.9)
The interior solution (ξ < ξ0 ) is given by 1 2 2 2 ζ 2 U = V0 + − C 1 − 2 − 2 , 2 a c
(2.10)
with
3M ξ0 (1 − ξ0 arccot ξ0 ). 20 The pressure distribution follows from the integrated Euler equation: 2 ζ 2 p = µ(V0 − V ) = µC 1 − 2 − 2 . a c C=
(2.11)
(2.12)
It can easily be veriﬁed that expressions (2.1) and (2.10) satisfy the Laplace and Poisson equations ∇ 2 U = 0 and ∇ 2 U = 4πµ respectively. The continuity of the potential at ξ = ξ0 is obvious; the continuity of its normal derivative can easily be checked as well. For a given value of the massdensity µ (i.e. for a given equation of state) the complete solution depends on two parameters, for example a and c or 0 and ξ0 . It is, of course, regular everywhere and has the correct asymptotic behaviour U →−
M r
as
r → ∞ (r 2 ≡ 2 + ζ 2 ).
(2.13)
In the following, we present some further useful parameter relations. The angular velocity given by (2.7) can also be related to the massdensity and the eccentricity
c2 1 (2.14) = 1− 2 = a 1 + ξ2 0
leading to the celebrated formula (Maclaurin 1742) 2 2 6 = 3 1 − 2 (3 − 2 2 ) arcsin − 2 (1 − 2 ), πµ
(2.15)
a plot of which can be found in Fig. 2.2. The angular momentum J and kinetic energy of rotation Erot are given by J = I ,
1 Erot = I 2 , 2
(2.16)
2.1 Maclaurin spheroids
37
Fig. 2.2. The normalized, squared angular velocity 2 /π µ for the Maclaurin sequence as it depends on eccentricity , see Equation (2.15). For a given massdensity µ and for some < max , there exist solutions for two different values of .
with the moment of inertia about the axis of rotation 2 I = Ma2 . 5 The gravitational (potential) energy is 1 3M 2 arcsin U µ d3 x = − . Eg ≡ 2 5a
(2.17)
(2.18)
It is interesting to discuss the two limiting cases of the Maclaurin spheroids: (i) Maclaurin spheres For → 0 we get a (nonrotating) sphere of radius R ≡ a = c with −2πµ(R2 − r 2 /3): r < R U = −M /r: r > R
(2.19)
and
2πµ2 2 (R − r 2 ) (r ≤ R). 3 The gravitational energy of a homogeneous sphere is p=
Eg = −
3M 2 . 5R
(2.20)
(2.21)
(ii) Maclaurin discs For → 1 we get a disc with radius 0 . The potential U is given by (2.1) everywhere. On the disc itself (characterized by ξ = 0 or, equivalently, ζ = 0
38
Analytical treatment of limiting cases
and ≤ 0 )
1 U = V0 + 2 2 , (2.22) 2 i.e. V = V0 = constant, results. Equation (2.9) shows that µ → ∞ in the disc limit.1 The corresponding surface massdensity σ is a function of : 3M 02 − 2 . (2.23) σ = 2π03 It can be calculated either by simple geometric means or via the jump in the normal derivative of U : U,ζ ζ =0+ − U,ζ ζ =0− = 2U,ζ ζ =0+ = 4πσ . Note that p/µ → 0 holds in the disc limit, since the central (maximal) pressure remains ﬁnite. Accordingly, the solution may also be interpreted as a ‘rigidly rotating disc of dust’, see Subsection 1.7.3. This corresponds to the fact that for each (surface) mass element the gravitational force and the centrifugal force are in equilibrium (∂U /∂ = 2 ). Angular velocity, mass and radius are related according to 2 =
3π M 403
(2.24)
and the relation V0 = −2 02
(2.25)
holds. The angular momentum is J =
8 3 5 0 15π
(2.26)
and the kinetic and gravitational energies are given by Erot =
3πM 2 200
and
Eg = −
3πM 2 . 100
(2.27)
Remarkably, analytic solutions of the Einstein equations exist for both limiting cases (spheres and discs) as well. These will be treated in the next two sections. 2.2 Schwarzschild spheres The global solution of Einstein’s ﬁeld equations describing a perfect ﬂuid sphere of constant mass(energy) density µ was found by Schwarzschild (1916). In our coordinates, the line element is ds2 = e2α (d2 + dζ 2 + 2 dϕ 2 ) − e2ν dt 2 , 1 We assume ﬁnite values for the disc’s radius and mass, of course.
(2.28)
2.2 Schwarzschild spheres
where α and ν depend on r=
39
2 + ζ 2
(2.29)
only. We denote the coordinate radius of the sphere by r0 . Inside the ﬂuid (r < r0 ) we have the ‘interior Schwarzschild solution’ given by 3 Mr 2 M M 1 − 1 + 2r 1 1 − 2r0 2r03 0 − (2.30) eα = , eν = 3 2 2 M 2 1 + 2r0 1 + Mr3 1 + Mr3 2r0
and
2r0
p = µ(eν0 −ν − 1),
eν0 ≡ eν(r0 ) .
(2.31)
The exterior solution (r > r0 ) is naturally the vacuum Schwarzschild metric given by 1− M M 2 α 2r e = 1+ , eν = . (2.32) 2r 1+ M 2r Of course, in the (‘isotropic’) coordinates used here, the metric coefﬁcients and their ﬁrst derivatives are continuous at the surface of the ﬂuid.2 M is the gravitational mass. For a given value of the massdensity µ, the solution depends on only one parameter, say ν0 . M and r0 are related to ν0 according to
3/2 3 1 1 − e2ν0 , (2.33) M = 4 2πµ 1 − eν0 M = . 2r0 1 + eν0
(2.34)
It turns out that
1 (2.35) 3 must hold. For eν0 → 1/3, the central pressure tends to inﬁnity.3 As a consequence, the relative redshift z = e−ν0 − 1 (2.36) eν0 >
of photons emitted from the ﬂuid’s surface and received at inﬁnity is bounded by the value 2. Even more importantly, the gravitational mass is also bounded: 4 M < Mmax = √ . 9 3πµ
(2.37)
2 The more familiar radial Schwarzschild coordinate r˜ is related to our r by r˜ = r(1+M /2r)2 in the vacuum region and r˜ = 2Br/(B2 + Ar 2 ) with A = 8π µ/3, B = 14 (1 + eν0 )3 inside the ﬂuid. In terms of the Schwarzschild 3 coordinate radius r˜0 , the gravitational mass is given by the simple expression M = 4π 3 µ˜r0 . Note that ∂gr˜ r˜ /∂ r˜ is discontinuous at r˜ = r˜0 . 3 In this limit, the Schwarzschild coordinate radius r˜ reaches the ‘Buchdahl limit’ 9M /4, see Subsection 1.7.2. 0
40
Analytical treatment of limiting cases
The baryonic mass (calculated under the assumption µ = µB , cf. Section 1.5) is
3 3 ν0 ν0 2ν 0 M0 = arccos e − e 1−e . (2.38) 8 2πµ The relative binding energy (M0 − M )/M0 is positive and reaches a maximal value of about 0.39 in the limit eν0 → 1/3. The Newtonian limit of the Schwarzschild spheres is approached for γ ≡ 1 − eν0 1 and leads back to the Maclaurin spheres. Of course, M /M0 → 1 as γ → 0. However, it is interesting to note that the leading order term in an expansion of M − M0 in powers of γ 1/2 gives exactly the Newtonian gravitational energy Eg , see (2.21). 2.3 The rigidly rotating disc of dust 2.3.1 The boundary value problem of the Ernst equation As has already been shown in Subsection 1.7.3, the model describing a uniformly rotating, inﬁnitely thin disc leads to a boundary value problem for the Ernst equation, see Fig. 2.3. On the disc, i.e. for ζ = 0 and ≤ 0 , the Ernst potential f in the corotating frame has to be equal to a real constant exp(2V0 ) with V0 < 0, f = e2V0
for
ζ = 0, ≤ 0 ,
(2.39)
see (1.107), whereas the Ernst potential f in the nonrotating frame has to be regular everywhere outside the disc and has to satisfy f →1
as
2 + ζ 2 → ∞,
(2.40)
see (1.108).
Fig. 2.3. The boundary value problem of the Ernst equation for a rigidly rotating disc of dust.
2.3 The rigidly rotating disc of dust
41
At ﬁrst glance, the boundary value problem seems to depend on the three parameters V0 , 0 and , where the angular velocity enters into the problem via the relation between f and f . However, it will turn out that the solution is not regular at the rim of the disc unless a certain relation between the three parameters is satisﬁed.4 Consequently, only two of them can be chosen arbitrarily. In the next subsections we shall derive the solution to this boundary value problem (Neugebauer and Meinel 1993, 1994, 1995, Neugebauer et al. 1996, Neugebauer and Meinel 2003). This will be done by applying the ‘inverse method’,5 which relies on the existence of a ‘linear problem’ for the Ernst equation (Belinski and Zakharov 1978, Harrison 1978, Maison 1978, Hoenselaers et al. 1979, Neugebauer 1979, 1980a). A further application of this method, namely the derivation of the Kerr metric as the unique solution for a stationary and axisymmetric vacuum black hole, including the degenerate case, will then be given in Section 2.4.
2.3.2 Solution via the ‘inverse method’ Neugebauer’s form of the linear problem (LP) for the Ernst equation reads (Neugebauer 1980b, Neugebauer and Kramer 1983) B 0 0 B ,z = +λ , 0 A A 0 (2.41) 1 0 A¯ A¯ 0 ,¯z = , + 0 B¯ λ B¯ 0 where (z, z¯ , λ) is a 2 × 2 matrix depending on the spectral parameter K − i¯z λ= K + iz
(2.42)
as well as on the complex coordinates z = + iζ ,
z¯ = − iζ ,
(2.43)
¯ B¯ are functions of z, z¯ (or , ζ ) and whereas A, B and their complex conjugates A, do not depend on K. From the integrability condition and the identities λ,z =
λ 2 λ −1 , 4
λ,¯z =
1 2 λ −1 , 4λ
(2.44)
4 In the Newtonian limit (V  1) this parameter relation reduces to (2.25). 0 5 For a general introduction to the ‘inverse method’ in the context of soliton theory, see e.g. Novikov et al. (1984).
42
Analytical treatment of limiting cases
it follows that a certain matrix polynomial in λ has to vanish. This yields the set of ﬁrst order differential equations ¯ − 1 (A + B), ¯ A,¯z = A(B¯ − A) 4
¯ ¯ − 1 (B + A). B,¯z = B(A¯ − B) 4
(2.45)
The system has the ‘ﬁrst integrals’ A=
f,z f + f¯
,
B=
f¯,z f + f¯
.
(2.46)
Resubstituting A and B in Equations (2.45), one obtains the Ernst equation 1 ( f ) f, + f,ζ ζ + f, = f,2 + f,ζ2 , (2.47) cf. (1.43). Thus, the Ernst equation is the integrability condition of the LP (2.41). Conversely, if f is a solution to the Ernst equation, the matrix calculated from (2.41) does not depend on the path of integration. The idea of the ‘inverse method’ is to discuss for ﬁxed, but arbitrary, values of z and z¯ as a holomorphic function of λ (or K) and to calculate A, B and ﬁnally f afterwards. To obtain the desired information about the holomorphic structure in λ, we shall integrate the LP along the dotted line in Fig. 2.4. A± are the two parts of the axis of symmetry ( = 0, ζ > 0 or ζ < 0), B represents the surface of the disc ( ≤ 0 , ζ = 0± ) and C stands for spatial inﬁnity. In this way, we shall solve the direct problem of the inverse method and obtain (z, z¯ , λ) for z, z¯ ∈ A± , B , C . It turns out that the holomorphic structure remains unchanged by an extension to values of z and z¯ off the axis of symmetry into the
Fig. 2.4. The line of integration for the direct problem, hugging the disc, along the axis of symmetry and at inﬁnity.
2.3 The rigidly rotating disc of dust
43
Fig. 2.5. The twosheeted Riemann Ksurface with moving branch cut (shown for some arbitrary z belonging to the path A+ CA− B of Fig. 2.4) and the ‘image’ of the disc, i.e. the contour(s) : K = 0, −0 ≤ K ≤ 0 (thick line).
entire vacuum region, meaning that one can construct functions with prescribed properties in λ from which one obtains the desired solution f (z, z¯ ) everywhere in the vacuum region. In some circumstances, λ may be replaced by K. For this purpose, it may be helpful to discuss the mapping (2.42) of the twosheeted Riemann surface of K onto the λplane for different values of , ζ (or equivalently z, z¯ ). Figure 2.5 shows the position of the branch points KB = i¯z , K¯ B = −iz for the marked path A+ CA− B of Fig. 2.4 by two dotted curves. It reﬂects the slice ϕ = constant, t = constant (Fig. 2.4) and indicates, in particular, the position of the disc. Consider now a Riemann surface with conﬂuent branch points KB = K¯ B = ζ ∈ + A . Here λ degenerates and takes the values λ = −1 for K’s in the lower sheet, say, and λ = +1 for K’s in the upper sheet (K = KB ). We shall now travel along the dotted line of Fig. 2.4 starting from and returning to any point = 0, ζ ∈ A+ . In Fig. 2.5 this corresponds to the bold faced points on the real axis. Note that λ = −1 for all K’s (K = ζ ) in the lower and λ = +1 for all K’s (K = ζ ) in the upper sheet of the Riemann Ksurface belonging to axis values = 0, ζ ∈ A± . The corresponding branch points cling to either side of the real axis in Fig. 2.5. For , ζ ∈ C , the cut between the branch points (e.g. the right solid line in Fig. 2.5) sweeps over the entire Ksurface and puts ‘upper’ K values into the lower sheet and ‘lower’ K values into the ‘upper’ sheet. As a consequence, λ will change from ±1 to ∓1 between = 0, ζ = +∞ and = 0, ζ = −∞. This ‘exchange of sheets’ is important for the solution of the LP: The initial value (0 , ζ0 , λ) can (and must) be ﬁxed in only one sheet of the Ksurface. The dependence on K in the other sheet follows by integrating the LP (2.41) along a suitable path.
44
Analytical treatment of limiting cases
We shall divide the integration of the LP (2.41) along the closed dotted line of Fig. 2.4 into two steps: (i) Integrating along A+ CA− . This step leads to a ‘general solution’ for on the regular parts A± of the symmetry axis and will be used in Section 2.4 as well. (ii) Integrating along B.
Axis and inﬁnity It can easily be veriﬁed that the matrix coefﬁcients U and V of the LP ,z = U, ,¯z = V, as given in (2.41), satisfy the relations 1 0 1 0 U(−λ) = U(λ) (2.48) 0 −1 0 −1 and
1 0 1 0 1 = U(λ) . (2.49) U 1 0 1 0 λ¯ Therefore, without loss of generality, the matrix may be assumed to have the structure ψ(, ζ , λ) ψ(, ζ , −λ) = (2.50) χ(, ζ , λ) −χ (, ζ , −λ) together with
1 = χ (, ζ , λ). ψ , ζ , λ¯ The particular form of (2.50) is equivalent to 1 0 0 1 (−λ) = (λ) 0 −1 1 0
and (2.51) can be written as 1 0 = ¯λ 1
1 1 0 (λ) . 0 −1 0
(2.51)
(2.52)
(2.53)
For K → ∞ and λ = −1, the functions ψ, χ may be normalized by ψ(, ζ , −1) = χ (, ζ , −1) = 1.
(2.54)
Finally, the solution to the Ernst equation can be read off at λ = 1 (K → ∞), f (, ζ ) = χ(, ζ , 1), cf. (2.41) for λ = 1 and (2.46).
f (, ζ ) = ψ(, ζ , 1),
(2.55)
2.3 The rigidly rotating disc of dust
45
As discussed in Section 1.4, the Ernst equation retains its form in the corotating frame of reference. This ensures the existence of an LP (2.41) in the corotating system. The matrices of the two systems of reference are connected by the relation 0 1 + a − e−2U = −2U 0 1 + a + e −1 −λ + i(K + i z)e−2U (2.56) . λ 1 As in Section 1.3, a prime marks ‘corotating’ quantities. We are now in a position to integrate the LP (2.41) along the part A+ CA− of the dotted line in Fig. 2.4. Using (2.41) along A± and (2.46) one ﬁnds for the axis values of F(K) 0 f (ζ ) 1 A+ : = , (2.57) G(K) 1 f (ζ ) −1 1 G(K) f (ζ ) 1 A− : = , (2.58) 0 F(K) f (ζ ) −1 where f (ζ ) = f ( = 0, ζ ) is the axis value of the Ernst potential and F(K), G(K) are integration ‘constants’ depending on K alone. The particular form of (2.57) is due to the initial condition ψ = χ = 1 for some 0 = 0, ζ = ζ0 ∈ A+ , λ = −1 (K in the lower sheet), which ﬁxes the second column of in (2.57), cf. (2.50). The ﬁrst column corresponds to the upper sheet (λ = 1) and represents a general integral with the two integration ‘constants’ F(K), G(K) which cannot be speciﬁed here. Along C , = (K) does not depend on and ζ , since A and B vanish because of (2.40) and (2.46). The ‘exchange of sheets’6 along C , together with (2.52), leads to the particular form of on A− , see Meinel and Neugebauer (1995). The representations (2.57), (2.58) describe the behaviour of ψ and χ in both sheets. Nevertheless, one may wish to consider the whole matrix as a unique function of λ, which is therefore deﬁned on both sheets of the Ksurface. From this point of view, Equations (2.57), (2.58) describe on one sheet only (say, on the upper sheet). Its values on the other (lower) sheet follow from (2.52). 6 The path C can be parameterized by = R sin ϑ, ζ = R cos ϑ, 0 ≤ ϑ ≤ π (R → ∞). According to (2.42) this
gives λ = ± exp(i ϑ).
46
Analytical treatment of limiting cases
Combining (2.57) and (2.58) with (2.56), we obtain the axis values in the corotating system, −1 −1 −2U + A : = 1 + i (K − ζ )e 1 1 F(K) 0 f (ζ ) 1 , (2.59) × G(K) 1 f (ζ ) −1 −1 −1 A− : = 1 + i (K − ζ )e−2U 1 1 1 G(K) f (ζ ) 1 , (2.60) × 0 F(K) f (ζ ) −1 where 1 is the 2 × 2 unit matrix. At the branch points KB = ζ of Ksurfaces belonging to axis values = 0, ζ ∈ A± , ψ and χ must be unique, i.e. ψ ψ F(ζ ) 0 f (ζ ) 1 + A (KB = ζ ) : = = , (2.61) χ −χ G(ζ ) 1 f (ζ ) −1 ψ ψ 1 G(ζ ) f (ζ ) 1 A− (KB = ζ ) : = = , (2.62) χ −χ 0 F(ζ ) f (ζ ) −1 whence A+ :
F(ζ ) =
A− :
F(ζ ) =
2 f (ζ ) + f (ζ ) 2f (ζ )f (ζ ) f (ζ ) + f (ζ )
,
G(ζ ) =
,
G(ζ ) =
f (ζ ) − f (ζ ) f (ζ ) + f (ζ ) f (ζ ) − f (ζ ) f (ζ ) + f (ζ )
,
(2.63)
.
(2.64)
Thus, F(K) and G(K) consist in a unique way of analytic continuations of the real and imaginary parts of the axis values of the Ernst potential f (ζ ). Vice versa, f (ζ ) follows from F(K), G(K) for K = ζ . Interestingly, the determinants of and can be expressed in terms of f , f and F(K). From (2.41) [Tr,z −1 = (ln det ),z ], (2.46) and (2.57)–(2.60), we have det = −2e2U F(K),
det = −2e2V F(K),
(2.65)
where e2U = f and e2V = f [U = U (, ζ ), V = V (, ζ )]. We can now interpret the result of (2.59)–(2.64) of the integration of the LP along A+ CA− : On the regular parts A± of the symmetry axis, and can explicitly be expressed in terms of the axis values f (ζ ) of the Ernst potential and its analytic continuations F(K), G(K). To calculate f (ζ ), however, one needs boundary values on B .
2.3 The rigidly rotating disc of dust
47
Surface of the disc On the disc, the LP in the corotating system of reference takes the form 0 f ,ζ B : , = − , 2 2 0 K + ( f + f ) f,ζ
(2.66)
where and f are the ‘corotating’ matrix and the ‘corotating’ Ernst potential on the disc. This relation must be considered alongside the ‘boundary conditions’ ( = 0, ζ = 0+ , λ)
B
= ( = 0, ζ = 0+ , λ)
A+
,
( = 0, ζ = 0− , λ)
B
= ( = 0, ζ = 0− , λ)
A−
,
cf. (2.59) and
cf. (2.60). The analysis of this problem will allow for the construction of F(K) and G(K) and, ultimately, via (2.63) and (2.64), for the construction of the axis values f (ζ ) of the Ernst potential. We ﬁrst take advantage of the symmetry of the problem, which implies f (, ζ ) = f (, −ζ ) and connects the ζ derivatives of f above (ζ = 0+ ) and below (ζ = 0− ) the disc (2.67) B : f ,ζ ζ =0+ = −f,ζ ζ =0− . A
As a consequence, the LP (2.66) connects the matrix above the disc, A
B
B
= (, ζ = 0+ , K), with the matrix below the disc, = (, ζ = 0− , K), B:
A
=
0 1 B H(K), −1 0
(2.68)
where the matrix H(K) (the ‘integration constant’) does not depend on ∈ B . At the rim of the disc we have A
B
r
(0 , 0, K) = (0 , 0, K) = . (2.69) r 0 −1 r −1 Because of (2.68), the rim matrix can be expressed in terms 1 0 A
B
of = (, ζ = 0+ , K), = (, ζ = 0− , K). Note that is considered to be a holomorphic function of λ and therefore a function living on the twosheeted Riemann Ksurface of Fig. 2.5. Hence we have to discuss the rim matrix as a function of K on both sheets.
48
Analytical treatment of limiting cases
Any multiplied from the right by a matrix function of K is again a solution of the LP. The discussion of the rim matrix simpliﬁes upon using the following renormalization −1 −1 0 1 0 1 F 0 r −1 0 −1 r F 0 . (2.70) R= 1 0 G 1 −1 0 −1 0 G 1 Using (2.59) and (2.60) we obtain ! e2V0 MS −1 on the upper sheet R= −e2V0 S −1 M on the lower sheet, where
and
G(K) [G(K)2 − 1]/F(K) , M= −F(K) −G(K) f0 f0 − 42 K 2 i b0 + 2iK S= i b0 − 2iK −1
(2.71)
f0 ≡ e2V0 + i b0 ≡ f ( = 0, ζ = 0+ ).
(2.72)
(2.73)
Obviously, Tr R = Tr R−1 = 0 and M2 = 1, whence Tr MS −1 = Tr SM = 0.
(2.74)
This relation implies one between F(K) and G(K) and, because of (2.63) and (2.64), thus also between the real and imaginary parts of the axis values f (ζ ) of the Ernst potential (Neugebauer and Meinel 1993, 1994). We next wish to determine F(K) and G(K), which in turn determine f (ζ ). To this end, we consider (, ζ , λ) for ﬁxed coordinates , ζ as a function of λ. We have already used the initial conditions ψ = χ = 1 for some = 1 = 0, ζ = ζ1 ∈ A+ prescribed in one sheet (λ = −1) of the Kplane. In principle, the behaviour of in the other sheet and at all points in the , ζ plane can be calculated by integrating the LP along a suitable path. However, the coefﬁcients A(, ζ ), B(, ζ ) in the LP (2.41) are not explicitly known. Nevertheless, their regular behaviour outside the disc together with the boundary values on the disc provides us with deﬁning properties for . One of them may be taken from Fig. 2.5: Since the domain of the disc, 0 ≤ ≤ 0 , ζ = 0± is a nonvacuum domain, where the LP fails, the matrix cannot remain continuous through the contour(s) : −0 ≤ K ≤ 0 at the branch point pairs K = i + 0± , −i + 0± , i.e. has a welldeﬁned jump between opposite points along the contour . For ﬁxed coordinate values , ζ outside the disc (, ζ ∈ / B ), a careful discussion would show that (, ζ , λ) is a regular function in
2.3 The rigidly rotating disc of dust
49
λ outside and jumps along , i.e. satisﬁes a Riemann–Hilbert problem. Here we assume to have this behaviour and are justiﬁed in our assumption in that we obtain the solution to the problem. Let us now consider the jump −1 + − , where the signs refer to the two sides of , cf. Fig. 2.5. The LP tells us that −1 + − does not depend on the coordinates and is therefore a function D of the contour alone, −1 + − = Du (K),
K ∈ u ,
−1 + − = Dl (K),
K ∈ l ,
(2.75)
where u denotes the upper and l the lower sheet. Since the jump contours u , l and the jump matrices Du , Dl are the same for all values of , ζ (i.e. for all Riemann surfaces with different branch points), we can express Du and Dl in terms of the axis values of , −1 F− 0 F+ 0 Du (K) = (2.76) , K ∈ u . G+ 1 G− 1 A similar relation for Dl may be obtained via (2.52). As a consequence of −1 F 0 does not jump along u . Because of (2.56), (2.75), (2.76) the matrix G 1 −1 F 0 and, ﬁnally, for R as deﬁned in (2.70). Consider the same holds for G 1 now the Riemann Ksurface corresponding to the disc’s rim = 0 , ζ = 0. The cut between the branch points KB = ±i 0 coincides with the contour u , l which are on the two ‘bridges’ connecting crosswise the upper with the lower sheet. Since R does not jump on u , we have, according to (2.71), (MS −1 )− = −(S −1 M)+ . Though R does not jump, F and G do jump, cf. (2.76). Note that F(K) and G(K) are unique functions of K. Hence, there is only one contour : K = 0, −0 ≤ K ≤ 0 where M does jump. Since is analytic outside u , l , the matrix M must be analytic outside . Thus we obtain F(K) and G(K) from the Riemann–Hilbert problem K ∈:
SM− = −M+ S ,
K∈ /:
M(K) analytic in K,
(2.77)
S and M as in (2.72). (Note that the elements of S , which are polynomials, and the elements of S −1 , which are rational functions in K, do not jump along .) There is no jump at the end points of the contour K = ±i0 , M(±i0 )− = M(±i0 )+ . As a consequence, one obtains Tr S (±i0 ) = 0, i.e. the parameter relation
f0 f 0 + 42 02 = 1.
(2.78)
50
Analytical treatment of limiting cases
It turns out that the Riemann–Hilbert problem (2.77) has a unique solution M(K) in the parameter region7 0 < µ ≡ 22 e−2V0 02 < µ0 = 4.62966184 . . .
(2.79)
An important step on the way to this solution is the diagonalization of S . Finally, one obtains F(K), G(K) and the axis values of the Ernst potential f (ζ ) in terms of elliptic theta functions. We need not go along this route. As we shall see in the following, we can use the Riemann–Hilbert problem (2.77) to formulate a more general Riemann–Hilbert problem, which will yield the complete disc of dust solution in terms of hyperelliptic theta functions. Ernst potential everywhere In order to construct the matrix for arbitrary values of , ζ and λ, let us return to −1 F 0 the Riemann–Hilbert problem (2.77). As we have seen, the matrix G 1 −1 1 G does not jump along u . Analogously, does not jump along l . The 0 F images λ of u and −λ of l inherit these properties, which is essential to the following deductions. To formulate a Riemann–Hilbert problem in the λplane, we deﬁne two matrices, −1 1 0 1 0 1 G 1 G −1 L := M 0 −1 0 −1 0 F 0 F −1 0 −1 0 1 F 0 F 0 (2.80) −1 = M 1 0 −1 0 G 1 G 1 0 1 = −1 , 1 0 −1 1 1 G Q := e 0 0 F −1 0 F 0 = e−2V0 1 G 1 −2V0
where
0 1 0 1 G (S + w1) −1 −1 0 F 0 −1 (2.81) −1 0 1 F 0 (S + w1) −1 , 0 −1 0 G 1
1 w = − Tr S = 22 (K 2 + 02 ). 2
(2.82)
7 In this section the symbol µ denotes a parameter deﬁned in (2.79), and should not be confused with the
massdensity. The meaning of µ0 will be discussed in Subsection 2.3.3.
2.3 The rigidly rotating disc of dust
51
Here we have made use of the parameter relation (2.78). Since S and w are polynomials in K, and therefore rational functions in λ, the matrix Q does not jump at all. Taking the asymptotics of S and w into account, Q must have the following structure in λ: q2 q , (2.83) Q = (K + i z)2 1 q3 −q2 with the polynomials q1 = kλ + lλ3 , q2 = m + nλ2 + pλ4 , q3 = q + rλ2 + sλ4 ,
(2.84)
where k, l, m, n, p, q, r and s are functions of , ζ alone. From the deﬁnitions (2.80), (2.81) and condition (2.74), we can derive QL = −LQ,
(2.85)
whereas the particular Riemann–Hilbert problem (2.77) has the continuation λ ∈ λ :
(Q + e−2V0 w1)L− = −L+ (Q + e−2V0 w1),
λ ∈ −λ :
(Q − e−2V0 w1)L− = −L+ (Q − e−2V0 w1),
λ∈ / λ , −λ :
(2.86)
L analytic in λ.
The following solution of the regular Riemann–Hilbert problem (2.86) is based on the diagonalization of Q. We consider a function deﬁned by √ 1 Lˆ 22 + 1 + w2 e−4V0 Lˆ 21 := √ ln , (2.87) √ w2 + e4V0 Lˆ 22 − 1 + w2 e−4V0 Lˆ 21
Q 1 11 . Lˆ = L 0 Q21
where
(2.88)
Note that has no branch points at the zeros K1 , K2 , K 1 = −K2 and K 2 = −K1 of w2 + e4V0 , K12 = 02
i−µ , µ
K22 = 02
i+µ µ
[ K1 < 0,
K2 > 0,
µ as in (2.79)],
(2.89) √ since is unaffected by a change in the sign of w2 + e4V0 . It is an odd function of λ, thus vanishing at λ = 0 and at λ = ∞. Therefore, the function ˆ := /[λ(K + i z)] = / (K − i¯z )(K + i z) (2.90)
52
Analytical treatment of limiting cases
is a unique function of K, which is characterized by the following properties: (i) There is a jump along , which because of (2.86) reads ˆ− = ˆ+ + √
2
√ (K − i¯z )(K + i z) w2 + e4V0
√ w2 + e4V0 + w . ln √ w2 + e4V0 − w
(2.91)
(ii) Because of Lˆ 221 (1 + w2 e−4V0 ) − Lˆ 222 = Q221 , Q21 = −
2f 2 e−2V0 f +f
(K − Ka )(K − Kb ),
(2.92) (2.93)
the behaviour for K → Ka/b is given by ˆ →
±2 2 + e4V0 ) (Ka/b − i¯z )(Ka/b + i z)(wa/b
ln(K − Ka/b ) as K → Ka/b . (2.94)
(The ambiguity of sign can be compensated for by the square root.) (iii) Because of the deﬁnitions of Q and L, the behaviour for K → ∞ is given by ˆ →
ln f 2 K 3
as
K → ∞.
(2.95)
ˆ that possesses all these properties is A representation of 1 ˆ = πi
i0 −i0
Ka −2 K1
Kb −2 K2
√ √ ln[ w2 + e4V0 + w ]/[ w2 + e4V0 − w ] dK √ √ (K − i¯z )(K + i z) w2 + e4V0 (K − K) 1 dK 2 4V (K − i¯z )(K + i z)(w + e 0 )(K − K)
(2.96)
1 dK , (K − i¯z )(K + i z)(w2 + e4V0 )(K − K)
ˆ = O(K −3 ). The lower limits where Ka and Kb have to be determined such that of integration in the last two integrals have been ﬁxed to obtain the correct result in the Newtonian limit µ → 0 where Ka /K1 = 1 + O(µ2 ) and Kb /K2 = 1 + O(µ2 ). For a systematic postNewtonian expansion of the solution, see Petroff and Meinel (2001). Note that the last two terms in Equation (2.96) may also be
2.3 The rigidly rotating disc of dust
53
interpreted as follows, K1 Ka Kb Kb Kb Kb 2 + = 2 {−} + = {1} + {2}, K1
Ka
K2
K2
Ka
(2.97)
Ka
showing that nothing special happens at K1 and K2 . In this symbolic notation, {−} indicates that the square root is meant to have the opposite sign from that of the ﬁrst term; {1} and {2} denote different paths in the complex Kplane, which are chosen such that the closed integral "
Kb Kb K2 = {1} − {2} = 2 Ka
Ka
(2.98)
K1
is √ performed around a contour enclosing the branch points K1 and K2 of w2 + e4V0 . In the subsequent formulae we normalize K by introducing K , 0
X =
Ka/b , 0
Xa/b =
X1/2 =
K1/2 . 0
(2.99)
According to (2.95), an asymptotic expansion of Equation (2.96) for X → ∞ (K → ∞) leads to Xa 2 Xb 2 i 2 X dX hX dX X dX ln f = µ + − (2.100) , W W W1 X1
Xa X1
dX + W
Xb
dX = W
X2
i −i
hdX , W1
−i
X2
Xa X1
X dX + W
Xb
X dX = W
X2
i −i
hX dX , W1
(2.101)
where the lower limits of integration X1 , X2 are given by X12 =
i −µ , µ
X22 = −
i +µ µ
( X1 < 0,
X2 > 0),
(2.102)
whereas the upper limits Xa , Xb must be calculated from the integral equations (2.101). Here we have introduced the abbreviations W = W1 W2 , W1 = (X − ζ /0 )2 + (/0 )2 , (2.103) W2 = 1 + µ2 (1 + X 2 )2
54
Analytical treatment of limiting cases
1 + µ2 (1 + X 2 )2 + µ(1 + X 2 ) h= . (2.104) πi 1 + µ2 (1 + X 2 )2 The third integral in (2.100) as well as the integrals on the right hand sides in (2.101) have to be taken along the imaginary axis in the complex Xplane with h and W1 ﬁxed according to W1 < 0 (for , ζ outside the disc) and h = 0. The task of calculating the upper limits Xa , Xb in (2.101) from and
ln
i u= −i
hdX , W1
i
hX dX W1
v= −i
(2.105)
is known as Jacobi’s inversion problem. Göpel (1847) and Rosenhain (1850) were able to express the hyperelliptic functions Xa (u, v ) and Xb (u, v ) in terms of (hyperelliptic) theta functions. Later on it turned out that even the ﬁrst two integrals in (2.100) can be expressed by theta functions in u and v ! A detailed introduction to the related mathematical theory which was founded by Riemann and Weierstrass may be found in Stahl (1896), Krazer (1903) and Belokolos et al. (1994). The representation of the Ernst potential (2.100) in terms of theta functions can be taken from Stahl’s book, see Stahl (1896), p. 311, Equation (5). Here is the result: Deﬁning a theta function ϑ(x, y; p, q, α) by ϑ(x, y; p, q, α) :=
∞ #
∞ #
2
2
(−1)m+n pm qn e2mx+2ny+4mnα ,
(2.106)
m=−∞ n=−∞
one can reformulate the expressions (2.100), (2.101) to give f =
ϑ(α0 u + α1 v − C1 , β0 u + β1 v − C2 ; p, q, α) −(γ0 u+γ1 v+µw) e , ϑ(α0 u + α1 v + C1 , β0 u + β1 v + C2 ; p, q, α)
(2.107)
with u and v as in (2.105) and i w= −i
hX 2 dX . W1
(2.108)
The normalization parameters α0 , α1 ; β0 , β1 ; γ0 , γ1 , the moduli p, q, α of the theta function and the quantities C1 , C2 are deﬁned on the two sheets of the hyperelliptic Riemann surface related to W = µ (X − X1 )(X − X¯1 )(X − X2 )(X − X¯2 )(X − i¯z /0 )(X + i z/0 ), (2.109)
2.3 The rigidly rotating disc of dust
55
Fig. 2.6. Riemann surface with cuts between the branch points X1 and X¯ 1 , X2 and X¯ 2 , −i z/0 and i¯z /0 . Also shown are the four periods ai and bi (i = 1, 2). Solid/dashed lines belong to the upper/lower sheet deﬁned by W → ±µX 3 as X → ∞.
see Fig. 2.6. There are two normalized Abelian differentials of the ﬁrst kind dX X dX + α1 , W W dX X dX + β1 , dω2 = β0 W W dω1 = α0
deﬁned by
(2.110) (2.111)
" dωn = πi δmn
(m = 1, 2; n = 1, 2).
(2.112)
am
Equation (2.112) consists of four linear, algebraic equations and yields the four parameters α0 , α1 , β0 , β1 in terms of integrals extending over the closed (deformable) curves a1 , a2 . It can be shown that there is one normalized Abelian differential of the third kind, dω = γ0
X 2 dX dX X dX + γ1 +µ , W W W
(2.113)
56
Analytical treatment of limiting cases
with vanishing aperiods " dω = 0 ( j = 1, 2).
(2.114)
aj
This equation deﬁnes γ0 and γ1 , again via a linear, algebraic system. The Riemann matrix (Bij ) =
ln p 2α
2α ln q
(i = 1, 2; j = 1, 2)
(2.115)
(with negative deﬁnite real part) is given by " Bij =
dωj
(2.116)
bi
and deﬁnes the moduli p, q, α of the theta function (2.106). Finally, the quantities C1 , C2 can be calculated by ∞+ Ci = −
dωi
(i = 1, 2),
(2.117)
−i z/0
where ‘+’ denotes the upper sheet. Obviously, all the quantities entering the theta functions and the exponential function in (2.107) can be expressed in terms of welldeﬁned integrals and depend on the three parameters /0 , ζ /0 , µ. The corresponding ‘tables’ for αi , βi , γi , Ci , Bij , u, v and w can easily be calculated by numerical integration. Fortunately, theta series like (2.106) converge rapidly. For 0 < µ < µ0 , the solution (2.107) is analytic everywhere outside the disc, even at the rings −i z/0 = X1 , X2 . The limit µ → µ0 will be discussed in detail in Subsection 2.3.5. Note that for µ > µ0 , the Ernst potential (2.107) is no longer singularityfree outside the disc. This corresponds to the fact that the boundary value problem (2.39) has a unique solution, since the range 0 < µ < µ0 already covers the full range 0 > V0 > −∞, as will be shown in Subsection 2.3.3.
2.3 The rigidly rotating disc of dust
57
The complete metric e2U , a
and e2k calculated from the Ernst potential (2.107) The metric functions according to (1.44) and (1.110) are given as follows: e2U = 1+
ϑ(c)ϑ ∗ (c)ϑ(a)ϑ ∗ (a) e−(γ0 u+γ1 v+µw) , ϑ(0)ϑ ∗ (0)ϑ(a + c)ϑ ∗ (a + c)
(2.118)
(1 + a)e2U ϑ(0)ϑ ∗ (0)ϑ(a + 2c)ϑ ∗ (a) = , ϑ(c)ϑ ∗ (0)ϑ(a + c)ϑ ∗ (a + c) κ(, ζ ) e2k(,ζ ) = κ(0, 0)
with
(2.120)
2 ∂ 2 ln ϑ(x)ϑ ∗ (x) 1 # κ(, ζ ) = exp 2k − a a 0 i k ϑ(0)ϑ ∗ (0) 2 ∂xi ∂xk
ϑ(a)ϑ ∗ (a)
(2.119)
i,k=1
, (2.121) x=0
where µ2 2k0 = 4
i i
(X − X1 )(X − X2 )(X + X1 )(X + X2 )
−i −i
(λ − λ )2 h(X )h(X ) dX dX , λλ (X − X )2
X − iz/0 X − iz/0 λ= , λ = , X + i z/0 X + i z/0 ×
ϑ(x) = ϑ(x; p, q, α) = ϑ(x1 , x2 ; p, q, α), iπ iπ , x2 + ; p, q, α), 2 2 a = (a1 , a2 ) = (α0 u + α1 v , β0 u + β1 v ), 0 = (0, 0), c = (C1 , C2 ). ϑ ∗ (x) = ϑ(x1 +
(2.122) (2.123) (2.124) (2.125) (2.126)
2.3.3 Mathematical discussion of the solution The aim of this subsection is to perform a careful analysis of the mathematical properties of the solution of the rigidly rotating disc of dust in terms of theta functions. The main focus of our attention will be the Ernst potential (2.107), but we shall also consider Equations (2.119) and (2.120) for the complete metric. We begin by presenting a general method permitting the use of these equations in
58
Analytical treatment of limiting cases
a direct numerical calculation (Kleinwächter 2001). The result is theta formulae that contain only deﬁnite real integrals as arguments. Besides providing for easy numerical applicability, all the related equations are well suited for specializations to regions of special interest, such as the axis and the disc itself as well as the whole symmetry plane. An alternative representation of the solution can be found in Appendix 4. Throughout this and the next subsection, we shall restrict ourselves to the region ζ > 0. The symmetry relations U (, −ζ ) = U (, ζ ), a(, −ζ ) = a(, ζ ), k(, −ζ ) = k(, ζ ) and f (, −ζ ) = f (, ζ ), see (1.88) and (1.111), can be used to calculate the solution for ζ < 0. Note that the metric functions U , a and k are continuous at ζ = 0 but the imaginary part of the Ernst potential b is discontinuous at the disc: b(, ζ = 0− ) = −b(, ζ = 0+ ). The same holds for the function v of (2.105). Discussion of the theta formula for the Ernst potential In what follows, the elliptic theta functions ϑ1 , . . . , ϑ4 introduced by Jacobi and some of the ultraelliptic theta functions ϑn,k (n, k = 1, 2, 3, 4) introduced by Rosenhain are used. Moreover Jacobi’s elliptic functions sn, cn, dn, the elliptic integrals F and E, the Jacobian zeta function Z and Heuman’s lambda function 0 are used. All these functions are deﬁned in Appendix 2 and some basic properties are provided there. In this notation, (2.107) for the Ernst potential reads ϑ4,4 (α0 u + α1 v − C1 , β0 u + β1 v − C2 ; B11 , B22 , B12 ) . ϑ4,4 (α0 u + α1 v + C1 , β0 u + β1 v + C2 ; B11 , B22 , B12 ) (2.127) Before starting, we recapitulate the meaning of the arguments in the above formula making use of the normalized coordinates f = e−(γ0 u+γ1 v+µw)
x := /0 ,
y := ζ /0 .
(2.128)
The quantities α0 , α1 , β0 , β1 , γ0 , γ1 , C1 , C2 , B11 , B22 and B12 of the ‘theta formula’ for f depend via ultraelliptic line integrals on the above normalized coordinates and on the parameter µ. The functions u, v and w, which depend on x, y and µ, are given by8 i u= −i
h dX , W1
i v= −i
h X dX , W1
i w= −i
h X 2 dX , W1
8 The convention for the root W is W < 0 for (x, y) ∈ {(r, 0) : 0 ≤ r ≤ 1}. 1 1
(2.129)
2.3 The rigidly rotating disc of dust
with h=
ln
1 + µ2 (1 + X 2 )2 + µ(1 + X 2 ) , i π 1 + µ2 (1 + X 2 )2
W1 =
59
(X − y)2 + x2 . (2.130)
The normalization parameters α0 , α1 , β0 , β1 , γ0 , γ1 , the moduli Bmn of the theta functions and the quantities C1 , C2 are deﬁned via integrals on the two sheets of the hyperelliptic Riemann surface (see Fig. 2.6) related to W = µ (X − X1 )(X − X1 )(X − X2 )(X − X2 )(X − y − i x)(X − y + i x) (2.131) with 1 X1 = − √ 1 + µ2 − µ + i 1 + µ2 + µ 2µ and X2 = −X1 . The upper sheet of the surface is characterized by W → +µX 3 for X → ∞. The two normalized Abelian differentials of the ﬁrst kind are deﬁned by dω1 := α0
dX X dX + α1 , W W "
with
dω2 := β0
!
dωn = i πδmn
dX X dX + β1 W W
(m = 1, 2; n = 1, 2).
(2.132)
(2.133)
am
The values of γ0 and γ1 have to be determined by integrating over an Abelian differential of the third kind dω := γ0 with the requirement:
X 2 dX dX X dX + γ1 +µ W W W
"
!
dω = 0 (m = 1, 2).
(2.134)
(2.135)
am
The remaining quantities are deﬁned to be y−i x
" Bmn :=
dωn bm
and
Cm :=
dωm ∞+
(m = 1, 2; n = 1, 2).
(2.136)
60
Analytical treatment of limiting cases
Our ﬁrst goal is to calculate all the arguments once and for all so that for applications it is not necessary to consider the Riemann surface and Abelian differentials any more. Using these results, the theta formula will then be rewritten. With the notation 1 ξ1 := − √ 1 + µ2 − µ, 2µ 1 1 + µ2 − µ, ξ2 := √ 2µ
1 η1 := √ 2µ 1 η2 := √ 2µ
1 + µ2 + µ,
(2.137)
1 + µ2 + µ,
(2.138)
η3 := x,
ξ3 := y,
(2.139)
the cuts in the Riemann surface (Fig. 2.6) are denoted by V (ξ1 , η1 ), V (ξ2 , η2 ) and V (ξ3 , η3 ) with V (ξi , ηi ) := {ξi + i s ηi : s ∈ [−1, 1]} .
(2.140)
Let us carefully consider the behaviour of the ‘fundamental root’ (2.131). First we introduce the ‘elementary root’ for the complex quantity X = r + i s jumping at the cut V (a, b)
W (r, s; a, b) := (X − a) 1 +
b2 (X − a)2
(r, s and a, b real),
(2.141)
√ where the choice . . . ≥ 0 is made. The fundamental root (2.131) can be built up from (2.141) using the deﬁnitions (2.137)–(2.139) and (2.140) W (r, s; µ, x, y) = µW (r, s; ξ1 , η1 )W (r, s; ξ2 , η2 )W (r, s; ξ3 , η3 ).
(2.142)
The radicand R(r, s; a, b) of (2.141) is given by b2 = RR (r, s; a, b) + i RI (r, s; a, b), (X − a)2
b2 (r − a)2 − s2 RR (r, s; a, b) ≡ 1 +
2 , (r − a)2 + s2
1+
(r − a)b2
RI (r, s; a, b) ≡ −2s
(r − a)2 + s2
2 .
(2.143) (2.144)
(2.145)
2.3 The rigidly rotating disc of dust
With the notation W = (X − a)(SR + i SI ), one ﬁnds9 1 R2R + R2I + RR , SR (r, s; a, b) ≡ √ 2 1 R2R + R2I − RR , SI (r, s; a, b) ≡ sign(RI ) √ 2 sign(RI ) = −sign(s) sign(r − a). The deﬁnition 1 g± (r, s; a, b) := √ 2
R2R + R2I ± RR
61
(2.146) (2.147) (2.148)
(2.149)
1 ( 2 !
2
b2 (r − a)2 − s2 1 (r − a)2 + s2 + b2 − 4b2 s2 ± 1 + =√
2 (r − a)2 + s2 2 (r − a)2 + s2 leads to the following representation of the elementary root W (r, s; a, b) = WR (r, s; a, b) + i WI (r, s; a, b),
(2.150)
WR (r, s; a, b) = sign(r − a) [r − ag+ (r, s; a, b) + sg− (r, s; a, b)] , WI (r, s; a, b) = sign(s) [sg+ (r, s; a, b) − r − ag− (r, s; a, b)] .
(2.151)
From these equations, one can work out the following properties for W and W , which will be used later to calculate the integrals on the Riemann surface: lim W (a ± ε, s; a, b) = ± b2 − s2 (s ≤ b), ε→0 W (r, s = 0; a, b) = sign(r − a) (r − a)2 + b2 , (2.152) W (r, −s; a, b) = W (r, s; a, b).
For the fundamental root, this ﬁnally leads to W (r, −s; µ, x, y) = W (r, s; µ, x, y), W (r = ξi − ε, s; µ, x, y) = −W (r = ξi + ε, s; µ, x, y) (i ∈ {1, 2, 3}, s < ξi , ε sufﬁciently small), W (r, s = 0; µ, x, y) = sign [(r − ξ1 )(r − ξ2 )(r − ξ3 )] 2 2 2 2 × µ (r − ξ1 ) + η1 (r − ξ2 ) + η2 (r − ξ3 )2 + η32 .
(2.153)
9 The roots that appear here are roots of positive, real quantities and must always be taken to have positive sign.
62
Analytical treatment of limiting cases
In the following formulae, (i, j, k) is any permutation of (1, 2, 3). With GR (s; ai ) := WR (ξi , s; ξj , ηj )WR (ξi , s; ξk , ηk ) − WI (ξi , s; ξj , ηj )WI (ξi , s; ξk , ηk ), GI (s; ai ) := WR (ξi , s; ξj , ηj )WI (ξi , s; ξk , ηk )
(2.154)
+ WI (ξi , s; ξj , ηj )WR (ξi , s; ξk , ηk ), one ﬁnds W±ai (s; µ, x, y) := lim W (ξi ± ε, s; µ, x, y) ε→0 = ±µ ηi2 − s2 W (ξi , s; ξj , ηj )W (ξi , s; ξk , ηk ) = ±µ ηi2 − s2 [GR (s; ai ) + i GI (s; ai )]
(2.155)
for the fundamental root along the cut V (ξi , ηi ). For the calculation of the normalizing coefﬁcients α0 , α1 , β0 , β1 , γ0 , γ1 and the quantities B11 , B12 , B22 the following notation is introduced: " Ani (µ, x, y)
:= ai
X n dX = 2i W (X )
ηi −ηi
(ξi + i s)n ds . W+ai (s; µ, x, y)
(2.156)
Consider, for example, A0i :
A0i (µ, x, y)
2i = µ 2 = µ
ηi −ηi
ds GR (s; ai ) − i GI (s; ai ) 2 2 GR (s; ai ) + GI (s; ai ) η2 − s2 i
ηi
ds GI (s; ai ) 2 2 GR (s; ai ) + GI (s; ai ) η2 − s2 −ηi i
2i + µ
(2.157)
ηi
ds GR (s; ai ) . 2 2 GR (s; ai ) + GI (s; ai ) η2 − s2 −ηi i
Due to the symmetry properties of GI (s; ai ), the real part of the equation above vanishes. Nevertheless, for the calculation of B11 , B12 and B22 , it is useful to use
2.3 The rigidly rotating disc of dust
63
the corresponding integrals over the interval [0, ηi ]. Therefore the integrals L0i (µ, x, y)
1 := µ
L1i (µ, x, y) :=
L2i (µ, x, y)
Ii0 (µ, x, y)
Ii1 (µ, x, y)
Ii2 (µ, x, y)
1 µ
1 := µ 1 := µ 1 := µ 1 := µ
ηi
ds GR (s; ai ) , GR2 (s; ai ) + GI2 (s; ai ) η2 − s2
(2.158a)
ds ξi GR (s; ai ) + sGI (s; ai ) , GR2 (s; ai ) + GI2 (s; ai ) 2 2 η −s
(2.158b)
i
0
ηi
i
0
ηi
(ξi2 − s2 )GR (s; ai ) + 2ξi sGI (s; ai ) GR2 (s; ai ) + GI2 (s; ai )
0
ηi 0
ηi
ds , 2 2 ηi − s
(2.158c)
ds GI (s; ai ) , 2 2 GR (s; ai ) + GI (s; ai ) η2 − s2 i
(2.158d)
ds ξi GI (s; ai ) − s GR (s; ai ) , 2 2 GR (s; ai ) + GI (s; ai ) 2 2 η −s
(2.158e)
i
0
ηi
(ξi2 − s2 )GI (s; ai ) − 2ξi s GR (s; ai ) GR2 (s; ai ) + GI2 (s; ai )
0
ds ηi2 − s2
(2.158f )
are deﬁned and hence Ani (µ, x, y) = 4i Lni (µ, x, y).
(2.159)
For the calculation of the moduli Bmn and the quantities C1 , C2 , integrals of the type d n Kc,d (µ, x, y)
:= c
r n dr W (r, s = 0; µ, x, y)
(n = 0, 1)
(2.160)
are also used. These integrals, which run along parts of the real axis, can be computed using the last equation of (2.153). Due to Equations (2.132), (2.133) and (2.159) we have α0 L01 + α1 L11 =
π , 4
α0 L02 + α1 L12 = 0,
β0 L01 + β1 L11 = 0, β0 L02 + β1 L12 =
π . 4
(2.161) (2.162)
64
Analytical treatment of limiting cases
A consequence of these relations and of " " " dωn + dωn + dωn = 0 a1
a2
(n = 1, 2)
(2.163)
a3
is
π π and β0 L03 + β1 L13 = − , 4 4 which will be used in the context of the calculation of Bmn and Cm . α0 L03 + α1 L13 = −
(2.164)
Determination of α0 , α1 , β0 , β1 : Due to (2.161) and (2.162), we have to solve the two linear systems 0 L1 L11 α0 π/4 = , (2.165) 0 α1 L02 L12 0 L1 L11 β0 0 = . (2.166) π/4 L02 L12 β1 The solutions are π 1 π 0 π 1 π 0 L2 , α1 = − L2 and β0 = − L1 , β1 = L , α0 = 4Da 4Da 4Da 4Da 1
(2.167)
where Da := L01 L12 − L02 L11 . Determination of γ1 and γ2 : The deﬁnition (2.134) and the condition (2.135) lead to the system 0 2 L1 L11 L1 γ0 = −µ (2.168) 0 1 L2 L2 γ1 L22 with the solutions γ0 = −µ
L21 L12 − L22 L11 , Da
γ1 = −µ
L01 L22 − L02 L21 . Da
(2.169)
Determination of B11 , B22 , B12 and C1 , C2 : It turns out that for the calculation of these quantities, two cases have to be discussed separately, namely y > X2 (corresponding to the cut V (ξ3 , η3 ) being to the right of V (ξ2 , η2 ), see Fig. 2.7) and 0 < y < X2 (corresponding to V (ξ3 , η3 ) being to the left of V (ξ2 , η2 ), see Fig. 2.8). In both cases the path of integration is divided into suitable parts. We use the following representation for Bmn and Cm (m, n = 1, 2) " Bmn =
dωn = 2 bm
# l b m(l)
P dωn ,
Cn =
dωn = +∞
# l c(l)
dωn .
(2.170)
2.3 The rigidly rotating disc of dust
65
Fig. 2.7. Complex plane with cuts V (ξ1 , η1 ), V (ξ2 , η2 ) and V (ξ3 = y, η3 = x) for the case y > X2 . This case corresponds to the Riemann surface shown in Fig. 2.6.
The structure Bmn = 2{. . .} is due to the fact that on the Riemann surface (see Fig. 2.6), one also has to integrate the corresponding return path in the lower sheet for calculating the moduli. The case y > X2 : From Fig. 2.7, one can read off the corresponding Bmn : , 0 + 2I20 + Kξ02 ,ξ1 + I10 ) B11 = 2 α0 (I30 + Ky,ξ 2 1 + 2I21 + Kξ12 ,ξ1 + I11 ) , + α1 (I31 + Ky,ξ 2 , 0 1 + I20 ) + β1 (I31 + Ky,ξ + I21 ) , B22 = 2 β0 (I30 + Ky,ξ 2 2 . 0 B12 = 2 β0 (I30 + Ky,ξ + 2I20 + Kξ02 ,ξ1 + I10 ) 2 iπ/ 1 1 1 1 + 2 I + K + I ) − , + β1 (I31 + Ky,ξ ξ2 ,ξ1 2 1 2 4 . iπ/ 0 0 1 1 1 + I ) + α ( I + K + I ) − , B21 = 2 α0 (I30 + Ky,ξ 1 y,ξ 2 3 2 2 2 4
(2.171)
and C1 , C2 are given by iπ , 4 iπ 0 1 . + I30 ) + β1 (K∞,y + I31 ) + C2 = β0 (K∞,y 4
0 1 + I30 ) + α1 (K∞,y + I31 ) + C1 = α0 (K∞,y
(2.172)
66
Analytical treatment of limiting cases
Fig. 2.8. Complex plane with cuts V (ξ1 , η1 ), V (ξ2 , η2 ) and V (ξ3 = y, η3 = x) for the case 0 < y < X2 .
The case 0 < y < X2 : Analogously, from Fig. 2.8 one ﬁnds Bmn : , 0 1 B11 = 2 α0 (I30 + Ky,ξ + I10 ) + α1 (I31 + Ky,ξ + I11 ) , 1 1 , 1 0 − I20 ) + β1 (−I31 + Ky,ξ − I21 ) , B22 = 2 β0 (−I30 + Ky,ξ 2 2 . iπ/ 0 0 1 1 1 B12 = 2 β0 (I30 + Ky,ξ , + I ) + β ( I + K + I ) + 1 y,ξ1 1 3 1 1 4 . iπ/ 0 0 1 1 1 , − I ) + α (− I + K − I ) + B21 = 2 α0 (−I30 + Ky,ξ 1 y,ξ2 2 3 2 2 4
(2.173)
and C1 , C2 are given by 0 + 2I20 + Kξ02 ,y + I30 ) C1 = α0 (K∞,ξ 2 1 + α1 (K∞,ξ + 2I21 + Kξ12 ,y + I31 ) + 2
iπ , 4
0 C2 = β0 (K∞,ξ + 2I20 + Kξ02 ,y + I30 ) 2 1 + 2I21 + Kξ12 ,y + I31 ) − + β1 (K∞,ξ 2
(2.174)
iπ . 4
The imaginary parts of the equations follow directly from (2.161), (2.162) n and (2.164). The symbol ∞ in K∞,y stands for ‘+ real inﬁnity’. Note that B21 = B12 holds in general, see e.g. Krazer (1903). Since different expressions for each of
2.3 The rigidly rotating disc of dust
67
these quantities can be found in (2.171) and (2.173), one can derive useful identities and one has a further test for verifying a numerical evaluation of these integrals. Using symmetry properties, the functions u, v and w (2.129) are found to be10 2 u(µ, x, y) = π 2 v (µ, x, y) = π
1 0
1 0
2 w(µ, x, y) = − π
ln[ 1 + µ2 (1 − s2 )2 + µ(1 − s2 )] WR (0, s; y, x) ds, WR2 + WI2 1 + µ2 (1 − s2 )2 ln[ 1 + µ2 (1 − s2 )2 + µ(1 − s2 )] sWI (0, s; y, x) ds, WR2 + WI2 1 + µ2 (1 − s2 )2
1 0
ln[ 1 + µ2 (1 − s2 )2 + µ(1 − s2 )] s2 WR (0, s; y, x) ds. WR2 + WI2 1 + µ2 (1 − s2 )2 (2.175)
Transformation of the theta formula: Now, using relations between Rosenhain’s theta functions, the solution for the Ernst potential is transformed in such a way that all the arguments become real. First we write down the relation ϑ4,4 (x1 , x2 ; B11 , B22 , B12 ) = ϑ3,3 (x1 + x2 , x1 − x2 ; B11 + 2B12 + B22 , B11 − 2B12 + B22 , B11 − B22 ) − ϑ2,2 (x1 + x2 , x1 − x2 ; B11 + 2B12 + B22 , B11 − 2B12 + B22 , B11 − B22 ), (2.176) which can be deduced from the deﬁnitions (A2.10). It will be convenient to introduce the following combinations of the arguments: L := exp {−(γ0 u + γ1 v + µw)} ,
S := (α0 + β0 )u + (α1 + β1 )v ,
(2.177)
T := B11 + B22 + 2 B12 ,
A := (α0 − β0 )u + (α1 − β1 )v ,
(2.178)
B := B11 + B22 − 2 B12 ,
C := (C1 + C2 ),
(2.179)
R := B11 − B22 ,
D := (C1 − C2 ).
(2.180)
The previous results lead immediately to the equations (upper sign corresponds to y > X2 and lower sign to 0 < y < X2 ) π π π B12 = B12 ∓ i , C1 = C1 + i , C2 = C2 ± i , 2 4 4 (2.181) B11 + B22 + 2B12 = T ∓ i π, B11 + B22 − 2B12 = B ± i π , 10 The arguments of the denominator W 2 + W 2 are always the same as for the corresponding numerator W R/I . R I
68
Analytical treatment of limiting cases
and hence π , C1 − C2 = (C1 − C2 ), 2 π 0 < y < X2 : C1 + C2 = (C1 + C2 ), C1 − C2 = (C1 − C2 ) + i . 2 (2.182) y > X2 : C1 + C2 = (C1 + C2 ) + i
Applying (2.176) to (2.127) and using the properties (A2.11), (A2.12) leads to the results : Case y > X2 : f =L Case
ϑ3,4 (S − C, A − D; T , B, R) + i ϑ1,2 (S − C, A − D; T , B, R) . ϑ3,4 (S + C, A + D; T , B, R) − i ϑ1,2 (S + C, A + D; T , B, R)
(2.183)
0 < y < X2 :
f =L
ϑ4,3 (S − C, A − D; T , B, R) + i ϑ2,1 (S − C, A − D; T , B, R) . ϑ4,3 (S + C, A + D; T , B, R) − i ϑ2,1 (S + C, A + D; T , B, R)
(2.184)
± := ϑn,k (S ± C, A ± D; T , B, R), one ﬁnds for e2U and b : Finally, with ϑn,k
Case
y > X2 : e
Case
2U
− + − + ϑ3,4 − ϑ1,2 ϑ1,2 ϑ3,4 =L 2 2 , + + + ϑ1,2 ϑ3,4
− + + − ϑ1,2 + ϑ3,4 ϑ1,2 ϑ3,4 b=L 2 2 . + + + ϑ1,2 ϑ3,4
(2.185)
− + + − ϑ2,1 + ϑ4,3 ϑ2,1 ϑ4,3 b=L 2 2 . + + ϑ4,3 + ϑ2,1
(2.186)
0 < y < X2 : − + − + ϑ4,3 − ϑ2,1 ϑ2,1 ϑ4,3 e2U = L 2 2 , + + ϑ4,3 + ϑ2,1
Applications: The arguments (2.177)–(2.180) of Equations (2.186), (2.185) are given by (2.167), (2.169), [(2.171) and (2.172) for y > X2 ] or [(2.173) and (2.174) for 0 < y < X2 ] and (2.175). These quantities can be calculated as shown above with the three types of integrals (2.158a)–(2.158c), (2.158d)–(2.158e) and (2.160). These integrals themselves can be numerically evaluated very easily. Whether or not the ycoordinate approaches X2 from the left or the right, (2.186) and (2.185) converge to the same result.11 In this region, a little more attention must be paid to 11 The limit (ξ = y, η = x) → (ξ , η ) can be calculated analytically. 3 3 2 2
2.3 The rigidly rotating disc of dust
69
the numerical evaluation of the integrals. The theta functions, on the other hand, are easy to handle numerically, since the rapidly converging series make it possible 0j to replace the inﬁnite sums (A2.1) and (A2.10) by ﬁnite sums of the type n=−j , with, say j = 10. Equations (2.186) and (2.185) and the formulae for the related arguments are also very well suited to deriving simpler formulae for the special cases of the plane of symmetry (y = 0) and the axis (x = 0). The simplest form of the Ernst potential and the whole metric is to be found within the disc (y = 0, x ≤ 1) (Kleinwächter 2000). How these formulae can be derived will be explained in some detail for the latter case. For the other two cases, we just list the results. Disc metric In this case, one has to consider y = 0+ , x ≤ 1 and thus to use the formulae for the case 0 < y < X2 . It turns out that β0 = α0 ,
C = 2 (C1 ),
T = 2 (B11 + B12 ),
β1 = −α1 ,
D = 0,
B = 2 (B11 − B12 ),
γ1 = 0,
S = 2α0 u,
R = 0,
B11 = B22 ,
A = 2α1 v ,
L = exp {−(γ0 u + µw)} .
(2.187)
Using the general identity ϑn,k (x1 , x2 ; a, b, c = 0) = ϑn (x1 ; a)ϑk (x2 ; b), one ﬁnds f =L
ϑ4 (S − C; T )ϑ3 (A; B) + i ϑ2 (S − C; T )ϑ1 (A; B) , ϑ4 (S + C; T )ϑ3 (A; B) − i ϑ2 (S + C; T )ϑ1 (A; B)
ϑ2 (S − C; T ) f =L ϑ2 (S + C; T )
ϑ4 (S−C;T ) ϑ2 (S−C;T ) ϑ4 (S+C;T ) ϑ2 (S+C;T )
+ i ϑϑ13 (A;B) (A;B) − i ϑϑ13 (A;B) (A;B)
(2.188) .
For convenience we introduce a new variable µ˜ := µ(1 − x2 ).
(2.189)
Within the disc, the functions u, v and w given by (2.175) in this new variable simplify to √ 1 µ ln( 1 + t 2 + t) dt u(µ, µ) ˜ =− , √ √ π µ˜ 1 + t2 µ − t t − µ˜ √ 1 µ˜ ln( 1 + t 2 + t) dt √ µv (µ, µ) ˜ = , (2.190) √ π 0 1 + t2 µ˜ − t √ √ 1 µ ln( 1 + t 2 + t) µ − t ˜ = µw(µ, µ) dt. √ π µ˜ 1 + t2 t − µ˜
70
Analytical treatment of limiting cases
Moreover, one can express the functions u and w through elliptic functions and elliptic theta functions 1 F(ϕ, l ), u(µ, µ) ˜ = − 4 (1 + µ2 )(1 + µ˜ 2 ) µ + µ˜ 2 µw(µ, µ) + 1 + µ˜ ˜ = − ln 2 1 2 π 1 + µ˜ 2 − (µ − µ) ˜ − F(ϕ, l ) − 0 (ψ, l) 4 K(l) (1 + µ2 )(1 + µ˜ 2 )
K(l ) π ϑ3 2K(l) F(ϕ, l ) − F(ψ, l ) ; −π K(l) . + ln ) π [F(ϕ, l ) + F(ψ, l )] ; −π K(l ϑ3 2K(l) K(l)
(2.191)
The main arguments are 3 4 4 (1 + µ2 )(1 + µ˜ 2 ) + µµ˜ − 1 ϕ(µ, µ) ˜ = arcsin 5 , (1 + µ2 )(1 + µ˜ 2 ) + µµ˜ + 1 3 4 4 (1 + µ2 )(1 + µ˜ 2 ) + µ˜ 2 + 1 − 1 + µ˜ 2 (µ − µ) ˜ 5 ψ(µ, µ) ˜ = arcsin , (1 + µ2 )(1 + µ˜ 2 ) + µµ˜ + 1 (2.192) the modulus l is given by 3 4 41 1 + µ µ ˜ l(µ, µ) ˜ =5 1− 2 1 + µ2 1 + µ˜ 2
(2.193)
and the complementary modulus l is l ≡
1 − l2.
(2.194)
Furthermore the moduli T and B of the Jacobian theta functions are T (µ, µ) ˜ = −π
K(l ) , K(l)
B(µ) ˜ = −π
with the additional moduli h˜ and h˜ 3 4 41 µ ˜ h˜ ≡ h(µ) ˜ ≡5 1+ , 2 1 + µ˜ 2
h˜ ≡
K(h˜ ) , ˜ K(h)
1 − h˜ 2 .
(2.195)
(2.196)
2.3 The rigidly rotating disc of dust
71
We make use of the following notation for functions that are closely related to v as given in (2.190), √ 1 µ ln( 1 + t 2 + t) dt 4 I (µ) := , Iˆ (µ) := 1 + µ2 I (µ). (2.197) √ √ 2 µ − t π 0 1+t All the other arguments in (2.188) follow: π 4 (1 + µ2 )(1 + µ˜ 2 ), α0 = 4K(l) π 4 √ 1 + µ˜ 2 µ, α1 = − ˜ 4K(h) 1 2 2 − (µ − µ) 1 + µ ˜ ˜ 1 π 4 γ0 = − − 0 (ψ, l) , (1 + µ2 )(1 + µ˜ 2 ) 1 + K(l) 4 π K(l) (1 + µ2 )(1 + µ˜ 2 ) π F(ϕ, l ), ˜ =− S = 2α0 u(µ, µ) 2K(l) π 4 π ˆ I (µ), ˜ ˜ =− (1 + µ˜ 2 )I (µ) ˜ =− A = 2α1 v (µ, µ) ˜ ˜ 2K(h) 2K(h) π C= K(l ) − F(ψ, l ) 2K(l) and
π µ + µ˜ 2 L= exp − F(ϕ, l ) 1 + µ˜ + 2 K(l) π ϑ3 2K(l) [F(ϕ, l ) + F(ψ, l )]; T . × π [F(ϕ, l ) − F(ψ, l )]; T ϑ3 2K(l)
(2.198)
Putting all these formulae into (2.188) and using the relations between the Jacobian theta functions and Jacobian elliptic functions (A2.20)–(A2.22), one ﬁnally ﬁnds µ − µ˜ ˜ h (µ) 2U 2 ˆ e (µ, µ) cn I (µ), , (2.199) ˜ = ˜ − ˜ h (µ) h(µ) ˜ 2 1 ˜ dn Iˆ (µ), ˜ h (µ) ˜ . (2.200) sn Iˆ (µ), ˜ h (µ) b(µ, µ) ˜ =− h(µ) ˜ Specialization to the origin (x = 0 is equivalent to µ˜ = µ) leads to h (µ) 2 ˆ e2V0 (µ) ≡ e2U (µ, µ) = cn I (µ), h (µ) , h(µ) 1 sn Iˆ (µ), h (µ) dn Iˆ (µ), h (µ) . b0 (µ) ≡ b(µ, µ) = − h(µ)
(2.201) (2.202)
72
Analytical treatment of limiting cases
From (2.199), (2.200), (2.201) and (2.202) one immediately ﬁnds the scaling property (Neugebauer and Meinel 1994) µ e2U (µ, x) = e2V0 µ(1 − x2 ) − x2 , b(µ, x) = b0 µ(1 − x2 ) , (2.203) 2 now reusing the original variables µ and x. The parameter function 0 follows from (2.79):
1 h2 (µ) ˆ cn I (µ), h (µ) . 0 (µ) = 1− 2 2 h (µ)
(2.204)
With (2.201), (2.202), (2.204) and using the general identities (A2.19) one can easily verify the parameter relation e4V0 + b20 + 42 ρ02 = 1
(see (2.78) with: f0 = e2V0 + i b0 ).
(2.205)
The value µ0 = 4.62966184 . . . (see (2.79)) is the smallest positive zero of cn Iˆ (µ), h (µ) , i.e. cn Iˆ (µ0 ), h (µ0 ) = 0. (2.206) The monotonic function V0 (µ) varies over the range 0 ≥ V0 > −∞ when µ takes the values 0 ≤ µ < µ0 . The parameter functions e2V0 , b0 and 0 are plotted in Fig. 2.9. In a similar manner as shown above for the Ernst potential, the general formulae for the metric functions a (2.119) and e2k (2.120) can be simpliﬁed within the
Fig. 2.9. The parameter functions e2V0 , b0 and 0 .
2.3 The rigidly rotating disc of dust
disc. The result for (1 + a)e2U is h h˜ ˆ (1 + a)e2U = cn I (µ), h cn Iˆ (µ), ˜ ˜ h˜ = eV0 (µ)eV0 (µ). h h˜ Hence, for e2V within the disc one gets e2V = e2U (1 + a)2 − 2 2 e−4U 2 1 (1 + a)e2U − (0 )2 x2 = 2U e / µ 1 . ˜ − x2 e2V0 (µ) = 2U e2V0 (µ)e2V0 (µ) 2 e 2V0 = e (µ)
73
(2.207)
(2.208)
thereby verifying that the boundary condition for the real part of the Ernst potential is indeed fulﬁlled, see (2.39). It turns out that three of the metric coefﬁcients gij can be expressed with the parameter function
h (µ) ˆ eV0 (µ) = cn I (µ), h (µ) (2.209) h(µ) in an especially simple manner. The complete disc metric is given by12 2 µx2 , 2 gϕϕ (µ, x) = − eV0 (µ) − eV0 (µ) ˜ + 2 µx2 gϕt (µ, x) = −eV0 (µ) , ˜ eV0 (µ) − eV0 (µ) ˜ − 2 µx2 gtt (µ, x) = −e2V0 (µ) , ˜ + 2 ˜ ˜ h) πK(h) 22 (F(ϕ, l ), l) 22 (Iˆ (µ), g (µ, x) = e2k0 eF , ˜ µ − µ˜ 24 (Iˆ (µ), h) K(h)K(l) with the function F deﬁned by 2 (0, l) 4 (0, h) 4 (0, l) 1 2 2 − − l h˜ F (ϕ, l ) l h˜ F = Iˆ (µ) 4 (0, h) h 4 (0, l) 2 (0, l) 2 1 ˜ ˜ (0, h) (0, h) + Iˆ 2 (µ) − l h˜ 2 . ˜ l h˜ 4 ˜ ˜ 4 (0, h) 2 (0, h)
(2.210)
(2.211)
12 The zero in the denominator of g ˜ = µ) is also a zero of the numerator and the metric for x = 0 (→ µ
coefﬁcient remains regular.
74
Analytical treatment of limiting cases
One should keep in mind that µ˜ = µ(1 − x2 ) and that the functions l, l and h, h are deﬁned by (2.193), (2.196) and ϕ, Iˆ are deﬁned by (2.192), (2.197). Furthermore h ≡ h(µ), h˜ ≡ h(µ) ˜ and h ≡ h (µ), h˜ ≡ h (µ). ˜ The theta functions 2 , 4 are deﬁned in Appendix 2 and 2 , 4 denote second derivatives with respect to the main argument. At the rim of the disc (x = 1 → µ˜ = 0), we have, due to e2V0 (0) = 1, a particularly simple form of gϕϕ , gϕt and gtt : 2 µ 2 gϕϕ (µ, 1) = − 1 − eV0 (µ) + , 2 µ V0 gϕt (µ, 1) = 1 − e (µ) − , 2 µ gtt (µ, 1) = −1 + . 2
(2.212)
Metric in the plane y = 0 outside the disc In this region (y = 0, x ≥ 1), it is also convenient to use µ˜ = µ(1 − x2 ) as an abbreviation (here: 0 ≥ µ˜ > −∞). It is easy to see that in this range v (µ, µ) ˜ = 0. For u and w one ﬁnds √ 1 µ ln( 1 + t 2 + t) dt u(µ, µ) ˜ =− , √ √ 2 π 0 1+t µ − t t − µ˜ √ √ 1 µ ln( 1 + t 2 + t) µ − t ˜ = µw(µ, µ) dt. √ π 0 1 + t2 t − µ˜
(2.213)
Since we have not been able to solve these integrals, they remain in the following formulae e2U = −e−G g =
1 (ˆu + F(ψ, l ), l) , 1 (ˆu − F(ψ, l ), l)
2h K(h) e2k0 eX eG 21 (ˆu − F(ψ, l ), l) , π 24 (Iˆ (µ), h) 21 (F(ψ, l ), l)
ˆ 2U − 1) − e2U , ˆ Qe ˆ 2U − 2) + 20 x(Qe 2 gϕϕ = −(0 )2 x2 Q( ˆ 2U − 1) + e2U , gϕt = −0 x(Qe gtt = −e2U
(2.214)
2.3 The rigidly rotating disc of dust
75
with 4 (1 + µ2 )(1 + µ˜ 2 )u, 1 2 π 1 + µ˜ 2 + µ˜ − µ G= uˆ + µw, 0 (ψ, l) − 4 K(l) (1 + µ2 )(1 + µ˜ 2 ) uˆ =
1 X =
π Iˆ (µ) 2K(h)
22
4 (0, h) π uˆ 2 ˜ 4 (0, l) ˜ 2 (0, l) − − hl , hl √ 4 (0, h) 2K(l) h 4 (0, l) 2 (0, l)
2 K2 (l)eG ˆ = − 4h Q π 2 41 (F(ψ, l ), l) ˜ 2 (ˆu − 2F(ψ, l ), l) h˜ 4 (ˆu − 2F(ψ, l ), l) − h × . ˜ 2 (ˆu, l) h˜ 4 (ˆu, l) − h
(2.215)
For the other quantities appearing above, see the remark after Equations (2.210); ψ is given by (2.192). Of course, in the special case x = 1, the above formulae for the metric potentials coincide with (2.210) for x = 1, in particular (2.212) holds.
Metric on the axis For x → 0, all Bperiods Bik diverge, but two of our combinations, namely B and R, remain ﬁnite: B = B11 + B22 − 2 B12 → ﬁnite value, R = B11 − B22 → ﬁnite value,
(2.216)
T = B11 + B22 + 2 B12 → −∞. On the axis, we have to deal with both the case 0 < y < X2 and the case y > X2 . In the ﬁrst case, (2.184) tells us that we have to calculate
f =L
ϑ4,3 (S − C, A − D; T , B, R) + i ϑ2,1 (S − C, A − D; T , B, R) . ϑ4,3 (S + C, A + D; T , B, R) − i ϑ2,1 (S + C, A + D; T , B, R)
(2.217)
76
Analytical treatment of limiting cases
Rosenhain’s theta functions in the above equation are given by ϑ4,3 (S ∓ C, A ∓ D; T , B, R) = ∞ #
. / (−1)m exp m2 T + 2m(S ∓ C) ϑ3 (A ∓ D + mR; B),
m=−∞
ϑ2,1 (S ∓ C, A ∓ D; T , B, R) = ∞ 2 # exp 12 (2m + 1) T + (2m + 1)(S ∓ C) m=−∞
(2.218)
× ϑ1 A ∓ D + 12 (2m + 1)R; B .
The limiting procedure x → 0 in the ﬁrst equation leads to lim ϑ4,3 (S ∓ C, A ∓ D; T , B, R) = ϑ3 (A ∓ D; B),
x→0
(2.219)
whereas in the second equation we have lim ϑ2,1 (S ∓ C, A ∓ D; T , B, R) = exp
x→0
6T
,T 4

+ (C ∓ S)
× ϑ1 (A ∓ D ∓ R2 ; B).
7 + C remains ﬁnite.) So in this case , ϑ3 (A − D; B) + i exp T4 − (S − C) ϑ1 (A − (D + R2 ); B) , f =L , ϑ3 (A + D; B) − i exp T4 + (S + C) ϑ1 (A + (D + R2 ); B)
(Note that limx→0
(2.220)
4
and similarly in the case y > X2 , ϑ4 (A − D; B) − i exp T4 − (S − C) ϑ2 (A − (D + R2 ); B) f =L . , ϑ4 (A + D; B) − i exp T4 + (S + C) ϑ2 (A + (D + R2 ); B)
(2.221)
(2.222)
On the axis, we write uA (µ, y) := u(µ, x = 0, y) and analogous expressions for v and w. It turns out that √ √ µ 2 dx ln x + 1 + x µy uA (µ, y) = − , √ √ π 1 + x2 µ − x(µ[1 + y2 ] − x) 0 I (µ) vA (µ, y) = yuA (µ, y) + √ , µ I (µ) . wA (µ, y) = yvA (µ, y) = y yuA (µ, y) + √ µ
(2.223)
2.3 The rigidly rotating disc of dust
77
Therefore the functions vA and wA , which enter some of the arguments of formulae (2.221) and (2.222) for the Ernst potential, can be expressed by the function uA . In addition to uA , one other integral that could not be given in terms of elliptic functions is Iˆ (2.197), which already entered the formulae for the plane of symmetry. It is convenient to introduce the new parameter
τ :=
4
1+
1 . µ2
(2.224)
For numerical calculations of the aforementioned remaining integrals the representation 2µτ Iˆ (µ) = π
π 2
g(µ sin2 ϕ) sin ϕ dϕ, 0 arcsinh 1 g µ 1 − y2 sinh2 α y 2 uA (µ, y) = − dα, π 0 cosh α
(2.225)
√ √ [here g is given by g(x) = ln(x + 1 + x2 )/ 1 + x2 ] is useful. If one calculates all the arguments of (2.221) and (2.222), splits these equations into real and imaginary parts, then one ﬁnds the following representation of the functions e2U and b: y4 + 2y2 + τ 4 − (1 + y2 ) (τ 2 − 1)cn+ cn− + , e2U (µ, y) = N (τ 2 − 1)cn2− + y4 + 2y2 + τ 4 − (1 + y2 ) N 2
2 4 2 4 2 b(µ, y) = (τ − 1) y + 2y + τ − (1 + y ) ×
(2.226)
N 2 cn+ − cn− . y4 + 2y2 + τ 4 − (1 + y2 ) N 2 (τ 2 − 1)cn2− +
Here we have made use of the notation
cn± := cn F(ϕ1 , h ) ± Iˆ , h ,
1 µ 1 1 1− 2 = 1− h = , 2 2 τ 1 + µ2 2
√ 2τ ϕ1 := arcsin , y2 + τ 2 + y4 + 2y2 + τ 4
(2.227)
78
Analytical treatment of limiting cases
and the function N is deﬁned by13 N := exp − µ y4 + 2y2 + τ 4 uA (µ, y) ˆ − 2I (µ) Z(Y , h ) +
h sn(Y , h )cn(Y , h )dn(Y , h ) πY − 2K(h)K(h ) 1 − h sn2 (Y , h ) y 3 + τ h y 2 + y − τ 3 h π + + 4K(h) 2τ y4 + 2y2 + τ 4 ) K(h ) π ϑ3 2K(h) Iˆ (µ) + Y + K(h , −π 2 K(h) , × (2.228) ) K(h ) π Iˆ (µ) − Y + K(h , −π ϑ3 2K(h) 2 K(h) with and
6 7 Y := sign(y − τ )F ϕ2 , h 4 + 2y 2 + τ 4 + 2hτ y y 1 . ϕ2 := arcsin 1 − h h (y + τ )2
(2.229)
(2.230)
Note that the moduli h and h are the same as in the equations for the Ernst potential at the origin (2.201), (2.202), and thus depend only on µ and not on y. For the other metric functions on the axis we have a = 0,
e2k = 1
(2.231)
and hence the complete metric is given by g = e−2U ,
gϕϕ = 0,
gϕt = 0,
gtt = −e2U .
(2.232)
A few remarks are in order: (i) At the origin, y = 0, Equations (2.226)–(2.228) lead to [see (2.201) and (2.202)] e2U (µ, y = 0) = e2V0 (µ), b(µ, y = 0+ ) = b0 (µ).
(2.233)
(ii) limy→∞ e2U (µ, y) = 1, limy→∞ b(µ, y) = 0. (iii) A series expansion of the above formulae at inﬁnity can be used to calculate all the multipole moments of the rigidly rotating disc of dust. This will be shown in the next subsection. 13 Here we use the following deﬁnition: sign(x) = 1 for x ≥ 0 and sign(x) = −1 for x < 0.
2.3 The rigidly rotating disc of dust
79
2.3.4 Physical properties of the disc In this subsection, we shall discuss physical quantities of the rigidly rotating disc of dust such as surface massdensity, baryonic and gravitational mass, angular momentum and also higher multipole moments. Furthermore, characteristic effects in connection with ergospheres and geodesic motion are considered. Gravitational mass M , angular momentum J and baryonic mass M0 can be found by specializing the corresponding formulae of Subsection 1.6.1 and calculating the resulting integrals. The fundamental quantity in this context is the surface massdensity, the derivation of which will be presented in detail using the results of Subsection 2.3.3.Alternatively, M and J can be obtained from the far ﬁeld behaviour of the metric. After deriving these and the other multipoles, we shall verify that the results for M and J obtained by these different methods agree. Later in this subsection the appearance of an ergosphere for sufﬁciently large µ and other characteristic relativistic effects are studied. Finally we shall discuss the Newtonian limit and the motion of test particles. In our calculations, the parameter functions e2V0 (µ), b0 (µ) and ρ0 (µ), as given in (2.201), (2.202) and (2.204), play an important role and the reader is referred back to Fig. 2.9. Surface massdensity, mass and angular momentum The invariant (proper) surface massdensity σp for the rigidly rotating disc of dust was already introduced in Subsection 1.7.3. Here we rewrite the important equations (1.103) and (1.104). The volume energydensity is given by = σp (ρ)eU −k δ(ζ ),
(2.234)
where δ(ζ ) is the usual Dirac delta distribution and σp can be calculated from σp =
1 U −k ∂V e 2π ∂ζ
ζ =0+
.
(2.235)
From σp we can calculate the total baryonic mass M0 , the gravitational mass M and the total angular momentum J using the general formulae already given in Subsection 1.6.1 together with (1.56): M0 = −
µB ui n dV = −
M =2
j
i
Tik − 21 Tj gik ni ξ k dV
−V0
0
ui n dV = 2πe
i
σp ek−U d, 0
(2.236)
80
Analytical treatment of limiting cases
0 = 2π
σp e
k−U
d + 4πe
−V0
0
J =−
0 σp ek−U ui ηi d, 0
Tik ni ηk dV = 2πe−V0
0 σp ek−U ui ηi d. 0
The relation M = eV0 M0 + 2J
(2.237)
follows immediately, cf. (1.157). As in Subsection 2.3.3, we use normalized coordinates x = /0 , y = ζ /0 and the combination µ˜ = µ(1 − x2 ). The aim now is to express the above quantities entirely in terms of the parameter functions eV0 (µ), b0 (µ) and 0 (µ) derived in Subsection 2.3.3. We ﬁrst consider the two parts of the integrands, namely σ ≡ σp ek−U and ui ηi of Equations (2.236). The transition formulae (1.15a), (1.15b) to the corotating coordinate system14 read e2V = e2U (1 + a)2 − 2 2 e−4U , (2.238) (1 − a )e2V = (1 + a)e2U . Differentiating the second equation with respect to y and using this relation again leads to15 a,y . (2.239) V,y = U,y + 2(1 + a) Combining Equations (2.238), we have e2V e2U = e4V (1 − a )2 − 2 02 x2 .
(2.240)
= 0 results in Differentiating with respect to y and using a,y
2U,y e2U e2V + 2V,y e2U e2V = 4V,y (1 + a)2 e4U . Combining this with (2.239) gives 2 e2V − e2U (1 + a)2 V,y = e2V
a,y . 2(1 + a)
(2.241)
(2.242)
Using (1.41) in the form a,y = −0 xe−4U b,x and the ﬁrst line of (2.238) we have b,x e2V V,y = . (2.243) 40 x (1 + a)e2U 14 Note that e2V ≡ e2U and W ≡ . 15 Note that a = 0, see (1.97). ,y
2.3 The rigidly rotating disc of dust
81
˜ and e2V = e2V0 , Finally we make use of the relations (1 + a)e2U = eV0 (µ)eV0 (µ) which are valid only within the disc, to arrive at b,x V,y = √ . 2 2µxeV0 (µ) ˜
(2.244)
˜ and hence From (2.203) one immediately concludes that b,x = −2µxb0 (µ) V,y = −
˜ 1 b0 (µ) , 2µ V ˜ 2 e 0 (µ)
(2.245)
where a prime denotes the derivative with respect to the argument. This derivative can be calculated using (A2.27) and (A2.28). Now the surface massdensity σ ≡ σp ek−U can be written as b (µ) ˜ 1 1 1 σ . V,y (x) = − e−V0 (µ) V0 (x) = 2π e 0 (µ) ˜ 2π 0
(2.246)
The proper surface massdensity σp / is plotted in Fig. 2.10 for several values of µ. The remaining term is easily seen to be ui ηi = −e−V e2V − e2U (1 + a) (2.247) ˜ = eV0 (µ) ˜ − eV0 (µ). = −e−V0 (µ) e2V0 (µ) − eV0 (µ)eV0 (µ)
Fig. 2.10. The normalized proper surface massdensity σp / is shown as a function of /0 for different values of µ (after Neugebauer and Meinel 1994).
82
Analytical treatment of limiting cases
The second equality is only valid within the disc and (2.207) was used. So recalling µ = 2(0 )2 e−2V0 and µ˜ = µ(1 − x2 ), the baryonic mass is found to be 1 σ 2 −V0 M0 (µ) = 2π(0 ) e (µ) xdx 0 1 b0 (µ) ˜ (2.248) xdx = −(0 )2 e−2V0 (µ) V ˜ 0 e 0 (µ) ˜ 1 µ b0 (µ) dµ. ˜ =− 4 0 eV0 (µ) ˜ This remaining integral will be solved in the context of the multipole moments, see (2.277), (2.274) and the subsequent discussion. For the angular momentum one ﬁnds b (µ) ˜ 1 µ V0 2 J (µ) = − b0 (µ) ˜ dµ˜ e (µ) V0 e 0 (µ) ˜ 4 0 (2.249) 1 V0 = −e (µ)M0 (µ) − b0 (µ). 4 Finally (2.237) and (2.249) lead directly to the gravitational mass 1 M (µ) = −eV0 (µ)M0 (µ) − b0 (µ). 2
(2.250)
Multipole moments The gravitational multipole moments, introduced in Geroch (1970), were generalized to the case of asymptotically ﬂat stationary spacetimes in Hansen (1974) and Thorne (1980). In the case of axisymmetric systems, Fodor et al. (1989) developed a scheme for calculating the multipole moments by expanding the Ernst potential along the axis of symmetry. Using this scheme, the moments of the rigidly rotating disc of dust were calculated by Kleinwächter et al. (1995). For calculating the expansion coefﬁcients of the Ernst potential, it was, however, necessary to integrate over functions, which had to be determined as solutions of an integral equation. Here we shall provide an explicit scheme, again following the algorithm of Fodor et al. (1989), for calculating all the multipole moments as given functions of the parameter µ. The starting point for the calculation is a power series expansion of the function ∞
1 − f (0, ζ ) # mn g( = 0, ζ ) := ζ = , 1 + f (0, ζ ) ζn
ζ >0
(2.251)
n=0
at inﬁnity, where f (0, ζ ) is the Ernst potential on the axis. The multipole moments Pn = Mn + i Jn
(2.252)
2.3 The rigidly rotating disc of dust
83
can be obtained as functions of the coefﬁcients mn of this expansion. The real parts of the Pn are called the mass moments Mn and the imaginary parts the rotational moments Jn . It turns out that, as in the case of the mn , all Pn with even index are real and the others are purely imaginary. This is a general property of solutions with reﬂectional symmetry (Kordas 1995, Meinel and Neugebauer 1995). Thus we have Ml = 0, l = 1, 3, 5, . . .
and
Jl = 0, l = 0, 2, 4, . . . .
(2.253)
Here Mn n=0 = M is the gravitational mass (not to be confused with the baryonic mass) and J1 = J is the ζ component of the angular momentum. Inserting the Ernst potential f = e2U + i b into (2.251) yields g(x = 0, y) = 0 y
1 − f (0, y) 1 + f (0, y)
1 − (e4U + b2 ) 20 yb = 0 y −i . 2U 2 2 (1 + e ) + b (1 + e2U )2 + b2
(2.254)
Using the dimensionless coefﬁcients m ˜ n := n+1 mn , the power series expansion (2.251) reads ∞ # m ˜n g(0, y) = 0 . (2.255) (0 )n+1 yn n=0
For the Ernst potential on the axis, a slightly different representation from (2.226) will be inserted in (2.254), where again the parameter functions b0 (µ) and 0 (µ) will be used:16 MN 2 + P N (1 − MP ) − i , N 2 + P2 N 2 + P2 1 − 2(1 + y2 + y4 + 2y2 + τ 4 )(0 )2 M= , 2(0 )y − b0 1 − 2(1 + y2 + y4 + 2y2 + τ 4 )(0 )2 P=− . 2(0 )y + b0
f (x = 0, y) =
(2.256)
The function N = N (µ, y) is given in (2.228) and can be expanded as follows: c c3 c5 c7 1 N = exp + 3 + 5 + 7 + ... . (2.257) y y y y 16 This form of the axis potential is discussed by Kleinwächter (2000). Because b is negative, there is a zero in 0
the denominator of P, which is compensated for by the numerator, however.
84
Analytical treatment of limiting cases
Carrying out the expansion of (2.254) we can read off the coefﬁcients m ˜ n (µ) as functions of the parameter b0 , the newly deﬁned 0 := 0
(2.258)
and the above coefﬁcients cn , which all depend only on µ. The expansion of (2.254) leads to ˜2 m m ˜1 m ˜3 m ˜4 g = 0 m + 2 + 3 + 4 + ... , ˜0 + y y y y m ˜ 0 = −b0 − 0 c1 , m ˜ 1 = −i (b0 + 20 c1 ), m ˜ 2 = b0 + (2 + b20 )0 c1 + b0 20 c12 + (30 (c13 − 12c3 ))/3, m ˜ 3 = (i/3)(3b30 + 12b20 0 c1 + 3b0 20 (4 + 3c12 ) + 230 (12c1 + c13 − 12c3 )), m ˜ 4 = b0 − 2b30 + (2 − 7b20 )0 c1 − b0 20 (8 + (5 + b20 )c12 ) − (230 (24c1 + (1 + 2b20 )c13 − 6(2 + b20 )c3 ))/3 + b0 40 ((−2c14 )/3 + 8c1 c3 ) + 50 ((−2c15 )/15 + 4c12 c3 − 16c5 ). (2.259) It is possible to calculate the coefﬁcients cn (µ) up to arbitrary order and hence the same is valid for the quantities m ˜ n . Instead of the multipole moments (2.252), we use the dimensionless and normalized moments ˜ n + iJ˜n ; P˜ n := M
n = 0, 1, 2, . . .
with:
˜ 2l := (−1)l (2)2l+1 M2l , M J˜2l+1 := (−1)l (2)2l+2 J2l+1 ;
(2.260) l = 0, 1, 2, . . . .
It turns out that P˜ 0 = m ˜ 0,
P˜ 1 = m ˜ 1,
P˜ 2 = −m ˜ 2,
P˜ 3 = −m ˜3
(2.261)
holds, thus implying ˜0 = m M ˜ 0,
J˜1 = m ˜ 1,
˜ 2 = −m M ˜ 2,
J˜3 = −m ˜3
(2.262)
(i.e. P˜ n  = m ˜ n , n = 0, 1, 2, 3). But for i > 3 one has the general structure P˜ n  = m ˜ n + fn (m ˜ j : j < n).
(2.263)
2.3 The rigidly rotating disc of dust
85
The exact form of the functions fn up to n = 10 is given in Fodor et al. (1989) and can be calculated with the methods given there up to arbitrary order.17 Summarizing the above we have
P˜ n (µ) = P˜ n b0 (µ), 0 (µ), cj (µ) : j < jmax (n) , (2.264) where jmax (n) refers to the ﬁnite maximal value of j for a given n. Because the coefﬁcients cj can also be obtained as functions of µ for arbitrary j, all multipole moments of the rigidly rotating disc of dust can be calculated explicitly. From the structure of the function g, it follows that all the coefﬁcients m ˜ n with an even index are real and all the others are purely imaginary. This property transfers to the P˜ n . For all µ < µ0 and n > 3, the quantities P˜ n  differ from the absolute values m ˜ n . But in the limit µ → µ0 , for which 0 → 0 and b0 → −1 hold, P˜ i (µ0 ) = m ˜ i (µ0 ) = 1(i = 0, 1, 2, . . .), or more precisely ˜ 2l (µ0 ) = (−1)l m M ˜ 2l (µ0 )
= 1,
(2.265)
˜ 2l+1 (µ0 ) = 1 J˜2l+1 (µ0 ) = (−1)l m
(2.266)
is valid for all l. These are exactly the multipole moments of the extreme Kerr solution (in our normalization).18 We shall come back to this important point in Subsection 2.3.5. The remaining task is to calculate the coefﬁcients cn of (2.257). The function ln N is built up from the functions uA (2.223) and the function Y , which appears as the argument of different elliptic functions (2.228). The expansion of the term involving uA can be expressed with the integrals19 √ 1 µ ln( 1 + x2 + x) xk dx. (2.267) Ik (µ) := √ √ π 0 µ−x 1 + x2 As an illustration, we provide the ﬁrst terms of the expansion 1 1 √ −µ y4 + 2y2 + τ 4 uA (µ, y) = µI0 (µ)y + √ I1 (µ) µ y 1 I0 − 2µI1 + 2I2 + + O . y5 2µ3/2 y3
(2.268)
For the function Y deﬁned in (2.229), Y =
1 K(h ) 1 1 − 2h2 1 − 6h2 + 6h4 + + + + O 2 y y7 3y3 5y2
17 For n = 4, 5, 6 the functions f can be found in Appendix 3. n 18 For a more detailed discussion see Appendix A3.3. 19 Note the identity I (µ) = I (µ). 0
(2.269)
86
Analytical treatment of limiting cases
follows directly. From this series, one can in turn obtain the series expansions of the Jacobian zeta function Z(Y , h ) and the elliptic functions sn(Y , h ), cn(Y , h ) and dn(Y , h ). Finally we calculate the logarithm of the theta quotient Iˆ (µ) + Y + ln TQ := ln π Iˆ (µ) − Y + ϑ3 2K(h) ϑ3
π 2K(h)
K(h ) 2 K(h ) 2
.
) , −π K(h K(h) ) , −π K(h K(h)
(2.270)
Introducing Yˆ := Y − K(h )/2 and using (A2.7), one can write K(h ) π ˆ ˆ I (µ) + Y , −π ln TQ = ln ϑ2 2K(h) K(h) K(h ) π ˆ π ˆ ˆ I (µ) − Y , −π + I (µ). − ln ϑ2 2K(h) K(h) K(h)
(2.271)
Using (A2.29) it is possible to perform the expansion of the last equation. The coefﬁcients of this series consist of the complete elliptic integrals K(h ) and Jacobian elliptic functions with argument Iˆ (µ) and modulus h . Inserting all these series for the components of ln N , we can derive the desired coefﬁcients cn . These coefﬁcients have the structure cn = cn [µ, Ik (µ) : k ≤ kmax (n)] .
(2.272)
In addition to a simple algebraic dependence on µ and Ik (µ) (2.267), cn depends on 4 ˆ elliptic functions and elliptic integrals with main argument I (µ) = 1 + µ2 I0 (µ) and modulus h (µ). The explicit results for c1 , c3 , c5 , c7 , c9 and c11 , which are needed for the ﬁrst eleven multipole moments are given in Appendix 3. The coefﬁcient c1 is especially interesting since it is the only one that enters the formulae for the ﬁrst two moments yielding the gravitational mass and the angular momentum of the disc. This coefﬁcient reads 1 . E(h) c1 = √ 2 1 + µ2 I (µ) h2 − + I1 (µ) µ K(h) / π 4 4 + 1 + µ2 0 (am( 1 + µ2 I (µ), h ), h) , K(h)
(2.273)
which can be simpliﬁed using ψ1 := am(Iˆ (µ), h ) 1 c1 (µ) = √ µ
! √ ( 2 2ˆ E(ψ1 , h ) − h I (µ) + I1 (µ) . √ hh
(2.274)
2.3 The rigidly rotating disc of dust
87
Thus for the normalized gravitational mass M and ζ component of the angular momentum 2 J we have, 1
c1 (µ)0 (µ) + b0 (µ) , 2 1 1 2 J (µ) = − c1 (µ)0 (µ) + b0 (µ) . 2 2 M (µ) = −
(2.275) (2.276)
From the relation (2.237) between M , 2 J and M0 , one immediately ﬁnds M0 (µ) =
1 2µc1 (µ). 4
(2.277)
On the other hand we have the result (2.248) for M0 which was obtained by integrating over the proper surface massdensity. This means that the identity b (µ) d 2µc1 (µ) = − V0 (2.278) dµ e 0 (µ) must hold, which can easily be veriﬁed by differentiating (2.274) with the help of (A2.26) to provide an expression for the left hand side. The right hand side can be found by differentiating b0 , taking into account (A2.27) and (A2.28). Conversely, we have thus solved for the integral in (2.248) and have proved that the results that were obtained in connection with the multipole expansion (2.275), (2.276) and (2.277) coincide exactly with the results (2.248), (2.249) and (2.250) calculated by integrating over the source. The relative binding energy is M0 − M 2b0 (µ) (µ) = 1 + eV0 (µ) + √ . M0 2µc1 (µ)
(2.279)
For µ = µ0 , one ﬁnds 0 = 0, b0 = −1 and c1 = 1.0487435 . . . Therefore, in the extreme relativistic limit, we have the values 1 M (µ0 ) = , 2
2 J (µ0 ) =
1 4
(2.280)
and M0 (µ0 ) = 0.797809010 . . . ,
M0 − M (µ0 ) = 0.373283588 . . . (2.281) M0
The normalized mass M (2.275), angular momentum 2 J (2.276) and baryonic mass M0 (2.277) are plotted as functions of µ in Fig. 2.11, and the graphs of the relative binding energy (2.279) and of the ratio M 2 /J can be found in Fig. 2.12. The ﬁrst six mass multipole moments are plotted in Fig. 2.13 and the ﬁrst ﬁve rotational moments are shown in Fig. 2.14. The complete corresponding formulae are given in Appendix 3.
88
Analytical treatment of limiting cases
Fig. 2.11. The normalized baryonic mass M0 , gravitational mass M and angular momentum 2 J are given as functions of µ (after Neugebauer et al. 1996).
Fig. 2.12. Relative binding energy and M 2 /J (after Neugebauer et al. 1996).
The ergosphere For sufﬁciently large values of µ, the rigidly rotating disc of dust possesses an ergosphere, i.e. a region in which no static observer (seen from inﬁnity) can reside, see Subsection 1.6.2. Within this region dϕ/dt > 0 must hold for any timelike worldline. The ergosphere is characterized by ξi ξ i = g44 > 0,
(2.282)
2.3 The rigidly rotating disc of dust
89
˜ 0, M ˜ 2, M ˜ 4, M ˜ 6, M ˜ 8 and M ˜ 10 Fig. 2.13. The normalized mass multipole moments M (after Kleinwächter et al. 1995).
Fig. 2.14. The normalized rotational multipole moments J˜1 , J˜3 , J˜5 , J˜7 and J˜9 (after Kleinwächter et al. 1995).
meaning that the Killing vector ξi of stationarity [normalized by (1.3)] becomes spacelike there. The occurrence of an ergosphere is observed for µ > µe = 1.68849467 . . .
(2.283)
90
Analytical treatment of limiting cases
For µ = µe , we ﬁnd a ring at xe = (/0 )e = 0.7617018 . . . , where g44 = 0 holds. Values for both µe and xe can be obtained simultaneously by searching for the ﬁrst zero e2U (µe , xe ) = 0 using (2.203). From the last line of (2.212) it is obvious that the rim of the disc belongs to the ergosphere for µ > 2. For µ → µ0 , the shape of the ergosphere of the extreme Kerr solution is approached. The general toruslike shape can be obtained by searching for the zeros of the numerators of e2U in (2.185) and (2.186). Five characteristic shapes of the ergosphere are shown in Fig. 2.15.
Fig. 2.15. Ergospheres for different values of µ. An ergosphere appears for µ > µe = 1.68849 . . . and reaches the rim of the disc at µ = 2. For µ → µ0 the ergosphere of the extreme Kerr solution (with M = 1/2) is approached. The horizontal lines represent the disc. Its coordinate radius 0 vanishes in the limit µ → µ0 , see Subsection 2.3.5 (after Meinel and Kleinwächter 1995).
2.3 The rigidly rotating disc of dust
91
Further characteristic relativistic effects We start by considering circular geodesic motion at the rim of the disc, which can be solved particularly elegantly. For x = 1, we ﬁnd from (2.210) the expressions for the required metric coefﬁcients and their derivatives: 2 gϕϕ = −(1 − eV0 )2 + 2 gϕϕ,x = µ(2eV0 − 1),
µ µ µ , gϕt = 1 − eV0 − , gtt = −1 + , 2 2 2 V0 gϕt,x = µ(1 − e ), gtt,x = −µ,
(2.284) (2.285)
where eV0 (µ) is given by (2.209). From (2.284) we immediately obtain the angular velocity ω, as seen from inﬁnity, of the socalled locally nonrotating observer20 ω=−
gϕt µ − 2(1 − eV0 ) = . gϕϕ µ − 2(1 − eV0 )2
(2.286)
The value of ω approaches in the extreme relativistic limit µ → µ0 (i.e. eV0 → 0). This property holds not only at the rim of the disc, but throughout the whole disc as we can see by rewriting (2.210)
˜ µx2 + 2eV0 (µ) ˜ eV0 (µ) − eV0 (µ) ω(µ, x) = (2.287)
2 . µx2 − 2 eV0 (µ) − eV0 (µ) ˜ We have to distinguish between corotating particles (direct orbits: + > 0) and counterrotating particles (retrograde orbits: − < 0). Of course, + = , since the dust particles of the disc move on circular geodesics themselves. For − the result is − = − V . (2.288) 2e 0 − 1 i and the speciﬁc angular momenta The corresponding speciﬁc energies E± = −ξi u± i i L± = ηi u± (where u± is the corresponding 4velocity) turn out to be E+ = 1, 1−µ E− = √ , 1 − 2µ
1 (1 − eV0 ), 1 µ − 1 + eV0 L− = − . √ 1 − 2µ L+ =
(2.289) (2.290)
For the linear velocity of rotation v± measured in the locally nonrotating frame of reference we get from (1.32)
2 2 µ − 1 + eV0 v+ = (1 − eV0 ), v− = − , (2.291) µ µ 2eV0 − 1 20 See Section 1.4, especially footnote 6.
92
Analytical treatment of limiting cases
Fig. 2.16. The linear velocities of rotation v± of test particles moving on circular geodesic orbits at the rim of the disc measured in the locally nonrotating frame of reference. Retrograde motion (v− ) is possible only for µ < 1/2 (after Meinel and Kleinwächter 1995).
a plot of which can be found in Fig. 2.16. It turns out that for µ ≥ 1/2 no geodesic retrograde motion is possible. From (2.290) it is obvious that E− and L− become inﬁnite and (2.291) shows that v−  approaches 1, i.e. the speed of light in our units, for µ = 1/2. As we already pointed out, for µ ≥ 2 the rim of the disc belongs to the ergosphere, meaning that for such µ, no motion at all of a particle or observer in the retrograde direction is possible there. Figure 2.17 combines the parameter relations between M and M 2 /J for the rigidly rotating disc of dust as well as for the Kerr black hole (1.136). The two branches meet at the point M 2 /J = 1. In this ﬁgure, the corresponding relation for the Newtonian disc, which follows from (2.24) and (2.26), is also plotted. We see right away that in the extremely relativistic regime, the curve for the relativistic disc of dust differs from its Newtonian counterpart signiﬁcantly. We observe the parametric transition to the Kerr black hole for M 2 /J = 1 (µ → µ0 ), see Subsection 2.3.5. The Einsteinian disc possesses characteristic relativistic features even in regions where the curves for both discs differ by only very little. For M 2 /J (µ = 1/2) = 0.50457 . . ., where as discussed no geodesic retrograde motion at the rim is possible, both curves nearly coincide and even for M 2 /J (µ = 2) = 0.86656 . . ., where no retrograde motion at all is possible, one ﬁnds only a moderate deviation.
2.3 The rigidly rotating disc of dust
93
Fig. 2.17. Relation between M and M 2 /J for the classical Maclaurin disc (dashed line), the general relativistic disc of dust and the Kerr black hole. In the latter case has been identiﬁed with the ‘angular velocity’ h of the horizon, see Subsection 1.8.2 (after Neugebauer and Meinel 1993).
The Newtonian limit Of course, for all functions discussed here, such as the parameter functions e2V0 , b0 , 0 ≡ 0 or the mass M and angular momentum J as well as the higher multipole moments, a Taylor expansion around the Newtonian limit can be given. For example, from (2.209) and (2.204) we can derive 1 8 4 µ µ2 5 3 e (µ) = 1 − + + − + µ + − µ4 + O(µ5 ), 2 8 16 9π 2 128 9π 2 √ 128 512 11 µ µ3/2 µ5/2 (9− π 2 )µ7/2 (−45 + π 2 )µ9/2 0 (µ) = √ − √ + √ + + +O µ 2 √ √ 2 2 2 8 2 144 2 1152 2 (2.292) V0
and (2.275), (2.276), (2.277) lead to √ 3/2 11 µ5/2 2µ 19µ7/2 (−640 + 33π 2 )µ9/2 M (µ) = − √ +O µ2 , + + √ √ 3π 5 2π 420 2π 3240 2π 3 √ 3/2 √ 5/2 √ 11 µ7/2 2µ 2µ 16 2(−35 + 3π 2 )µ9/2 2 , − + √ + µ + O M0 (µ) = 3π 15π 2835π 3 35 2π √ 5/2 11 µ7/2 2µ (896 − 39π 2 )µ9/2 2 − √ (2.293) +O µ2 . J (µ) = + √ 15π 15 2π 2268 2π 3
94
Analytical treatment of limiting cases
In order to compare these with the corresponding Newtonian formulae it is instructive to invert the series for 0 :21 256 80 + 40 + 9π 2
µ(0 ) =
220
+ 440
+ 1260
2560 12 10 + 140 + 0 + O (0 ) 2 9π
(2.294)
and insert this result in the above series M (0 ) =
430 1650 108870 256(40 + 57π 2 )90 + O(11 + + + 0 ), 3π 5π 105π 405π 3
M0 (0 ) =
430 5250 121470 2(31360 + 58791π 2 )90 + + O(11 + + 0 ), 105π 3π 15π 2835π 3
2 J (0 ) =
850 3270 128(140 + 177π 2 )90 + + + O(11 0 ). 3 15π 15π 2835π
(2.295)
Comparing with (2.24) for M and with (2.26) for J (remembering 0 = 0 ), we observe that the leading terms coincide with the expressions for the Maclaurin disc. In the same way, the proper surfacemass density √ √ 2 1 − x2 0 4 1 − x2 (x2 + 2)30 σp (0 , x) = + π2 3π 2 √ 4 1 − x2 (3x4 + 14x2 + 28)50 + O(70 ) + 15π 2
(2.296)
can be seen to coincide with (2.23) to leading order by taking (2.24) and x = /0 into account. In 4 16 −3 − 2 60 9 π 512 80 + O(10 + −16 − 0 ) 2 9π
V0 (0 ) = −20 − 240 +
(2.297)
21 Note that is not a monotonic function in µ (see Fig. 2.9). Hence an inversion is not possible in the whole 0 range (0 ≤ µ ≤ µ0 ). Near the Newtonian limit (µ 1), this presents no problem, however.
2.3 The rigidly rotating disc of dust
95
the parameter relation (2.25) of the Maclaurin disc is also found to leading order. Finally, we consider the power series expansion for the relative binding energy
3809 8 − µ3 + O(µ4 ), 126000 27π 2 20 1940 21449 64 M0 − M + + − 60 + O(80 ). (0 ) = M0 5 50 15750 27π 2 µ2 µ M0 − M − + (µ) = M0 10 200
(2.298)
The leading term of the last series gives the Newtonian value −(Erot + Eg )/M as follows from (2.27) and (2.24). An impression of the accuracy of the above formulae can be gained from Fig. 2.18, in which the expressions (2.293) are compared with the corresponding exact curves for the range 0 ≤ µ ≤ 3. A systematic postNewtonian expansion of the whole solution is given in Petroff and Meinel (2001). There, as in Bardeen and Wagoner (1971), Padé approximants were used to obtain a more accurate representation of the physical parameters from their truncated series representations. Motion of test particles In the context of the study of physical spacetime properties, it is natural to investigate the relativistic motion of test particles in the vicinity of the gravitational source. Here we concentrate on circular orbits within the plane of the disc, centred about
Fig. 2.18. The postNewtonian expansions of M , M0 and 2 J up to the order given in (2.293), as compared to the exact curves (dashed lines), cf. Fig. 2.11.
96
Analytical treatment of limiting cases
the rotational axis.22 One ﬁnds that there is always a (stable or unstable) circular orbit for positive angular momentum and a given radius. However, for sufﬁciently relativistic discs (the more precise condition follows below), there are regions within the plane of the disc in which a particle with negative angular momentum cannot follow a circular path. If the disc is still more strongly relativistic, then one ﬁnds circular orbits with negative energies of arbitrary magnitude. We start the study of geodesic motion by considering the corresponding Hamiltonian system, which reads 1 H = g ij pi pj 2
with pi = gij x˙ , j
d , ˙= dτ
τ : proper time .
(2.299)
For the timelike geodesic equations i k l x˙ x˙ = 0, x¨ i + kl
x˙ k x˙ k = −1,
(2.300)
this system can be reduced, for axisymmetry and stationarity, to a conservative Hamiltonian of two degrees of freedom of the form H=
1 1 2(U −k) 2 1 e 1 − 2 (L2 gtt + 2LEgϕt + E 2 gϕϕ ) . ( p + pζ2 ) + 2 2
(2.301)
The constants of motion L and E have the same meaning as they did a few pages ago, i.e. L = pϕ = constant is the (ζ component of the) speciﬁc angular momentum of the particle and E = −pt = constant its speciﬁc energy. The above Hamiltonian is invariant under a simultaneous change of the signs of L and E. We ﬁx these signs by the condition ˙t > 0 and this condition is then satisﬁed along the whole trajectory. The Hamiltonian (2.301) is chosen such that H = 0 holds along the trajectory of the test particle being considered. Therefore a restriction for the region in which the motion takes place is given by 2(U −k)
0≤e
( p2
+ pζ2 )
1 2 2 = (L gtt + 2LEgϕt + E gϕϕ ) − 1 . 2
(2.302)
22 For a more general presentation of the associated Hamiltonian mechanics, see Ansorg (1998). The stochastic
behaviour of the geodesics turns out to be related to the position of the region containing all the crossing points of the particle through the plane of the disc. If this region contains points lying inside the disc as well as points outside, the geodesic motion shows highly stochastic behaviour. However, if the crossing region is completely inside or outside the disc, the motion proves to be nearly integrable. In such cases, the corresponding Hamiltonian system is close to an integrable system of the socalled Liouville class, see e.g. Contopoulos (1994).
2.3 The rigidly rotating disc of dust
97
Fig. 2.19. Example of geodesic motion in the (, ζ )space within the gravitational ﬁeld of a disc with parameter µ = 3 and radius 0 . Speciﬁc angular momentum and energy are chosen to be L = 2.78, E = 0.77. The bounded area within which the motion takes place, characterized by Equation (2.302), is shown (after Ansorg 1998).
This region of motion shows the following properties (for an illustrative example see Fig. 2.19): (i) The region is bounded for E 2 < 1. For E 2 ≥ 1 there are unbounded areas allowing an escape of the test particle to inﬁnity. However, even in the case E 2 ≥ 1, there may be additional, bounded regions in which the particle is trapped. (ii) Particles with L = 0 cannot cross the rotational axis. (iii) The momenta p and pζ as well as the velocities ˙ and ζ˙ vanish at the rim of the region under consideration. Furthermore, at the boundary of the region, the acceleration vector (, ¨ ζ¨ ) is perpendicular to the curve of the boundary and points inwards. (iv) Because of the property gij (, −ζ ) = gij (, ζ ), the regions of motion are symmetric with respect to reﬂection through the plane of the disc.
Motions within the plane of the disc are described by a Hamiltonian with merely one degree of freedom. Due to the conditions ˙ = 0 = , ¨ circular orbits with centres on the rotation axis are ﬁxed. These lead to the equations L2 gtt (, 0) + 2LEgϕt (, 0) + E 2 gϕϕ (, 0) = 2
(2.303)
98
Analytical treatment of limiting cases
and L2 gtt, (, 0) + 2LEgϕt, (, 0) + E 2 gϕϕ, (, 0) = 2,
(2.304)
which serve to determine the parameter functions L± = L± () and E± = E± (). For a circular orbit of given radius , these functions yield the associated pairs of angular momentum and energy. In general, there are two such pairs, denoted by (L+ (), E+ ()) and (L− (), E− ()) where (L+ , E+ ) refers to an orbit of positive angular momentum and similarly (L− , E− ) to an orbit with negative angular momentum. It turns out that the functions L+ () and E+ () exist for an arbitrary choice of ≥ 0. However, for sufﬁciently large values of the parameter µ, there are no corresponding (L− , E− )pairs for values within a certain interval (˜ 1 , ˜ 2 ). In Figs 2.20–2.24, the functions L± = L± () and E± = E± () are displayed for different values of the parameter µ. A discussion of these pictures follows. (i) The pictures for (L+ , E+ ) are similar for all values of µ, see Fig. 2.20 for a particular example. There is a monotonic growth of the functions L+ and E+ for < 0 , where 0 is the radius of the disc. At = 0 , where E+ = 1, both functions turn back and decrease monotonically until they reach another turning point at = . For still greater values of the parameter , the functions L+ and E+ grow monotonically again.
Fig. 2.20. For prograde circular orbits, the corresponding parameter curve (L+ (), E+ ()) is displayed for a disc with µ = 1.61 (after Ansorg 1998).
2.3 The rigidly rotating disc of dust
Fig. 2.21. For retrograde circular orbits, the corresponding parameter curve (L− (), E− ()) is displayed for a disc with µ = 0.01 (after Ansorg 1998).
Fig. 2.22. For retrograde circular orbits, the corresponding parameter curve (L− (), E− ()) is displayed for a disc with µ = 0.4 (after Ansorg 1998).
99
100
Analytical treatment of limiting cases
Fig. 2.23. For retrograde circular orbits, the corresponding parameter curve (L− (), E− ()) is displayed for a disc with µ = 1.21 (after Ansorg 1998). The turning points are characterized by the conditions dL+ /d = 0 = dE+ /d. As the radius → ∞, E+ tends to 1 and L+ to +∞. Circular orbits with a radius ∈ (0 , ) are unstable. Stable circular orbits are those with a radius ∈ [0, 0 ) or ∈ ( , ∞); the remaining circular orbits with radius 0 or are marginally unstable. For circular orbits, the condition ϕ˙ =
dE ˙t dL
(2.305)
holds. In all sections of the graphs of the functions L+ and E+ , the function E+ = E+ (L+ ) grows monotonically. Hence, all circular orbits with positive angular momenta possess positive angular velocities. Furthermore, circular orbits with radii < 0 are precisely those paths along which the dust particles of the disc move. Their 4velocity ui is given by (ui ) = (0, 0, ϕ, ˙ ˙t ) = e−V0 (0, 0, , 1), cf. (1.18), (1.105), from which dE/dL = = constant follows. Thus, the graph E+ (L+ ) is a straight line for ∈ [0, 0 ]. The region A contains all (L, E)pairs for which no motion of a particle is possible, i.e. either the restriction (2.302) cannot be met or the motion possesses a negative ˙t . If one chooses a (not necessarily circular) motion corresponding to an (L, E)pair inside the small region B, then there exist two separate compact regions in which bounded motions are possible. At the intersection point of the stable and unstable parts of the
2.3 The rigidly rotating disc of dust
101
Fig. 2.24. For retrograde circular orbits, the corresponding parameter curve (L− (), E− ()) is displayed for a disc with µ = 3.5 (after Ansorg 1998).
(L+ , E+ )graphs, these regions degenerate to become two separate stable circular orbits with radii 1∗ < 0 and 2∗ > 0 . (ii) In Figs 2.21–2.24, the parameter functions L− () and E− () are displayed for different values of the parameter µ. For 0 < µ < 1/2, the pictures are similar to the graphs of the functions L+ and E+ . The region B now extends to E− values greater than 1. Thus, there are stable as well as unstable circular orbits and, furthermore, extended compact regions of motion with energies greater than 1 and corresponding negative angular momenta. From Equation (2.305) and the slopes of the functions E− = E− (L− ), see Figs 2.21 and 2.22 for two examples, one concludes that all particles moving along circular orbits with negative angular momenta have negative angular velocities. (iii) As µ approaches the value 12 , the region B grows and is unbounded for µ ≥ 12 . For radii ∈ (˜ 1 , ˜ 2 ) there are no circular orbits with negative angular momentum. As can be seen in Fig. 2.23, for sufﬁciently large values of µ (µ > µϕ˙ ≈ 0.7088) there are small intervals [0, ϕ˙ ) in which the functions E− (L− ) grow monotonically. Hence, circular orbits with radii < ϕ˙ possess positive angular velocities in spite of their having negative angular momenta, cf. page 171. At the radius ϕ˙ , a particle may remain at rest (in the chosen coordinate system). Such a particle has the smallest energy possible of all motions around the disc.
102
Analytical treatment of limiting cases
(iv) Finally, in the range µe < µ < µ0 , there are stable circular orbits with negative energies, see Fig. 2.24.23 As the slope of E− (L− ) is positive for < ˜ 1 , circular orbits with radii in this range have positive angular velocities. (v) As µ → 0, the curves L± () and E± () tend to their corresponding counterparts for the Maclaurin disc. Since negative and positive angular momenta are equivalent in Newtonian theory, these graphs possess a reﬂectional symmetry with respect to the axis L = 0.
For the functions L+ and E+ , the radii 1∗ , 2∗ and , as they depend on the parameter µ, can be seen in Fig. 2.25. Likewise, the µdependencies of the radii {1∗ , 2∗ , , ˜ 1 , ˜ 2 , ϕ˙ } for the functions L− and E− are displayed in Fig. 2.26. (E) (E) In addition, the radii 1 and 2 at which the function E− = E− () reaches (E) the value E− (i ) = 1 can be seen.24 For µ < 21 , these radii are located in the
Fig. 2.25. The µdependency of the radii 1∗ , 2∗ and for prograde circular orbits, described by the angular momentumenergy pair (L+ , E+ ). For each disc parameter ∗ (µ)) for which two separate stable µ ∈ (0, µ0 ) there is a unique pair (L∗+ (µ), E+ ∗ ∗ circular orbits of radii 1 and 2 exist. Circular orbits with radii ∈ (0 , ), and only such orbits, are unstable (after Ansorg 1998). 23 For µ > µ ≈ 1.69, the disc is sufﬁciently relativistic so as to produce an ergosphere, cf. Fig. 2.15 and (2.283). e 24 Note that E ( ) < 1 always holds. −
2.3 The rigidly rotating disc of dust
103
(E)
(E)
Fig. 2.26. The µdependency of the radii {1∗ , 2∗ , , ˜ 1 , ˜ 2 , ϕ˙ , 1 , 2 } for retrograde circular orbits, described by the angular momentumenergy pair (L− , E− ). For each disc parameter µ ∈ (0, µe ), there is a unique pair ∗ (µ)) for which two separate stable circular orbits at the radii ∗ and (L∗− (µ), E− 1 ∗ 2 exist. Circular orbits with radii ∈ (0 , ), and only such orbits, are unstable. (E) For µ > 12 , no circular orbits exist with radii ∈ (˜ 1 , ˜ 2 ). At the radii 1/2 , the corresponding circular orbit is characterized by the energy E− = 1. For the circular orbit at the radius ϕ˙ , the angular velocity ϕ˙ vanishes, i.e. the test particle remains at rest (in the chosen coordinate system). Additionally, the radii (e) (e) 1 and 2 at which the ergosphere intersects the plane of the disc are plotted (after Ansorg 1998).
intervals (E)
< 0 ,
0 < 2
(E)
< ˜ 1 ,
˜ 2 < 2
1∗ < 1 whereas for
1 2
(E)
< ,
(E)
<
< µ < µe , 1∗ < 1
(E)
hold. In the range µe < µ < µ0 , the radius 2 is still within (˜ 2 , ) whereas the (E) radii 1∗ , 2∗ and 1 cannot be deﬁned since the curve (E− (), L− ()) does not possess an intersection point.
104
Analytical treatment of limiting cases
Circular orbits around a Kerr black hole are characterized by the functions, see e.g. Bardeen (1973b), r 3/2 − 2Mr 1/2 ± JM −1/2 , √ r 3/4 r 3/2 − 3Mr 1/2 ± 2JM −1/2 √ M 1/2 [r 2 ∓ 2J r/M + J 2 M −2 ] L± = ± , √ r 3/4 r 3/2 − 3Mr 1/2 ± 2JM −1/2
E± =
(2.306) (2.307)
√ r 2 − 2Mr + J 2 M −2 , M denotes the mass and J the angular where = momentum of the black hole (see Subsection 2.4). For a vanishing denominator in the above expressions, when = ˜ say, L± → ±∞ and E± → +∞. Circular orbits with radii smaller than ˜ do not exist. For > , ˜ the (L± , E± )curves are similar to those of Fig. 2.24 for > ˜ 2 . Again there is a that separates unstable (for < ) and stable orbits (for > ). Furthermore, for circular orbits within the Kerr metric, the signs of the angular momentum and angular velocity always coincide. Generally, one ﬁnds that the qualitative behaviour of circular motions at sufﬁciently large radii is similar for the Kerr black hole and the disc. However, circular motions at small radii are quite different.
2.3.5 Black hole limit It will turn out that 0 → 0 in the black hole limit. Therefore, we ﬁrst return to the nonnormalized quantities K, Ka/b and K1/2 , cf. (2.99), and introduce H≡
µ h, 02
Z1 ≡ 0 W1 ,
03 W µ
(2.308)
v2 ≡ µw.
(2.309)
Z≡
together with v0 ≡
µ u, 02
v1 ≡
µ v, 0
With these expressions, the Ernst potential of the disc solution represented in the form (2.100), (2.101) reads
f = exp
Ka K 2 dK
K1
Z
Kb + K2
K 2 dK − v2 Z
(2.310)
2.3 The rigidly rotating disc of dust
with
(K + i z)(K − i¯z )(K 2 − K12 )(K 2 − K22 ),
i−µ ( K1 < 0). K1 = −K¯ 2 = 0 µ Z=
105
(2.311) (2.312)
The upper limits of integration Ka and Kb in (2.310) have to be calculated from Ka Kb Ka Kb dK dK KdK KdK (2.313) + = v0 , + = v1 Z Z Z Z K1
K2
K1
K2
with +i 0
v0 = −i0
H dK, Z1
+i 0
v1 = −i0
H KdK, Z1
+i 0
v2 = −i0
H 2 K dK, Z1
2 2 2 2 2 2 1 + µ (1 + K /0 ) + µ(1 + K /0 ) µ ln ( H = 0), H= 2 2 2 2 2 πi 0 1 + µ (1 + K /0 ) Z1 = (K + i z)(K − i¯z ) ( Z1 < 0 for , ζ outside the disc).
(2.314)
(2.315) (2.316)
The solution depends on the two parameters 0 and µ. The original parameters V0 and of the boundary value problem are related to 0 and µ as follows: V0 ≡ U ( = 0, ζ = 0) depends on µ alone. Using (2.118), it can be expressed as ( ! 1 + µ2 1 , (2.317) V0 = − arcsinh µ + 2 ℘[I (µ); 43 µ2 − 4, 83 µ(1 + µ2 /9)] − 32 µ 1 I (µ) = π
µ 0
√ ln(x + 1 + x2 )dx (1 + x2 )(µ − x)
(2.318)
with the Weierstrass function ℘ deﬁned by ∞ ℘ (x;g2 ,g3 )
dt 4t 3 − g2 t − g3
= x.
(2.319)
Note that Equation (2.317) is equivalent to (2.201). The range 0 < µ < µ0 = 4.62966184 . . . corresponds to 0 > V0 > −∞ [µ0 is the ﬁrst zero of the
106
Analytical treatment of limiting cases
denominator in (2.317)]. Recalling the deﬁnition of µ in (2.79), µ = 22 02 e−2V0 ,
(2.320)
one can then ﬁnd the relation = (µ, 0 ). The limit µ → µ0 As was shown in Subsection 2.3.4, one obtains M2 →1 J
2M → 1
and
(2.321)
in the limit µ → µ0 , i.e. V0 → −∞. From (2.320) we ﬁnd 0 → 0,
(2.322)
0 → 0.
(2.323)
which means that for ﬁnite M , Note that remains a free parameter in the limit. For ease of discussion, we introduce sphericallike coordinates R, ϑ, = R sin ϑ,
ζ = R cos ϑ,
0 ≤ ϑ ≤ π.
(2.324)
The disc (ζ = 0, ≤ 0 ) shrinks to the origin R = 0 of the coordinate system. For R = 0, the disc metric becomes exactly the r > M part of the extreme Kerr metric with the radial Boyer–Lindquist coordinate r related to R by r = R + M.
(2.325)
This can be shown as follows (Meinel 2002). Let us ﬁrst rewrite (2.310) and (2.313) in the equivalent form Ka Ka Ka K 2 dK dK KdK f = exp − v˜2 , = v˜0 , = v˜1 , (2.326) Z Z Z Kb
Kb
with
K2 v˜n = vn −
Kb
K n dK Z
(n = 0, 1, 2).
(2.327)
K1
(Kb is now on the other sheet of the Riemann surface.) In the limit µ → µ0 , one obtains for R > 0, using (2.320) and (2.317), v˜0 =
2 πi cos ϑ − , R 2R2
v˜1 = −
πi , 2R
v˜2 = 0
(2.328)
2.3 The rigidly rotating disc of dust
107
(modulo periods). In the above integrals from Kb to Ka , Z can be replaced by √ Z = K 2 (K + i z)(K − i¯z ) since K1 and K2 both tend to zero, cf. (2.312). Hence, all integrals become elementary and the unique result is f =
2R − 1 − i cos ϑ 2R + 1 − i cos ϑ
(R > 0),
(2.329)
which is the Ernst potential of the extreme Kerr solution. Note that R = 0 (r = M ) characterizes the horizon (and the throat) of the extreme Kerr black hole. = h = 1/(2M ) is the ‘angular velocity of the horizon’, see (1.149). A completely different limit of the spacetime for µ → µ0 is obtained for ﬁnite values of R/0 (corresponding to the previously excluded R = 0). Therefore, we consider a coordinate transformation (Bardeen and Wagoner 1971, Meinel 2002) r˜ = e−V0 R,
ϑ˜ = ϑ,
ϕ˜ = ϕ − t,
˜t = eV0 t.
(2.330)
Note that ﬁnite R/0 correspond to ﬁnite r˜ in the limit, as can be seen from (2.320). For µ < µ0 , this is nothing other than the transformation to the corotating system combined with a rescaling of R and t. The transformed Ernst potential f˜ is related to the Ernst potential f in the corotating system (R = R, ϑ = ϑ, ϕ = ϕ −t, t = t) according to f˜ = f exp(−2V0 ), i.e. f˜ f = r˜ 2 R2
for
µ < µ0 .
(2.331)
However, for µ → µ0 , the solutions f (ﬁnite R > 0) and f˜ (ﬁnite r˜ ) separate from each other. For ﬁnite R > 0, the extreme Kerr solution arises,25 while ﬁnite r˜ values lead to a solution that still describes a disc, whose ﬁnite coordinate radius is ˜ 0 = lim (e−V0 0 ) = 2µ0 M . (2.332) µ→µ0
Note that the proper radius of the disc remains ﬁnite in the limit µ → µ0 as well. Its √ circumference is 4πM µ0 /2 − 1, which is larger26 than the circumference 4πM of the extreme Kerr throat. The metric corresponding to f˜ , which can be expressed in terms of theta functions, is regular everywhere outside the disc, but it is not asymptotically ﬂat. The spacetime structure of both solutions ( f and f˜ ) coincides at R → 0 (the throat) and r˜ → ∞ (spatial inﬁnity). The relation (2.331) survives in the form f˜ f lim 2 = lim 2 as µ → µ0 . (2.333) R→0 R r˜ →∞ r˜ 25 Accordingly, in the limit µ → µ , all gravitational multipole moments assume the extreme Kerr values, cf. 0
Subsection 2.3.4.
26 An analogous situation will be found for ﬂuid rings approaching the black hole limit, see Fig. 3.22.
108
Analytical treatment of limiting cases
[The limits have to be taken consistently with (2.330).] The Ernst potential f of the extreme Kerr solution in the corotating system reads
f = − R
2 2
2(1 + i cos ϑ)2 2 + sin ϑ . 2R + 1 − i cos ϑ
Accordingly, for µ = µ0 and r˜ → ∞, 2 1 2 ˜ 2(1 + i cos ϑ) f˜ → f˜as = −2 r˜ 2 + sin2 ϑ˜ . 1 − i cos ϑ˜
(2.334)
(2.335)
Note that f˜as belongs to the family of solutions to the Ernst equation of the type f = Rk Yk (cos ϑ) presented by Ernst (1977). The corresponding asymptotic line element is given by the ‘extreme Kerr throat geometry’ or ‘nearhorizon geometry’ (1.150). The analytic solution to the disc of dust problem thus provides an explicit example for a parametric transition from a normal matter equilibrium conﬁguration to a black hole as discussed in Subsection 1.8.3. Further examples, obtained by numerical means, will be presented in Chapter 3. 2.4 The Kerr metric as the solution to a boundary value problem In this section, we want to show that the ‘inverse method’ has another remarkable application: the derivation of the Kerr metric as the unique solution to a welldeﬁned boundary value problem (Neugebauer 2000, Neugebauer and Meinel 2003). To obtain the stationary and axisymmetric, asymptotically ﬂat solution describing the vacuum exterior of a black hole, we have to solve the following boundary value problem of the Ernst equation (see Subsection 1.8.1 and Fig. 2.27): On the horizon
Fig. 2.27. The boundary value problem for a black hole, and the line of integration.
2.4 The Kerr metric as the solution to a boundary value problem
109
H, which in canonical Weyl coordinates covers the domain
= 0,
H:
K1 ≥ ζ ≥ K2 ,
(2.336)
the real part e2V of the Ernst potential f in the corotating frame (rotating with the angular velocity = h of the horizon) has to vanish:27 e2V ≡ −χ i χi = 0,
H:
(2.337)
whereas the Ernst potential f in the nonrotating frame has to be regular everywhere outside the horizon and to satisfy f → 1 as Because of
2 + ζ 2 → ∞.
e2V = e2U (1 + a)2 − 2 2 e−4U ,
(2.338)
(2.339)
cf. (2.238), this implies28 H:
1 + a = 0.
(2.340)
Remember that a vanishes on the regular parts of the axis, A± :
a = 0,
(2.341)
i.e. a is not continuous at = 0, ζ = K1/2 . However the metric coefﬁcients, e.g. gϕt = −ae2U , are continuous (e2U vanishes at = 0, ζ = K1/2 ). It should be mentioned in this connection, that our derivation of the Kerr solution makes use of the assumption = 0. However, the ﬁnal result will contain the Schwarzschild solution (corresponding to = 0) as a limiting case. As a ﬁrst step, we calculate and along the horizon H. From (2.41), (2.336), (2.337), (2.340) and (2.56) we obtain U (K) V (K) f (ζ ) 1 = , W (K) X (K) f (ζ ) −1 H: (2.342) −1 0 U (K) V (K) = 2i(K − ζ ) . 1 0 W (K) X (K) The Ernst equations have to hold at K1 and K2 too. Hence, and must be continuous at K1 and K2 . Considering (2.57), (2.58), (2.59), (2.60) and (2.342), we 27 It turns out that the imaginary part of f (up to an arbitrary constant) vanishes on the horizon as well. 28 Note that e2U = 0 holds on the horizon of rotating black holes except at the poles, since e2U = −2 ηi η
follows from (1.125)–(1.127).
i
110
Analytical treatment of limiting cases
are led to the conditions
f1
−1
f1 + 2i(K − K1 ) f1
−1
2i(K − K1 )
0
−1
f2
−1
f2 + 2i(K − K2 ) f2
−1
2i(K − K2 )
0
−1
F
0
G
1
U
V
W
X
1
G
0
F
U
V
W
X
= , (2.343)
= ,
where f1 = f (ζ = K1 ) and f2 = f (ζ = K2 ). Note that f1 and f2 are purely imaginary. Eliminating the UVWX matrix, we obtain F −G N ≡ (2.344) G (1 − G 2 )/F F2 F1 1+ , (2.345) = 1+ 2i(K − K1 ) 2i(K − K2 ) where F1 =
−f1 −f12
1 , f1
F2 =
f2 f22
−1 . −f2
(2.346)
tothe previously introduced matrix M, see (2.72), Note that thematrix N is related −1 0 0 −1 by M = . Obviously, the elements of N are regular N 0 1 1 0 everywhere in the complex Kplane with the exception of the two simple poles at K1 and K2 (K1 = 0 = K2 ). The sum of the offdiagonal elements in (2.344) must be zero. This requirement leads to the constraints f1 = −f2 ,
=
i f1 (1 + f12 ) (K1 − K2 )(1 − f12 )
.
(2.347)
F(K) and G(K) take the form F(K) = G(K) =
42 (K 2 − K12 ) + 4if1 K − 2f12 42 (K 2 − K12 ) 4iK1 + 2f1 . 42 (K 2 − K12 )
,
(2.348) (2.349)
2.4 The Kerr metric as the solution to a boundary value problem
111
Here, we have chosen K1 = −K2 , i.e. we have set the horizon in a symmetric position in the , ζ plane, see Fig. 2.27. Making use of (2.63) and (2.64) and eliminating by the second constraint equation, we obtain the axis potential +
A :
f =
ζ (1 + f12 ) + ( f12 − 1 + 2f1 )K1 ζ (1 + f12 ) + (1 − f12 + 2f1 )K1
.
(2.350)
It is useful to introduce the multipole moments mass M and angular momentum J by an asymptotic expansion of f , M =
1 − f12 1 + f12
K1 ,
2i f1 K1 J = M 1 + f12
(2.351)
and to replace f1 and K1 in (2.348), (2.349), (2.350) and (2.347): F(K) =
(K + M )2 + (J /M )2 , K 2 + (J /M )2 − M 2
G(K) =
−2i J . K 2 + (J /M )2 − M 2
(2.352)
To represent f (ζ ), a simplifying parameterization is advisable, f1 = i tan δ/2, J = −M 2 sin δ, K1 = −K2 = M 2 − (J /M )2 = M cos δ,
(2.353) (2.354)
with the real parameter δ satisfying −
π < δ < 0. 2
(2.355)
Note that we have taken into account the necessary condition M > 0. Moreover, we have chosen J > 0 (and thus > 0) without loss of generality.29 This yields A+ :
f =
ζ − M + i M sin δ . ζ + M + i M sin δ
(2.356)
Finally, the second constraint equation (2.347) becomes the wellknown relation,
2 M M4 2M = − 1, (2.357) − J J2 connecting the angular velocity of the horizon with mass and angular momentum, cf. (1.136). 29 The solution with negative J (and negative ) is simply given by the complex conjugate Ernst potential.
Solutions with negative M are not singularityfree outside the horizon.
112
Analytical treatment of limiting cases
Solutions for which the Ernst potential along the axis is a rational function of ζ can uniquely be continued to all space using ‘Bäcklund techniques’, see Neugebauer (1996). The continuation of (2.356) gives f (, ζ ) =
r1 e−iδ + r2 eiδ − 2M cos δ , r1 e−iδ + r2 eiδ + 2M cos δ
(2.358)
where the nonnegative quantities ri are deﬁned by ri2 = (Ki − ζ )2 + 2
(i = 1, 2)
(2.359)
with Ki as in (2.354). This is the Ernst potential f of the Kerr solution30 in Weyl–Lewis–Papapetrou coordinates including the limiting cases δ → 0 (Schwarzschild solution, J = 0) and δ → −π/2 (extreme Kerr solution, J = M 2 ). To obtain the latter limit one has to apply the Bernoulli–l’Hospital rule. The extreme Kerr black hole is characterized by a degenerate horizon (vanishing ‘surface gravity’). The black hole uniqueness theorems, see Heusler (1996), state that the Kerr black holes with J < M 2 are the only stationary vacuum black holes with nondegenerate horizon. It is an interesting question whether the extreme Kerr black hole is the only stationary vacuum black hole with degenerate horizon. At least if axisymmetry and the corresponding conditions (1.125), (1.126) are assumed to hold from the very beginning, this question can be answered in the afﬁrmative – as will be shown in the remainder of this section. The degenerate case The extreme Kerr solution results from the limit δ → −π/2, i.e. the interval K1 ≥ ζ ≥ K2 of the ζ axis characterizing the horizon in canonical Weyl coordinates shrinks to a point (K1 = 0 = K2 ). We now want to show that the extreme Kerr solution follows uniquely, if we start with the assumption that the horizon is located at a point on the ζ axis. Of course, the slice t = constant, ϕ = constant of the horizon still has to be onedimensional. Therefore, we have to parameterize the position along the horizon by another coordinate. A natural choice is the angle ϑ of sphericallike coordinates R, ϑ: = R sin ϑ,
ζ = R cos ϑ,
0 ≤ ϑ ≤ π,
(2.360)
cf. (2.324). The horizon is placed at the origin (R = 0) of our Weylcoordinate system, see Fig. 2.28. The ‘north pole’ of the horizon is at ϑ = 0, the ‘south pole’ 30 The full metric is given by Equation (1.130). It can be calculated from the Ernst potential (2.358)
using (1.44), (1.41) and (1.40).
2.4 The Kerr metric as the solution to a boundary value problem
113
Fig. 2.28. The black hole with degenerate horizon ( = ζ = 0).
at ϑ = π. The calculation of and along the horizon gives U (K) V (K) f (ϑ) 1 = , W (K) X (K) f (ϑ) −1 H: −1 0 U (K) V (K) . = 2iK 1 0 W (K) X (K)
(2.361)
We can now repeat our analysis of the continuity of and at the ‘poles’ of the horizon, which touch the regular parts of the axis A± . The unique (positive mass) result for the axis Ernst potential is A+ :
f =
ζ − M − iM , ζ + M − iM
M =
1 , 2
(2.362)
where we have again assumed > 0. This is the axis potential of the extreme Kerr solution. The continuation to all space is given by f =
2R − 1 − i cos ϑ , 2R + 1 − i cos ϑ
(2.363)
cf. (2.329). Since a ﬁnite interval and a single point on the ζ axis are the only possibilities for the (Killing) horizon, we can conclude that the Kerr black holes – including the extreme case – are the only stationary and axisymmetric black holes surrounded by a vacuum.
3 Numerical treatment of the general case
After introducing basic notions and equations concerning ﬂuid bodies in the ﬁrst chapter, we turned our attention to the analytical treatment of a small number of limiting cases in the second. The methods presented there are quite powerful, as was shown, but also have severe limitations. In particular, it does not seem to be possible to use them for treating genuinely threedimensional ﬂuid bodies in which the shape of the free boundary emerges out of the simultaneous solution to an interior and an exterior problem along with transition conditions. We thus devote the third chapter to a numerical treatment of the problem, which is capable of handling the general case. The form of the metric we use in this chapter is not (1.5) as with the disc, but rather (1.6) ds2 = e2α (d2 + dζ 2 ) + W 2 e−2ν (dϕ − ω dt)2 − e2ν dt 2 .
(3.1)
The beneﬁt of this line element in the context of numerical solutions is that ν remains real even within an ergosphere (see the related comment on page 14). The ﬁeld equations are, cf. (1.29) and (1.33), 1 2 3 −4ν 1 + v2 2 2α ∇ · (B∇ν) − B e (∇ω) = 4πe B ( + p) + 2p , 2 1 − v2 v , ∇ · (2 B3 e−4ν ∇ω) = −16πB2 e2α−2ν ( + p) 1 − v2 ∇ · (∇B) = 16πBe2α p
(3.2a) (3.2b) (3.2c)
and 2 α −
1 1 ∂ν + ∇ν∇u − 2 B2 e−4ν (∇ω)2 = −4π e2α ( + p) ∂ 4 114
(3.2d)
3.1 A multidomain spectral method
115
with B := W /,
eu := eν /B
and v := Be−2ν ( − ω).
The operator ∇ has the same meaning as in a Euclidean 3space in which , ζ and ϕ are cylindrical coordinates. Thus the ﬁrst three of the ﬁeld equations can be applied as they are in other, related coordinates such as r, θ, ϕ with = r sin θ and ζ = r cos θ . In (3.2d), however, the operator 2 := ∂ 2 /∂2 + ∂ 2 /∂ζ 2 is not coordinate independent. The vanishing divergence of the energymomentum tensor yields a relationship between the speciﬁc enthalpy h and the metric functions given by (1.26) and (1.31), h = h(0) eV0 −V , eV = eν 1 − v 2 , (3.3) which allows us to express (h) √ and p(h) on the right hand sides of Equations V ν (3.2) as functions of e = e 1 − v 2 . As was already discussed in Section 1.4, an alternative to the second order differential equation (3.2d) for determining α, is a line integral. Here we choose nonetheless to treat α on the same footing as the other functions since it renders the numerical methods more transparent without requiring prohibitively more computational resources. This chapter deals with numerical solutions to the above ﬁeld equations together with appropriate boundary, regularity and asymptotic conditions. The ﬁrst section provides, by way of simple examples, relevant speciﬁcs of the numerical methods to be employed. The second provides a list of the coordinate mappings used in various scenarios. In the remaining sections, the physical properties of selected numerical solutions are presented. These include a detailed look at homogeneous ﬁgures, a discussion of a selection of further equations of state and ﬁnally a treatment of a central black hole surrounded by a ring. 3.1 A multidomain spectral method Numerical methods for solving the system of equations (3.2) describing the equilibrium of rotating relativistic stars have been developed since the 1970s (Wilson 1972, Bonazzola and Schneider 1974, Butterworth and Ipser 1976). The end of the 1980s and the beginning of the 1990s have seen a considerable further development and reﬁnement of the techniques being used (Friedman et al. 1986, 1989, Komatsu et al. 1989a,b, Lattimer et al. 1990, Cook et al. 1992, Neugebauer and Herold 1992, Stergioulas and Friedman 1995; for a review see Stergioulas 2003). The application of pseudospectral methods within the realm of numerical relativity was initiated by Bonazzola et al. (1993). An improved version of
116
Numerical treatment of the general case
the corresponding code by Bonazzola et al. (1998) demonstrated the excellent convergence properties of pseudospectral methods, which can lead to extremely high accuracy. In this section, we describe the pseudospectral method that was developed for the calculation of the numerical results presented in the remainder of this chapter (Ansorg et al. 2002a, 2003a). A particular feature of this method is the compactiﬁcation of the entire spatial domain and the division into several subdomains, where one of the domain boundaries is chosen to coincide exactly with the surface of the ﬂuid conﬁguration. In this manner, it is possible to avoid the socalled Gibbs phenomenon entirely, which would affect the spectral convergence rate. As a starting point, we introduce and illustrate pseudospectral methods for ordinary differential equations. We then present coordinate mappings for each domain. The collection of ﬁeld equations, boundary and transition conditions in each subdomain provides a highdimensional algebraic system of equations for the potential values at speciﬁed discrete spatial gridpoints. We describe the structure of the solution vector from the Newton–Raphson scheme utilized in order to tackle this system of equations. The ﬁnal result is a highly accurate spectral approximation of the desired gravitational ﬁeld of rotating ﬂuid conﬁgurations in equilibrium.
3.1.1 Chebyshev expansions For a spectral expansion of a realvalued function f deﬁned on the interval [a, b] (a, b ∈ R), we write f (x) =
n−1
(n)
ck k (x) + R(n) (x) .
(3.4)
k=0 (n)
Here, the integer n is the spectral expansion order, ck are the spectral coefﬁcients (which depend on the order n in general) and R(n) (x) is the residual term. The spectral basis functions k are chosen according to the properties of the underlying function f (e.g. trigonometric functions for periodic f ). In this book, we use Chebyshev polynomials of the ﬁrst kind deﬁned on the interval [a, b] = [0, 1], i.e. k (x) = Tk (2x − 1)
(3.5)
Tk (ξ ) = cos(k arccos ξ ) .
(3.6)
with
3.1 A multidomain spectral method
Moreover we choose the residual R(n) to vanish at the spectral gridpoints πj (n) 2 ; j = 0, 1, . . . , n − 1 ; xj = sin 2(n − 1)
117
(3.7)
which implies 2 − δk,0 − δk,n−1 (n) ck = (−1)k n−1 n−2
πkj 1 (n) , f (xj ) cos × f (0) + (−1)k f (1) + n−1 2
(3.8)
j=1
where δk, j is the Kronecker symbol. Note that (n)
x0 = 0,
(n)
xn−1 = 1
(3.9)
and (n)
n−1 (xj ) = (−1)n−1−j , (n)
i.e. the xj
(3.10)
are the abscissa values of the (local) extrema of n−1 in [0, 1]. (n)
The reason for this choice is an extremely rapid falloff of the coefﬁcients ck provided that the function f is analytic within [0, 1] and that there is no strong growth of the function’s derivative values as one goes to higher and higher derivatives of f . In a certain sense, the Chebyshev expansion results in a polynomial representation that is close to the polynomial of best approximation (Arnold 2001). In the following we illustrate the behaviour of the spectral coefﬁcients for some representative example functions. 1. Let us ﬁrst consider f1 (x) =
1 1 + x2
(3.11)
as an example of a smooth analytic function on [0, 1]. In Fig. 3.1, the absolute (n) values of the coefﬁcients ck for n = 30 are displayed (solid line). Moreover, the maximal residuals (n) R(n) max = max R (x) x∈[0,1]
(3.12)
are plotted against the resolution orders n (dashed line). We see a rapid decrease in the magnitudes of the coefﬁcients and also of the maximal residuals. The similarity
118
Numerical treatment of the general case
(30)
(n)
Fig. 3.1. Chebyshev coefﬁcients ck  (solid line) and maximal residuals Rmax (30) (dashed line) of the function f1 . The values along the abscissa are k for ck  or (n) alternatively n for Rmax .
between two such curves is a general feature of pseudospectral methods when applied to analytic functions f . In addition, in this semilog plot the values oscillate slightly about a straight line, indicating an exponential convergence rate of the spectral expansions. 2. The next example demonstrates a situation that could model weak gravitational sources where steep gradients of the gravitational potentials are present. For > 0 the function f2 (x) =
+x
(3.13)
is analytic but the magnitudes of the derivatives become very large in the vicinity of x = 0 if 1, as revealed by the Taylor expansion f2 (x) =
∞ x k − ,
0≤x o . Constant svalues correspond to constant rvalues: r0 , s 2 s2 (1 − t) , x˜ 0 = 2 m 2 s2 (1 − 2t) r 0 + m r=
y˜ 0 =
2 s2 t m . 2 s2 (1 − 2t) r02 + m
3.2 Coordinate mappings
133
– subdomain k = 1 The boundaries of this domain are two spherical shells (with radii r0 and r1 < i ), a section of the rotation axis (ζ ∈ [r1 , r0 ]) and a toroidal curve around the ring, intersecting the equatorial plane at the radii r1 and r0 . As with subdomain 0, constant svalues correspond to constant rvalues: s r0 r = r1 σ1 = r1 , r1
x˜ 1 =
y˜ 1 =
2 (r 2 − r 2 ) − r 2 t[σ 2 ( 2 − r 2 ) + (r 2 − 2 )] m m 0 1 1 1 m 1 0 2 + r 2 σ 2 ) − 2r 2 t[σ 2 ( 2 − r 2 ) + (r 2 − 2 )] (r02 − r12 )(m m 1 1 1 1 m 1 0 2 − r 2 ) + (r 2 − 2 )] r12 t[σ12 (m m 1 0 2 + r 2 σ 2 ) − 2r 2 t[σ 2 ( 2 − r 2 ) + (r 2 − 2 )] (r02 − r12 )(m m 1 1 1 1 m 1 0
,
.
– subdomain k = 2 This domain is a spherical shell around the central object which can either be a black hole (with the horizon chosen to be a sphere in the coordinates (, ζ )) or an inﬁnitely ﬂattened disc of dust. In the case of a single toroidal conﬁguration without a central object, we may describe the central vacuum region by a disc of dust with vanishing surface massdensity, i.e. we may take the same coordinates as for conﬁgurations with central discs. For a central black hole (with horizon coordinate radius rh < r1 ), we take a shell with inner radius rh and outer radius r1 . Again, constant svalues correspond to constant rvalues: 1−s rh r = r1 σ2 = r1 , r1
x˜ 2 =
y˜ 2 =
2 σ 2 − r2t m 2 h 2 σ 2 + r 2 (1 − 2t) m 2 h
rh2 t 2 σ2 m 2
+ rh2 (1 − 2t)
,
.
For a central disc (with coordinate radius d < r1 ) we take a transformation that interpolates between oblate spheroidal coordinates in the vicinity of the disc and spherical coordinates in the vicinity of the shell r = r1 :
22 = d2 + (s − 1)2 ξ 2 (t) τ (t) , ζ22 = (s − 1)2 ξ 2 (t) [1 − τ (t)]
134
Numerical treatment of the general case with 2 " # r 1 2 ξ 2 (t) = r12 − d2 + ! r12 + d2 − 4tr12 d2 1 + 12 (1 − t) , 2 m & 2 ' + r12 (1 − t) r12 t m & ' . τ (t) = 2 2 + ξ 2 (t) m d
– subdomain k = 3 This domain is a toroidal shell around the ring. The outer toroidal boundary intersects the equatorial plane at the radii r1 and r0 , and the inner one at i and o : x˜ 3 = t x˜ 1 (s, t = 1) + (1 − t)˜x4 (s, t = 0) ,
y˜ 1 (s, t = 1) y˜ 3 = y˜ 4 (s, t = 0) y˜ 4 (s, t = 0)
t .
The coordinates x˜ 4 and y˜ 4 follow from 42 and ζ42 deﬁned below through transformation (3.42). • Interior subdomain k = 4 This domain covers the interior of the ring. Its surface is obtained for t = 0 whereas t = 1 yields a section of the equatorial plane that is inside the ring: 42 = i2 (o /i )2s , ζ42 = (1 − t)[G(s) − 42 ] . In this transformation, the function G appears explicitly, which, as with spheroidal conﬁgurations, describes the unknown surface shape of the perfect ﬂuid body. From G, the alternative representation y˜ b , introduced in (3.43), is given implicitly by the relation y˜ b [˜x4 (s, t = 0)] = y˜ 4 (s, t = 0) . As the inner and outer equatorial edges of the ring correspond to s = 0 and s = 1 respectively, the surface function G is subject to the conditions G(s = 0) = i2 ,
(3.44)
G(s = 1) = o2 .
(3.45)
The function G (and hence the coordinate radii i and o ) is not known, but is determined through the numerical solution procedure. In the case of a toroidal conﬁguration with a central object, this also applies to the coordinate radii rh or d . On the contrary, the parameters r0 and r1 can be prescribed freely for the calculation.
3.2 Coordinate mappings
135
In order to get good accuracy even when using low resolution, these parameters should be chosen within some optimal interval. It will be useful in later sections to be able to refer to a radius ratio, A, for spheroidal as well as toroidal conﬁgurations. We thus deﬁne rp re : for spheroidal conﬁgurations A := (3.46) i − o : for toroidal conﬁgurations to be the ratio of polar to equatorial coordinate radius for spheroidal conﬁgurations and the negative of the inner to outer coordinate radius for toroidal ones. The radius ratio runs from −1 in the thin ring limit to 1 in the spherical limit, and A = 0 marks the transition point from one topology to the other. 3.2.3 The solution vector f of the Newton–Raphson scheme The numerical scheme to treat the free boundary value problems for axisymmetric, stationary equilibrium conﬁgurations generalizes the pseudospectral collocation point method introduced in Subsection 3.1.2 for ordinary differential equations. All functions U κ (κ = 0, 1, . . . , neq − 1) to be determined by the free boundary value problem are considered at speciﬁc gridpoints (sk,i ; tk,j ) in the subdomains k = 0, 1, . . . , ndom . These gridpoints are given through Equation (3.7), that is: πi (s) sk,i = sin2 ; i = 0, 1, . . . , nk − 1 ; (s) 2(nk − 1) (3.47) πj (t) tk,j = sin2 ; j = 0, 1, . . . , nk − 1 . (t) 2(nk − 1) (s)
(t)
The numbers nk and nk of gridpoints in domain k with respect to the s and t directions (i.e. the spectral expansion orders) may assume different values in the various subdomains, but at common boundaries the numbers have to coincide, i.e. for spheroidal conﬁgurations we have (t)
(t)
n0 = n1 , and for toroidal ones: (t)
(t)
(t)
(s)
(s)
(s)
n0 = n1 = n2 , n1 = n3 = n4 . As in Subsection 3.1.2, we collect all function values κ = U κ (sk,i , tk,j ) Uk,ij
(3.48)
136
Numerical treatment of the general case
as well as the values of the unknown surface function G, Gj = G(t1, j )
for spheroidal conﬁgurations and
Gi = G(s4,i )
for toroidal conﬁgurations,
in order to build up a vector f . In addition, this vector is ﬁlled with a number npar of physical parameters that characterize the conﬁguration. For a given equation of state, npar = 2 for onebody conﬁgurations, and we choose them to be V0 and , cf. Section 1.4. If a central object is present, then there are another two parameters (i.e. npar = 4) for which we take the angular velocity and coordinate radius of the central object, that is ( h , rh ) for a black hole and ( d , d ) for a disc. Whereas the number of parameters npar is ﬁxed, it is sometimes possible to ﬁnd more than one solution to a given set of parameters. For example, the crossing of the dashed and the solid lines in the inset of Fig. 3.17 below demonstrates that (at least) two √ conﬁgurations exist for that / A pair. The collection of elliptic equations valid in the subdomains, the transition conditions at common domain boundaries, the vanishing pressure boundary condition at the ﬂuid’s surface and certain parameter relations that one wishes to fulﬁl, yield a discrete nonlinear system of the form (3.24) where n stands for the (s) (t) collection of all nk and nk , (s)
(t)
n = {(nk , nk );
k = 0, 1, . . . , ndom − 1} .
The dimension of this system is given by ntotal = neq
ndom −1
(s) (t)
nk nk + nG + npar ,
(3.49)
k=0 (t)
(s)
with nG = n1 for spheroidal and nG = n4 for toroidal conﬁgurations. In particular, the transition conditions require the U κ to be continuous and to possess continuous normal derivatives. At domain boundaries that correspond to portions of the rotation axis or the equatorial plane, we require regularity conditions, which follow from the elliptic equations when specialized to this boundary. Via the integrated relativistic Euler equation (3.3), the vanishing pressure boundary condition restricts the potentials at the ﬂuid’s surface. It adds nG equations to the system. Finally, we may include speciﬁc parameter relations that we wish to be satisﬁed. For example, we could simply prescribe certain values for the physical parameters contained in f . However, we may instead wish to prescribe other parameters, say gravitational mass and angular momentum of the objects. For this reason we include the npar physical parameters into the vector f and add the appropriate parameter relations to the system.
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
137
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids In the previous section, a numerical method was presented for calculating ﬁgures of equilibrium. In this section, such methods will be applied to single homogeneous bodies, i.e. those described by the equation of state = constant, where is the energydensity. Newtonian homogeneous ﬁgures of equilibrium have played an invaluable role in the history of gravitational theory because of their simplicity, tractability and multifariousness. A number of the key contributors to the ﬁeld were listed in the preface, and important books combining both old and new material were written by Lichtenstein (1933) and Chandrasekhar (1969). The latter went on to consider relativistic homogeneous ﬁgures of equilibrium from various vantage points. The importance of this equation of state for Newtonian as well as Einsteinian theory is underscored by the number of analytic solutions that can be found using it. The Maclaurin spheroids, including their spherical and disc limits, have been presented here. Schwarzschild spheres and the rigidly rotating relativistic disc of dust, which are attached to these end points in the Newtonian limit, were also treated at length. It was found that the exterior extreme Kerr metric results as the extreme relativistic limit for the disc of dust. The question that arises is how these motley solutions ﬁt into the larger picture of solutions to Einstein’s equations for homogeneous matter. To answer this question, we employ a combination of analytic and numerical techniques. Analytic results will provide us with our ﬁrst ‘good initial guess’ for the Newton–Raphson scheme, cf. page 123. Furthermore they will allow us to locate singular conﬁgurations that are important for distinguishing between distinct regions of solution space. By beginning at a known analytic solution, say the Schwarzschild spheres, and tracing out connected boundaries in this solution space, it will be possible to divide it into a countably inﬁnite number of distinct classes. A schematic portrayal of the picture that emerges is given in Fig. 3.14, which will serve us throughout this section as a roadmap. Our task now is to proceed to explain how this class structure was ascertained and what its basic properties are.
3.3.1 Bifurcation points We have already encountered Newtonian homogeneous ﬁgures of equilibrium in Section 2.1, namely the Maclaurin spheroids. The secular stability of these and other ﬁgures of equilibrium was studied by Poincaré (1885), who determined the change in energy associated with a change in the shape of the surface up to second order. He focused primarily on the second harmonic, nonaxially symmetric
138
Fig. 3.14. A schematic portrayal of the classes and their boundaries. The analytically known solutions are written in larger boldfaced letters (after Ansorg et al. 2004).
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
139
Table 3.1. Numerical values for the ﬁrst four solutions to Equation (3.50). In addition to the value of ξ0 , the corresponding values of the eccentricity and of the ratio of the minor to major axis of the ellipse are given. l
ξ0
c/a = Al−1
2 3 4 5
0.17383011 0.11230482 0.08303471 0.06588682
0.98522554 0.99375285 0.99657034 0.99783651
0.17126187 0.11160323 0.08274493 0.06574427
perturbation. On the other hand, Chandrasekhar (1967, 1968) studied the fourth harmonic axially symmetric mode and determined the value for the eccentricity at the onset of axially symmetric instabilities. Bardeen (1971) derived formulae to determine all marginal, axisymmetric instabilities, which were generalized to include nonaxisymmetric modes by Hachisu and Eriguchi (1984). Numerical results (Eriguchi and Hachisu 1982, Ansorg et al. 2003c) conﬁrm that the solutions to these equations indeed mark bifurcation points along the Maclaurin sequence. By considering an axisymmetric perturbation in the shape of the Maclaurin spheroids, it is possible to locate nonellipsoidal Newtonian stars deviating only inﬁnitesimally from an ellipsoid (see Ansorg et al. 2003c). Such conﬁgurations are found when the constant value of the elliptic coordinate describing the surface of the body is a solution of the equation iP2l (iξ0 )Q2l (iξ0 ) − ξ0 (1 − ξ0 arccotξ0 ) = 0;
l = 2, 3, 4, . . . ;
(3.50)
where Pn (x) and Qn (x) are the Legendre polynomials and Legendre functions of the second kind. For a given value of l, there is exactly one solution to (3.50), which will be denoted by ξ2l∗ . The ﬁrst few are listed in Table 3.1. These values for ξ0 mark a countably inﬁnite number of bifurcation points that have an accumulation point at ξ0 = 0. The nature of these bifurcation points within relativity is quite interesting. Chandrasekhar (1967) noticed that the ﬁrst postNewtonian expansion for the Maclaurin spheroids contains a singularity at the ﬁrst bifurcation point associated with l = 2. Upon reexamining this work, Bardeen (1971) conjectured that there exists a singularity in the postNewtonian expansion for every bifurcation point. This conjecture was veriﬁed by Petroff (2003), who showed that the singularity associated with the value l ﬁrst appears at the (l − 1)st postNewtonian order.
140
Numerical treatment of the general case
√ Fig. 3.15. Sequences of constant M are plotted in the βA parameter space in the vicinity of the ﬁrst bifurcation point A1 . For a deﬁnition of the massshed parameter β, see (3.51), and for the radius ratio A, see (3.46) (after Ansorg et al. 2004).
If one assumes that the postNewtonian expansion converges at least for some values of a relativistic expansion parameter, then these results imply that there exists no (relativistic) ﬁgure of equilibrium at these bifurcation points. This expectation is conﬁrmed by numerical results. For example, Fig. 3.15 shows lines of constant √ normalized gravitational mass M as it depends on the radius ratio A of (3.46) in the vicinity of the ﬁrst bifurcation point. For any nonzero value of the mass, there is a region about the bifurcation point containing no solutions. What is more, we shall see that it is not possible to cross over from the solutions on one side of this point to those on the other without passing through the bifurcation point itself. This behaviour of the solution space turns out to hold for each of the bifurcation points and allows us to divide the solution space into classes. Each class contains a ∗ section of the Maclaurin sequence running from ξ0 = ξ2l∗ to ξ0 = ξ2l+2 . The nature of the classes will become clearer in the next sections, when we examine the ﬁrst few of them, comment on the remaining ones and ﬁnally sketch out a picture of the solution space.
3.3.2 Class I: the generalized Schwarzschild class As can be seen in Fig. 3.14, the generalized Schwarzschild class is demarcated by ﬁve boundary sequences, one of which is the static limit. A detailed look at this class can be found in Schöbel and Ansorg (2003), which also provides some information
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
141
regarding the numerical techniques involved. We now proceed to traverse these ﬁve sequences, making note along the way of some of their interesting features. The Schwarzschild solution: Consider the interior and exterior Schwarzschild solution for a given energy density , as discussed in Section 2.2. The solution depends on one parameter, say the relative redshift z of zero angular momentum photons of Equation (1.28), which runs from 0 in the Newtonian limit to 2 in the limit of inﬁnite central pressure ((a) to (b) in Fig. 3.14). The value for this redshift can be further increased if we abandon the static limit, allowing the star to rotate. Such a rotating star is described by two parameters. Let us choose to consider a sequence of stars with inﬁnite central pressure and ask what happens when we increase the angular velocity . Inﬁnite central pressure: Equation (2.30) shows that e2ν → 0 at the star’s centre as the central pressure pc → ∞. As soon as such a star is set in motion, an ergosphere forms. The ergosphere grows outward from the centre as is increased and crosses the star’s surface. Increasing further, causes the shape of the surface at the equatorial plane to become more pointed until, ﬁnally, a cusp indicating a massshedding limit is reached ((b) to (c) in Fig. 3.14). This salient conﬁguration deserves more attention, since many important physical properties, such as mass, attain their maximum values for this class precisely here, where the sequence of inﬁnite central pressure meets the massshedding limit. As we can see in Table 3.2, the (polar) redshift can increase signiﬁcantly relative to the nonrotating case. Furthermore, the maximal attainable mass, given by the Buchdahl limit for the Schwarzschild solution, i.e. Equation (2.37), increases by roughly 34% due to rotation. Figure 3.16 shows the coordinate shape of this star and its ergosphere. The massshedding sequence: We can describe a sequence of stars by requiring that they all rotate at the massshedding limit. To do so, it is convenient to introduce the massshed parameter β, which provides a measure of ‘how pointy’ a conﬁguration is around the equator:
2 " 2 # d ζb spheroidal topology: − rrpe d(2 ) =re " 2# 2i d ζb β = toroidal topology, = i : o −i d(2 ) (3.51) =i # " 2 2o d ζb , toroidal topology, = o : i −o d(2 ) =o
where ζ = ζb () describes the conﬁguration’s surface. This parameter is chosen such that β = 0 in the massshedding limit and β = 1 for (Maclaurin) spheroids.
142
Numerical treatment of the general case
Table 3.2. Properties of the conﬁguration of maximal mass for the generalized Schwarzschild class. An asterisk indicates that the corresponding quantity is a global maximum. Physical quantity Gravitational mass Baryonic mass Angular velocity Angular momentum Polar radius Equatorial radius Radius ratio Redshift
Value M M0
J rp re A z
= 0.19435 = 0.27316 = 1.8822 = 0.03637 = 0.04856 = 0.08475 = 0.5730 = 7.378
−1/2 −1/2 1/2 −1 −1/2 −1/2
∗ ∗ ∗ ∗
∗
Fig. 3.16. Meridional crosssection (solid line) and ergosphere (dashed line) of the conﬁguration of maximal mass of the generalized Schwarzschild class. Axes are scaled identically (after Schöbel and Ansorg 2003).
Let us traverse the massshedding sequence by choosing β = 0 and varying a √ second parameter, say / , which changes monotonically along it. We know that inﬁnite central pressure marks one end point of the massshedding sequence and it turns out that the other end point is given by the Newtonian limit ((c) to (d) in Fig. 3.14). The ergosphere must shrink and vanish along this sequence since it ends in a Newtonian conﬁguration and it turns out that the normalized angular velocity √ √ falls from / = 1.8822 to / = 1.0775 while the ratio of polar to equatorial radius falls monotonically from A = 0.5730 to A = 0.1922.
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
143
The ﬁrst Newtonian lens sequence A+ 1 : Although the massshedding sequence ends in a Newtonian limit, it clearly cannot end in a Maclaurin spheroid. There exists a Newtonian sequence connecting the massshedding limit to the Maclaurin solution however, which reaches it precisely at the bifurcation point A1 ((d) to (e) in Fig. 3.14). We call this sequence the ﬁrst Newtonian lens sequence because of the shape the conﬁgurations take on as they approach the massshedding limit. Bardeen (1971) surmised that such a sequence exists and it was studied along with other Newtonian bifurcation sequences by Ansorg et al. (2003c). The Maclaurin segment: The portion of the Maclaurin sequence joining A1 to the spherical limit comprises a Newtonian sequence that brings us back to our starting point. We have succeeded in identifying ﬁve connected physical boundary curves that thus delimit the generalized Schwarzschild class of solutions. The class itself is made of all the conﬁgurations within this boundary, the borders of which constitute its closure. The classic Maclaurin curve, which follows from (2.15) and depicts how the square of the normalized angular velocity depends on the radius ratio A, is found in Fig. 3.17 as one of the Newtonian limits. Only three of the ﬁve boundary curves are discernable on the plot: the entire Schwarzschild sequence resides at the point A = 1 and the ﬁrst Newtonian lens sequence A+ 1 is too small to be seen except in the inset.
Fig. 3.17. The square of the normalized angular velocity 2 /4π is plotted versus the radius ratio A. Ergospheres exist for conﬁgurations above the light dashed line. The labels and line types used are the same as those used in Fig. 3.14 (after Schöbel and Ansorg 2003).
144
Numerical treatment of the general case
Fig. 3.18. The normalized gravitational mass is plotted versus the radius ratio. Ergospheres exist for conﬁgurations above the dashed line. The labels and line types used are the same as those used in Fig. 3.14 (after Schöbel and Ansorg 2003).
The dotted lines are lines of constant central pressure, which are equally spaced in pc /( + pc ). The light dashed line marks the boundary between conﬁgurations with an ergosphere and those without. In Fig. 3.18 the normalized mass is plotted versus the radius ratio. This allows us to represent the Schwarzschild sequence as a line instead of a point. In fact, all ﬁve boundary curves are visible in this plot although the A+ 1 sequence, situated between the points (d) and (e), overlaps entirely with a segment of the Maclaurin sequence. In this plot, we can see that the onset of ergospheres occurs at a roughly constant √ value of M .
3.3.3 Class II: the generalized Dyson ring class and its black hole limit As with the generalized Schwarzschild class, the generalized Dyson ring class can be deﬁned by locating connected boundary sequences. The members of the class and in particular the transition to a black hole were studied in Ansorg et al. (2003b). Beginning at the bifurcation point A1 , one ﬁnds that the associated Newtonian bifurcation sequence contains not only A+ 1 , but also a second arm, which we denote − by A1 . The shape of the surface along this sequence can be seen in Fig. 3.19. The ﬁrst three members in this sequence of pictures belong to A+ 1 and were discussed in the previous section. The subsequent ones belong to the bifurcation sequence
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
145
Fig. 3.19. Meridional crosssection of Newtonian bodies belonging to the ﬁrst bifurcation sequence with ζ /re (ζ /o ) plotted versus /re (/o ). The portion of the sequence running from the Maclaurin spheroid (denoted by hatching) to the − massshedding limit belongs to A+ 1 and the other portion to A1 . The transition point from spheroidal to toroidal topology is marked by an asterisk (after Ansorg et al. 2003b).
A− 1 ((e) to (f) in Fig. 3.14), which is a Newtonian boundary curve of the class of solutions being discussed here. The original ellipsoid grows increasingly dumbbell shaped and ﬁnally pinches together at the centre and takes on a toroidal topology. For toroidal objects, a numerical code using coordinates like those described in Subsection 3.2.2 is needed. The conﬁguration at the transition point provides possible ‘initial data’ for the program as does the Newtonian thin ring limit to be discussed below. As always, once one has succeeded in ﬁnding a single numerical solution, it can be used as the initial guess in the Newton–Raphson method for determining the solution to a neighbouring conﬁguration. Newtonian rings received considerable attention in the nineteenth century and a good deal is known about them. In particular, Dyson (1892, 1893) determined the gravitational potential by carrying out an expansion to fourth order about the thin ring limit in which the ring’s crosssection becomes circular. The results are remarkably accurate even when A deviates signiﬁcantly from −1 (see e.g. Fig. 1 in Ansorg et al. 2003c). We see in Fig. 3.19 that the rings in this sequence are tending toward the thin ring limit. In fact, A can be arbitrarily close to −1. In that limit,
146
Numerical treatment of the general case
Fig. 3.20. A plot demonstrating the behaviour of Newtonian rings in the thin ring limit A → −1. As the function plotted here tends to −∞, it approaches the straight line given by Equation (3.52) and denoted by a dotted line.
if the extent of the ring o and its mass M are kept ﬁnite, it turns out that V0 as a function of A tends logarithmically to minus inﬁnity such that 4πo V0 1 − ln(1 + A) = − ln 16 A→−1 5M 4 lim
(3.52)
holds (Dyson 1892, Horatschek and Petroff 2008; see also Equation (11) in Fischer et al. 2005). A graphical depiction of this behaviour can be found in Fig. 3.20. Since Newtonian theory is obtained as V0 → 0, we see that this thin ring limit poses a consistency problem. We shall return to this interesting point after considering the other boundary sequences. Moving back to the point A1 , we can traverse the Maclaurin sequence to the next bifurcation point A2 ((e) to (i) in Fig. 3.14). Bifurcating from this point is the A+ 2 sequence, which is very similar to A+ 1 . It too ends in a massshedding limit after a short time, short meaning that A, and J vary only slightly along it ((i) to (h) in Fig. 3.14). Travelling now along the massshed sequence, we ﬁnd that the surface of the body pinches in at the centre as we depart from the Newtonian limit. There is a change in topology, which occurs at 1 − eV0 ≈ 0.2. As we can see in Fig. 3.21, the value of this parameter can approach the value of one, i.e. the relative redshift can tend to inﬁnity for rings with a radius ratio less than A ≈ −0.56 ((h) to (g) in Fig. 3.14). As indicated in the ﬁgure, this is the limit in which the exterior metric becomes that of the extreme Kerr black hole (see Subsection 1.8.2 for a discussion of the exterior metric and the ‘inner world’).
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
147
Fig. 3.21. In this plot of 1 − eV0 versus A, the massshedding sequence marks a boundary for the conﬁgurations of Class II, which lie entirely within the shaded region. The line indicating the transition from a spheroidal to a toroidal topology is emphasized.
Let us consider the evidence supporting this claim. We recall that Meinel (2006) has shown that a necessary and sufﬁcient condition for the existence of a parametric transition to an extreme Kerr black hole is that eV0 → 0. It is not to be expected that it will be easy to come arbitrarily close to this limit using numerical means. After all, it becomes increasingly necessary to resolve two increasingly disjoint worlds. For a ring quite close to this limit, the metric functions in its vicinity resemble those of a nonextreme ring. The asymptotic behaviour of the ‘inner world’ describes the functions’ behaviour far away from the ring, but these functions must possess large gradients in order to assume their asymptotically ﬂat values. Making use of logarithmic coordinates such as those described by (3.15), it was nonetheless possible to ﬁnd rings with eV0 ≈ 10−3 . In light of the proofs mentioned above, this alone is convincing support of our claim. We choose to go further however and explore the behaviour of the metric. Examining multipole moments, it was possible to show that they indeed approach those of the extreme Kerr metric (Labranche et al. 2007). Moreover, they tend to these values in a way that is independent of the equation of state and independent of whether or not the transition body is a ring or a disc of dust. There also exists evidence showing that the metric in the vicinity of the ring approaches that of the inner world. The sequence of plots in Fig. 3.22 shows the development of the throat geometry and is inspired by Fig. 13 in Bardeen and Wagoner (1971). The ﬁnal plot in the sequence shows two separate worlds after
148
Numerical treatment of the general case
√ Fig. 3.22. The function gϕϕ in the equatorial plane is plotted versus proper (√ radial distance δ = 0 g d, both normalized with respect to the mass M , for a sequence approaching the extreme Kerr limit e2V0 = 0. All four plots were made for a homogeneous ring with a radius ratio A = −0.7 and with a value for e2V0 as indicated. The thin vertical lines indicate the positions of the inner and outer radii of the ring. In the last plot (e2V0 = 0), gives the proper radial distance in the √ Kerr metric to the reference point = M . In the throat region, gϕϕ /M tends to the constant value 2, cf. page 30. Note that the proper distance between any point in the ‘inner world’ region and any point in the ‘extreme Kerr’ region tends to inﬁnity as e2V0 → 0, cf. (1.147) (after Labranche et al. 2007).
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
149
having carried out the limiting process. To construct the inner world solution, a spectral program was used, which prescribes the correct nonﬂat asymptotic values for the functions as discussed in Subsection 1.8.2. The startup data that we used for this program were provided by simply cutting off the far exterior region for rings from the asymptotically ﬂat program for rings with eV0 1. The fact that such startup data work, provides additional evidence suggesting that the metric in the ring’s vicinity approaches that of the inner world. Let us return now to the point (f) in Fig. 3.14, marked as an open circle, where the extreme Kerr and the Newtonian A− 1 sequence approach each other. From the standpoint of Fig. 3.21, this amounts to representing the whole line A = −1 by this open circle. This symbolic representation has its origin in the singular nature of the thin ring limit A → −1. The physical requirements that one can impose on a ring for ﬁnite V0 are that its mass and its extent remain ﬁnite. As an invariant measure of the extent of the ring, we here choose to use the circumferential radius of the outer edge in the equatorial plane Rcirc o , which is simply the proper circumference at = o , ζ = 0 divided by 2π. Based on the Newtonian result (3.52), one might suppose that for a ﬁnite ﬁxed value of V0 , M /Rcirc → 0 as A → −1. Figure 3.23 provides an example o of the veriﬁcation of this expectation for eV0 = 0.6. The constant 4πV0 /5 is not of particular importance in this example and was simply chosen since the curve is then known to be linear in the Newtonian limit. What is signiﬁcant is that indeed M /Rcirc → 0 in the thin ring limit making it a physically unacceptable model. o This explains the reason for choosing a dashed line for A = −1 in Fig. 3.21. Two sequences with constant, ﬁnite M /Rcirc can be found in that ﬁgure. Each o
circ 4π Ro V0 Fig. 3.23. The function exp is plotted versus 1 + A for a sequence with 5M exp(V0 ) = 0.6, cf. Fig. 3.20 and Equation (3.52).
150
Numerical treatment of the general case
such sequence runs from the massshedding limit to that of the extreme Kerr black hole. The most relevant such curve in the current context would have a value M /Rcirc = constant 1. For conﬁgurations sufﬁciently far away from the thin o ring limit, this condition indicates the validity of the Newtonian limit. Thus the corresponding curve in Fig. 3.21 would ‘begin’ along the massshedding curve arbitrarily near the Newtonian limit 1 − eV0 = 0. As such a curve nears A = −1, Equation (3.52) tells us that V0  would grow rapidly, which turns out to hold even far away from the Newtonian limit. The curve would thus make a sharp bend upward and follow the line A = −1 right up to its end point at 1 − eV0 = 1. 3.3.4 Class III: the ﬁrst generalized corering class The conﬁgurations comprising the Newtonian bifurcation sequence associated with A2 and belonging to the third class begin to take on a rollingpin shape in crosssection as they depart from the spheroid. This effect becomes more pronounced as one proceeds along sequence A− 2 ((i) to (j) in Fig. 3.14) until a ring appears which is on the verge of separating from a central body. We consider such conﬁgurations to mark a boundary in the conﬁguration space, since our intention here is to classify single homogeneous bodies. Thus conﬁgurations at the transition point separating single bodies from two bodies form a boundary sequence running from the Newtonian limit to the massshedding limit ((j) to (k) in 3.14). As one moves √ away from the Newtonian limit, / increases along this sequence until shortly before reaching the massshedding limit, when it again falls slightly. Its global maximum in this class is found at the bifurcation point A2 however. Along the massshed sequence ((k) to (l) in Fig. 3.14), the conﬁgurations again become rollingpin shaped before ﬁnally resembling a spheroid with a cusp at the equator in the Newtonian limit. The Newtonian A+ 3 sequence, which can be seen in Fig. 3.24, then takes us to the true spheroid at the bifurcation point A3 ((l) to (m) in Fig. 3.14), which is then connected to A2 via the Maclaurin sequence ((m) to (i) in Fig. 3.14). Having again come full circle, we have identiﬁed a closed circuit of boundary sequences deﬁning the borders of this class of solutions. 3.3.5 Class IV: the ﬁrst generalized tworing class As one traverses the Newtonian A− 3 sequence ((m) to (n) in Fig. 3.14), the highly ﬂattened stars develop two bulges. Figure 3.24 depicts the surface of the conﬁgurations in the third bifurcation sequence. The A+ 3 portion of it, which belongs to class III, runs from the hatched Maclaurin spheroid to the massshedding star at the top of the ﬁgure. The sequence we are concerned with runs downward from the Maclaurin spheroid. Along the way there is a change in topology and
3.3 Equilibrium conﬁgurations of homogeneous ﬂuids
151
Fig. 3.24. Meridional crosssection of Newtonian bodies belonging to the third bifurcation sequence with ζ /re (ζ /o ) plotted versus /re (/o ). The portion of the sequence running from the Maclaurin spheroid (denoted by hatching) to the − massshedding limit belongs to A+ 3 and the other portion to A3 (after Ansorg et al. 2003c).
one ﬁnds rings with a bowling pin shape in crosssection. The tapering in this ‘bowling pin’ becomes more pronounced until one ﬁnally reaches a ring on the verge of the transition to two rings. Such twobody conﬁgurations again mark a boundary of our solution space and can be followed from the Newtonian limit to the massshedding limit ((n) to (o) in Fig. 3.14). Along the massshedding sequence ((o) to (p) in Fig. 3.14) the ﬁgures again pass through the transition point from toroidal to spheroidal topologies and one ends in a Newtonian limit that is barely distinguishable from a spheroid. For example, along the entire Newtonian A+ 4 sequence ((p) to (q) in Fig. 3.14), the relative change in physical parameters such as A, or J is only a few percent. This boundary sequence brings us back to the Maclaurin solutions at the point A4 . The ﬁnal boundary sequence enclosing the solutions in class IV is made up of the spheroids in the Newtonian segment joining bifurcations points A3 and A4 ((q) to (m) in Fig. 3.14). Relative changes in the parameters listed above are about a factor of ﬁve greater along this segment than along the A+ 4 sequence.
3.3.6 Overview of the solution space including the disc limit As was explained in Subsection 3.3.1, it is possible to divide the solution space of rigidly rotating, relativistic, homogeneous bodies into a countably inﬁnite number of classes of solutions (for a detailed discussion, see Ansorg et al. 2004). The ﬁrst three of them are depicted in Fig. 3.25, which constitutes a fuller version of Fig. 3.17. Each class contains a segment of the Maclaurin solution between adjacent bifurcation points as one of its boundaries and is comprised of all conﬁgurations
152
Numerical treatment of the general case
Fig. 3.25. The square of the normalized angular velocity 2 /4π is plotted versus the radius ratio A for the ﬁrst three classes. A blowup containing Class III provides greater detail (after Ansorg et al. 2004).
that can reach this boundary via a continuous transition. It turns out that each class has ﬁve boundary sequences enclosing it and we refer again here to the schematic overview in Fig. 3.14, which has played an important role throughout Section 3.3 in discussing individual classes. The ﬁrst two classes are exceptional in that each contains boundary sequences unique to that class. The generalized Schwarzschild class (Class I) contains the static limit and that of inﬁnite central pressure. Every other class contains a ﬁnite limit for the maximal pressure. The generalized Dyson ring class (Class II) allows for a transition to the extreme Kerr black hole, something otherwise possible only after reaching the disc limit. To reach this limit, an inﬁnite number of further classes must be traversed. Each of them contains three Newtonian boundary sequences: a
3.4 Conﬁgurations with other equations of state
153
Maclaurin segment, a bifurcation sequence ending in a massshedding limit and a bifurcation sequence ending in a twobody conﬁguration. In each of these higher classes, a twobody and simultaneously massshed conﬁguration exists, meaning that the class is indeed bounded by ﬁve sequences. This conﬁguration is likely to be the one with the highest relative redshift in the class. The higher classes can be further divided into even and odd ones. The twobody limit in the (2n + 1)st class consists of a central core with n humps surrounded by a ring just barely touching it. Similarly, the twobody limit in the (2n + 2)nd class consists of a ring with n humps surrounded by a ring just barely touching it. These statements are based on following the behaviour of the ﬁrst ten Newtonian bifurcation sequences (see Ansorg et al. 2003c), a detailed study of the ﬁrst four relativistic classes and then extrapolating. As one proceeds to ever higher classes, the conﬁgurations grow increasingly ﬂat and closely resemble a manyring system. Entire classes deviate only marginally from the Newtonian limit as measured by the maximum value for the relative redshift for example. As n → ∞, one approaches the Maclaurin disc limit. This boundary conﬁguration is simultaneously the Newtonian limit of the rigidly rotating disc of dust. The solution space presented here includes all homogeneous ﬁgures of equilibrium with a Newtonian limit. It may well be the entire solution space. It is difﬁcult to rule out the possibility that there exist relativistic solutions that do not possess a Newtonian limit whatsoever. Such solutions, if they exist, could be remote islands in some parameter space that are difﬁcult to chance upon with the numerical methods used here. 3.4 Conﬁgurations with other equations of state Figures of equilibrium of constant density have played a crucial historical role and are of particular academic interest because of their tractability to analytic methods. Models with other equations of state are of greater astrophysical relevance however, and we thus want to touch on some of their properties here. Arbitrary barotropic equations of state present no signiﬁcant difﬁculties to the numerical methods presented in this book. Indeed the inaccuracies inherent to massshedding conﬁgurations are less pronounced for equations of state without discontinuities in the density at the surface of the body, as explained in Section A1.1. Equations of state involving phase transitions may require the introduction of additional computational domains and tabulated ones the introduction of (ideally analytic) functions p = p(), but these are minor technical points that can be addressed as the need arises. Here we shall compare homogeneous bodies to those governed by three other equations of state: various polytropic ones
154
Numerical treatment of the general case
with particular emphasis on the one with the polytropic index n = 1, the one describing a completely degenerate ideal Fermi gas of neutrons and one describing strange matter based on the MIT bag model. Each of these equations of state was discussed in Section 1.5. The governing principle that will be used in this subsection for deciding which exemplary conﬁgurations to bring, is to restrict ourselves to the ﬁrst two classes of solutions and to provide a rough overview of the properties of the ﬁrst of them.
3.4.1 Polytropic equations of state The polytropic equation of state γ
1+ 1n
p = KµB = KµB
(3.53)
provides a mathematically simple relation between the pressure p and the baryonic mass density µB , where K is known as the polytropic constant. The polytropic index n determines the ‘stiffness’ of the equation, which can range from 0 for the stiffest case, homogeneous bodies, to large values for ‘soft’ equations of state, which exhibit a dense concentration of matter surrounded by a tenuous envelope. In Newtonian theory, in which µB simply becomes the mass density, closed form solutions are known in the static case for n = 0, n = 1 and n = 5, the last value also being the maximal attainable one for spherically symmetric Newtonian polytropes. When using the methods presented in this book for the numerical description of a polytropic conﬁguration, one has to keep in mind that (h) is an analytic function of h at h(0) only for n = 0, 1, 2, . . . , a property which also carries over to the metric functions. As was mentioned in the footnote on page 128, this affects the rate of convergence. Nonetheless, generic conﬁgurations with noninteger n can be calculated to high accuracy without great numerical effort. An example of the convergence rate for a chosen sequence as it depends on n is provided in Fig. 3.26. Using the measure Rm φ as deﬁned in (3.34), one sees that the solutions for small values of the spectral resolution m are extremely accurate when n = 0 or n = 1. For other values of n, one can choose m to be well in excess of 40 if one requires extremely high accuracy, though m = 15 is sufﬁcient for most purposes. Having come to know the nature of the solution space for homogeneous bodies in the last chapter, it is natural to wonder how a change in the equation of state affects that picture. We begin addressing that question by looking at what happens in the vicinity of the bifurcation point A1 when the polytropic index n is perturbed slightly. In the homogeneous case, it turned out that the Newtonian bifurcation
3.4 Conﬁgurations with other equations of state
155
Fig. 3.26. Convergence is shown in dependence of the polytropic index n for a sequence with constant radius ratio A = 0.8 and constant central pressure K n pc = 10−4 (pc / = 1 for n = 0). The index m indicates how many Chebyshev polynomials were used in each dimension of each domain and the measure of convergence, Rm φ , is deﬁned in (3.34). The polytropic index n was varied in steps of 0.05, which thereby determines the width of the spikes at n = 0 and n = 1. In the continuum limit, these spikes would be much narrower.
points are singularities in the postNewtonian expansion and that conﬁgurations with radius ratio A = Ak and massshed parameter β = 1 can only be reached in the Newtonian limit. Similarly, it seems that polytropic Newtonian conﬁgurations can only reach such points for n → 0. The behaviour of three Newtonian sequences in the vicinity of A1 is depicted in Fig. 3.27 for n = 10−2 , n = 10−3 and n = 10−4 . Comparison with Fig. 3.15 shows that the progression for homogeneous bodies from the Newtonian limit to slightly relativistic ones is very similar to the progression in the Newtonian limit from polytropes with n = 0 to slightly larger values. The similarities between Figs 3.15 and 3.27 suggest that for polytropes (with n > 0) it will also be possible to identify disjoint classes of solutions. Here, two ‘adjacent classes’ are not going to have a point in common however. We no longer have an anchoring sequence such as the Maclaurin sequence, binding together the classes and providing a natural means of ordering them linearly. It will still be the homogeneous classes that we use for providing an ordering scheme and a nomenclature for a more general solution space, however. To that end let us start by calling a conﬁguration a basis conﬁguration if it is related to the homogeneous bodies located at the points (b), (f), (j), (n), etc. of Fig. 3.14 by remaining at the intersection of the two pertinent sequences (e.g. twobody and Newtonian for the point (j)) and by continuous variation of the polytropic index n. For a
156
Numerical treatment of the general case
Fig. 3.27. Newtonian sequences in the vicinity of the bifurcation point A1 are depicted for polytropes with the indicated value of the polytropic index n. See text for a comparison with Fig. 3.15.
given polytropic equation of state, the kth class can then be deﬁned to consist of those conﬁgurations connected via continuous parameter variations to the basis conﬁguration associated with that class. Having introduced this broader notion of a class of solutions, let us proceed to look at the ﬁrst two classes for polytropes with n = 1. The choice of this value for the polytropic index is somewhat arbitrary, but has the advantages of being fairly widely discussed in the literature for one, and being sufﬁciently far away from n = 0 so as to have some qualitatively new features for another. Polytropes with n = 1 The ﬁrst class of solutions for polytropes with n = 1 is enclosed by four boundary sequences: static, inﬁnite central pressure, massshed and Newtonian. One sequence is missing compared to the ﬁve homogeneous ones since the Newtonian limit is no longer divided into segments by a bifurcation point. The static limit is the most accessible and best understood limit, merely requiring that one solves the TOV equation (see Section 1.7.2). As in the homogeneous case, the sequence of static polytropes with n = 1 also runs from the Newtonian limit to that of inﬁnite central pressure. The Newtonian solution, as was mentioned above, is analytically known. Starting from this limit and choosing central pressure to parameterize the sequence, one ﬁnds for a given equation of state (i.e. for a constant value of K) that gravitational mass is no longer a monotonically increasing function. The behaviour of mass as it depends on the circumferential radius is depicted in Fig. 3.28. There is
3.4 Conﬁgurations with other equations of state
157
Fig. 3.28. The mass of static polytropes with n = 1 is plotted as a function of circumferential radius Rcirc = gϕϕ ( = re , ζ = 0). The points labelled (a) and (b) are discussed in the text.
a pronounced global maximum and there are many local maxima as one spirals in toward the limit of inﬁnite central pressure. The global maximum, which has been labelled ‘(a)’, also marks a maximum for the baryonic mass and the onset of an instability with respect to axisymmetric perturbations as discussed in Chapter 4. A further point of interest, ‘(b)’, marks the transition point from positive to negative binding energies M0 − M . It is well into the unstable branch of solutions and is further evidence indicating that this branch is not astrophysically relevant. Negative binding energies in general relativity have been discussed in the context of static polytropes by Tooper (1964).4 Although in the example provided here, the binding energy becomes negative far away from the Newtonian limit, the phenomenon as such is not solely relativistic. The binding energy of Newtonian static polytropes with n > 3 is known to be negative for example. The inﬁnite central pressure sequence runs from the static to the massshedding limit. This sequence is not as relevant as the corresponding one for homogeneous conﬁgurations since physical parameters such as mass, redshift and angular velocity are maximal for ﬁnite values of the central pressure. Furthermore, the binding energy remains negative over the whole of the sequence. Joining on to the sequence of inﬁnite central pressure is the massshedding sequence. It is more astrophysically relevant, not least of all because the maximal values for mass, baryonic mass, redshift, angular momentum and angular velocity 4 Note that ‘polytrope’ in Tooper (1964) is taken to be p = K γ and not p = Kµγ [see (1.46)]. B
158
Numerical treatment of the general case
Table 3.3. Maximal values for physical quantities are provided for polytropes with n = 1, where ‘maximal’ refers here to a global maximum for Class I. Physical quantity
Value
Gravitational mass Gravitational mass Baryonic mass Angular velocity Angular momentum Redshift
Mstatic = 0.164 M = 0.188 M0 = 0.207
= 0.634 J = 0.0204 z = 0.529
A K 1/2 K 1/2 K 1/2 K −1/2 K
1.000 0.585 0.585 0.603 0.583 0.596
are all to be found along it. A list of such values can be found in Table 3.3. The most striking difference between homogeneous stars and polytropes with n = 1 is that the maximal redshift of the latter is a factor of 14 smaller. Because redshift depends so strongly on the stiffness of the equation of state, it could provide important clues as to the nature of matter under extreme conditions. The Newtonian limit, which marks the second end point of the massshed sequence, brings us full circle back to the static case. Class I for polytropes with n = 1 has thus been staked out and can be seen along with the second class in Figs 3.29 and 3.30. In the ﬁrst of these ﬁgures, one can see that the massshed sequence, starting from the Newtonian limit, reaches a minimum value along the yaxis, curves upward, ‘overshooting’ the pc = ∞ mark, before doubling back on itself again (a process which, like the inward spiral in Fig. 3.28 may be repeated ad inﬁnitum). Although the boundary curves are depicted, the possible solutions in Class I do not all reside within these curves in the parameter space chosen in Fig. 3.29. Consider, for example, the sequence generated by holding the value for the central pressure constant such that it meets the massshedding sequence at its smallest value for M 2 2 /z 3 . The curve described by this sequence lies below that of inﬁnite central pressure. In contrast, Class I conﬁgurations do all lie within the boundary curves of Fig. 3.30. The impression of a shrinking area for Class I that one gains by comparing this ﬁgure with Fig. 3.25 is thus not entirely unwarranted: All Class I stars for polytropes with n = 1 have a radius ratio greater than 0.5 whereas this ratio can be less than 0.2 for homogeneous stars. Turning our attention now to the second class, one can see that the gap between it and the ﬁrst class, which was illustrated in Fig. 3.15, is very pronounced in Figs 3.29 and 3.30. In fact, the entirety of the second class is made up of ring conﬁgurations, there being no transition point from toroidal to spheroidal topologies for polytropes with n > ∼ 0.36 (see also Fig. 3.35 below). As with the homogeneous second class,
3.4 Conﬁgurations with other equations of state
159
Fig. 3.29. The boundary curves for the ﬁrst two solution classes for polytropic conﬁgurations with n = 1. The dimensionless parameter M 2 2 /z 3 was chosen because it allows a representation of both the Newtonian sequence and that of inﬁnite central pressure. The signiﬁcation of the line types is given in the legend of Fig. 3.25.
Fig. 3.30. The boundary curves for the ﬁrst two solution classes for polytropic conﬁgurations with n = 1. This plot can be compared with Fig. 3.25, the legend of which also clariﬁes the meaning of the line types.
the limiting curves are here too made up of Newtonian, massshed and extreme Kerr sequences. The single (nonbifurcating) Newtonian curve means that there are a total of three as opposed to ﬁve boundary sequences however. Newtonian polytropic rings with n = 1 can be described analytically by making use of an expansion about the thin ring limit (Ostriker 1964, Petroff and Horatschek 2008),
160
Numerical treatment of the general case
but become increasingly inaccurate as one approaches the massshedding end of the sequence. The massshedding sequence runs between the Newtonian and the extreme Kerr limit. This third limiting sequence, for which the ‘interior’ and nonasymptotically ﬂat ring metric separates from the ‘exterior’ extreme Kerr metric, completes the triangle. The solution space for polytropes with arbitrary indices Acquiring as detailed a picture of the solution space for polytropic bodies as was possible for homogeneous ones would require an enormous amount of work. Here we merely intend to list some of what is known in order to gain an understanding of some of it. For one thing, we know from static Newtonian theory that polytropic stars with n > 5 do not exist and can assume that relativistic conﬁgurations, at least in a neighbourhood of this limit, do not exist either. We know further that for n > 3 the binding energy is negative, thus shedding light on their astrophysical relevance. It is interesting to note that rings with n > 3 do have positive binding energies however. As was mentioned on page 155, the transition from n = 0 to higher n in the Newtonian limit is similar to an increase in some relativistic parameter. In particular, the pinching together that is seen along Newtonian bifurcation sequences can also be found by increasing the polytropic index. An example of the shape of such a conﬁguration can be found in Fig. 3.31. What is apparent in this picture, is that the pinching together takes place at the innermost lobe as opposed to the outermost one as is the case for the homogeneous bifurcation sequences. Starting from a ‘spheroidallike’ basis conﬁguration and increasing n leads to a conﬁguration that is both at the massshedding limit and at the twobody transition limit. For the third class, this double limit is reached for n ≈ 1.2 meaning that there exists no third class for n > ∼ 1.2. As one proceeds to higher odd classes, the value of n for which one reaches the double limit decreases slightly and has fallen to a value of n ≈ 1.1 by the time one has reached Class XXI. It seems likely that similar behaviour is exhibited for ring topologies, so that one can conjecture that a ﬁnite number of classes exist for polytropes with n > n0 , where the index n0 may or may not be arbitrarily small, but where the upper limit n0 < 1.1 holds.
Fig. 3.31. Meridional crosssection of conﬁguration from Class IX with n = 1.
the
Newtonian
massshedding
3.4 Conﬁgurations with other equations of state
161
3.4.2 Completely degenerate, ideal gas of neutrons In order to present numerical results for an ideal neutron gas, it will be useful to make use of dimensionless quantities and we shall thus introduce an appropriate constant. Naming the constant of Equations (1.48) K˜ n :=
m4n , 24π 2 3
the series expansion for the pressure of (1.48a) about x = 0 reads 8 5 7 p = K˜ n x + O(x ) . 5
(3.54)
(3.55)
In the limit x → 0, the relation −2/3 K˜ n 5/3 p= µB 20
(3.56)
1 ˜ −2/3 Kn , 20
(3.57)
then follows. Introducing Kn :=
a constant has been deﬁned that will be used to render quantities dimensionless such that, in the Newtonian limit, the values for physical quantities for the completely degenerate neutron gas tend to the dimensionless analogues for a polytrope with n = 3/2. The constant Kn has, of course, the same unit as the polytropic constant K associated with polytropes with n = 3/2. We now present a brief overview of the ﬁrst class of solutions and shall mention the second one in the context of Subsection 3.4.3. The precise deﬁnition of a class of solutions for a completely degenerate, ideal Fermi gas follows naturally from the deﬁnition for polytropes. Because the basis conﬁgurations introduced on page 155 are located at the Newtonian limit, the equation of state being considered here coincides with that of the polytropes. The set of conﬁgurations connected via a continuous transition to a given basis conﬁguration is then said to comprise the associated class of solutions. The qualitative features of the ﬁrst class of solutions are much like those of polytropes with n = 1. The four boundary sequences are once again the static one, that of inﬁnite pressure, the massshedding and the Newtonian sequence. A plot of the square of the angular velocity divided by the central density as it depends on radius ratio can be found in Fig. 3.32 for the boundary curves. Here some of the detail exhibited by the massshedding sequence as one approaches the curve of inﬁnite central pressure is provided by the blowup. As a point of technical interest, it should
162
Numerical treatment of the general case
Fig. 3.32. The boundary curves for the ﬁrst solution class for conﬁgurations made up of a completely degenerate, ideal gas of neutrons. The blowup in the centre shows some of the detail of the curves in the vicinity of inﬁnite central pressure. This plot can be compared with Fig. 3.30.
be noted that it is necessary to span fourteen orders of magnitude in central pressure in order to produce Fig. 3.32. As with polytropes, the values for mass, angular velocity and redshift do not increase monotonically as one increases central pressure along sequences with constant angular momentum. They are however again maximal along the massshedding curve and some values are listed in Table 3.4. Unlike the constants of other equations of state discussed in this book, Kn of (3.57) is ﬁxed and thus enables us to provide dimensional values for physical quantities without having to make somewhat arbitrary choices.5 One ﬁnds, for example, that this model is capable of accounting for observed rotational periods, but not observed masses for neutron stars, which exceed 0.78 M . The maximal redshift is quite small compared to that of homogeneous stars and the increase in maximal mass due to rotation is only about 10%. A much greater value for this increase in maximal mass is found for the MIT bag model to be discussed in the next subsection. 3.4.3 Strange Matter The strange matter equation of state, = 3p + 4B,
(3.58)
5 It may be argued that the value for the MIT bag constant B of Equation (1.50) follows from the model within
fairly narrow margins. Since other models for strongly interacting matter can lead to equations of state with the same form, but radically different values for the constant, we choose not to specify its value further (see page 163 for a reference to other derivations of equations like (1.50)).
3.4 Conﬁgurations with other equations of state
163
Table 3.4. Extremal values for physical quantities are provided for conﬁgurations made up of a completely degenerate, ideal gas of neutrons, where ‘extremal’ refers here to a global maximum (minimum for the rotational period) for Class I. Physical quantity
Value
Gravitational mass Gravitational mass Baryonic mass Rotational period
Mstatic = 0.236 Kn −3/4 M = 0.260 Kn −3/4 M0 = 0.270 Kn −3/4 2π/ = 0.196 Kn
Redshift
A −3/4
z = 0.216
= 0.710 = 0.782 = 0.812 = 0.476
M M M ms
1.000 0.620 0.620 0.629 0.624
based on the MIT bag model and introduced in (1.50), is a conceivable model for quark matter in a deconﬁned state and presents an alternative to more traditional models for matter in dense stars. The same equation of state was considered for entirely different reasons by Sen (1934), who applied it to static stellar models. A detailed study of much of what we call the ﬁrst class was made by Gourgoulhon et al. (1999), who also provide references to earlier work on the subject. Similar equations of state, but with different values for the proportionality factor between and p, and with different expected values for the constant B, have been derived by different means, see e.g. Peshier et al. (2000). One can see that at a conﬁguration’s surface p = 0, the energydensity has a discontinuity, as in the homogeneous case. Furthermore, in the Newtonian limit p/ → 0 the equation of state reduces to = constant, meaning that the Maclaurin sequence and the Newtonian bifurcation sequences as well as the disc limit presented in Section 2.1 are all to be found as the Newtonian limits of the strange matter conﬁgurations. If in a generic class containing three Newtonian sequences, the two relativistic adjoining sequences meet, then the class structure here will be identical to that for homogeneous matter. That is not to say, however, that the properties of the member constituents need be similar in a highly relativistic regime. Figure 3.33 represents the counterpart to Fig. 3.17 and shows the boundary curves for the ﬁrst class of solutions as well as the entire Maclaurin sequence. The normalized square of the angular velocity 2 /16π B is plotted here versus the radius ratio A and is not divided by the central density as in Figs 3.32 and 3.30. Thus this plot clearly reﬂects the fact that does not reach its maximum along the sequence of inﬁnite pressure. Its maximal value is to be found for a central pressure pc /B =: p˜ c ≈ 64 along the massshedding curve. Curves of constant central pressure approach the limiting sequence p˜ c = ∞ in an oscillatory manner
164
Numerical treatment of the general case
Fig. 3.33. The dependence of 2 /16πB on A is shown for the boundary curves of the ﬁrst class of strange matter conﬁgurations. In addition, various sequences with constant central pressure p˜ c := pc /B are plotted using dotted lines. Line types are the same as those used in Fig. 3.14, with the exception of the massshedding curve in the inset, which is drawn using a solid line to render greater detail.
as shown vividly by the blowup. As with other equations of state considered here, further physical parameters such as angular momentum, mass, baryonic mass and redshift oscillate along the massshedding curve. An example of the typical inspiral behaviour along this sequence is provided in Fig. 3.34 for redshift z. The maximal values are listed in Table 3.5. It is notable that the increase in the maximal mass due to rotation is 44% and thus signiﬁcantly greater than for any other equation of state considered in this book, including homogeneous matter. The second class of solutions naturally possesses the same three Newtonian limiting sequences as with homogeneous matter. The pertinent piece of the Maclaurin sequence joins on to two bifurcating sequences ending in a massshedding limit in the one case and an inﬁnitely thin ring in the second (Fischer et al. 2005, Labranche et al. 2007). Emanating from these two points are the massshedding and extreme Kerr sequences, both of which meet at a point, thus enclosing the class. But for the fact that the three Newtonian curves merge to a single one for polytropes and for a completely degenerate ideal gas of neutrons, these qualitative features are shared by all the equations of state considered in this book. In particular, the existence of a transition to a black hole is a generic feature. A plot of the massshedding sequence running from the Newtonian to the black hole limit can be found for a variety of equations of state in Fig. 3.35. One can see that, at least in the parameter pair (V0 , A), the massshedding sequence for homogeneous rings
3.4 Conﬁgurations with other equations of state
165
Fig. 3.34. The inspiral behaviour of the redshift z versus the radius ratio A can be seen for the massshedding sequence of strange matter conﬁgurations.
Table 3.5. Maximal values for physical quantities are provided for strange matter stars, where ‘maximal’ refers here to a global maximum for Class I. Physical quantity
Value
Gravitational mass Gravitational mass Baryonic mass Angular velocity Angular momentum Redshift
Mstatic = 0.0258 M = 0.0372 h(0)M0 = 0.0444
= 4.72 J = 0.00123 z = 0.890
A B−1/2 B−1/2 B−1/2 B1/2 B−1
1.000 0.468 0.471 0.540 0.456 0.505
differs only very slightly from that for strange matter rings. Such rings, as well as polytropes with sufﬁciently small polytropic index n, exhibit a change of topologies. The massshedding curve indicates a boundary for A at a given V0 so that only the shaded region to the left of it is accessible. For every equation of state plotted here there exists a shaded region near the black hole limit. Rings approaching this limit can be considered to be sources of the extreme Kerr metric. The metric generated by them becomes arbitrarily close to that of the extreme Kerr black hole exterior to its horizon.
166
Numerical treatment of the general case
Fig. 3.35. Massshedding sequences from the second class for various equations of state are shown. All of them connect the Newtonian with the extreme Kerr black hole limit (after Fischer et al. 2005).
3.5 Fluid rings with a central black hole The ﬁgures of equilibrium that have occupied our attention until now have been single objects, namely discs, stars, rings or black holes. Through the detailed exploration of the solution space for homogeneous bodies, we were led naturally to the verge of a twobody problem, for example a star with a surrounding ring in Subsection 3.3.4, and chose this to mark a boundary in solution space. Although twobody systems are not what one ﬁrst thinks of upon hearing ‘ﬁgures of equilibrium’, they can be of considerable interest and will be considered here with a furtive glance toward more distant horizons. It is fairly certain that systems in equilibrium must be axially symmetric (see Section 1.2). In Newtonian theory, spheroidal bodies in axisymmetry cannot be stationary when aligned along the axis of symmetry. In general relativity, all attempts to ﬁnd multiple (uncharged) bodies aligned along the axis and kept in equilibrium by gravitomagnetic effects have failed. When considering stationary, twobody systems, one is thus led quite naturally to a central ﬁgure surrounded by a ring. In the remainder of this chapter we study such systems. The central object considered here will either be a Newtonian point mass, a black hole or an inﬁnitely ﬂattened, rigidly rotating disc of dust. Spacetimes with central black holes are of interest concerning the study of (i) the collapse of a single neutron star to a black hole and (ii) the coalescence of two compact objects, for it is expected that such systems exist, if only for a short time, see e.g. Shibata et al. (2003). In addition to astrophysical motivations, there is also interest in studying a blackhole–ring system in order to see how matter affects the properties of the black hole.
3.5 Fluid rings with a central black hole
167
Studies of systems consisting of a central rotating black hole and surrounding matter have been carried out by various authors (Will 1974, 1975, Abramowicz et al. 1983, 1984, Bodo and Curir 1992, Rezzolla et al. 2003, Montero et al. 2004, Zanotti et al. 2005, Lanza 1992). Using a formulation of Einstein’s equations in terms of integral equations, Nishida and Eriguchi (1994) numerically solved the problem of a differentially rotating polytropic perfect ﬂuid ring surrounding a black hole. The methods used (about ﬁfteen years ago now) did not allow for an accuracy high enough to resolve the impact of the matter distribution on the black hole completely, and the authors were misled into making incorrect conjectures regarding the shape of black holes with zero angular momentum, see Ansorg and Petroff (2005). Horizon and disc boundary conditions We begin by revisiting the horizon boundary conditions (1.129) and adding corresponding expressions for the remaining metric potentials. For disc–ring systems, a similar introduction of the disc boundary conditions will be provided. The boundary conditions that hold for a stationary, axisymmetric, asymptotically ﬂat spacetime containing a black hole and a ﬂuid with purely rotational motions were discussed lucidly and at length in Bardeen (1973a). Let, in our coordinates, the central black hole’s horizon be described by a constant radius r = rh , with spherical coordinates (r, ϑ) introduced through = r sin ϑ, ζ = r cos ϑ. By virtue of the ﬁeld equations (3.2), the regularity requirements of the metric functions imply the following boundary conditions valid at the horizon r = rh : W = Brh sin ϑ = 0
⇒
B = 0,
(3.59)
ω = h = constant ,
(3.60)
1 ∂u = , ∂r rh ∂α 1 =− . ∂r rh
(3.61) (3.62)
Note that in contrast to ν, the auxiliary metric function u = ν −ln B remains regular at the horizon. It is furthermore possible to introduce an additional constant κ deﬁned through the following horizon values: −α ∂ ν κ = e e . (3.63) ∂r r=rh This constant is called the surface gravity of the horizon and describes a rescaled gravitational acceleration of a zero angular momentum observer on the horizon, see
168
Numerical treatment of the general case
Bardeen (1973a). We characterize a degenerate nonvanishing black hole (a black hole with degenerate horizon) by κ = 0 with ﬁnite horizon area A. For a system consisting of a rigidly rotating central disc of dust and a surrounding ring of matter, the conditions for the metric, valid within the disc, are given in Section 1.7.3. In terms of the coefﬁcients (α, B, ω, ν) they read: e2V = e2ν (1 − v 2 ) = e2Vd = constant ,
(3.64)
B,ζ = 0 ,
(3.65)
4( d − ω)ν,ζ + (1 + v 2 )ω,ζ = 0 ,
(3.66)
(1 − v 2 )ν,ζ + (1 + v 2 )α,ζ = 0 .
(3.67)
Here d is the constant angular velocity of the disc and Vd the constant value of the function V deﬁned in (1.19). This value is related to the relative redshift of zero angular momentum photons emitted from the disc’s surface and received at inﬁnity, see (1.28). A Newtonian ring surrounding a point mass Before turning our attention to a black hole surrounded by a ring, it is natural to study a related Newtonian problem ﬁrst. As discussed in Ansorg and Petroff (2005), a Newtonian, rigidly rotating testring of ﬁnite size cannot exist in equilibrium. Since we know that rings without a central body exist within Newtonian theory (Kowalewsky 1885, Poincaré 1885, Dyson 1892, 1893, Wong 1974, Eriguchi and Sugimoto 1981, Ansorg et al. 2003c), there must exist a maximum for the ratio of the mass of the central body to that of the ring if the ring is to remain a ﬁnite size. One would expect the gravitational pull towards the central object to grow ever stronger, until massshedding at the inner edge sets in, i.e. until the gradient of pressure at the inner edge of the ring in the equatorial plane vanishes and a cusp develops, marking the point at which a ﬂuid element is about to be pulled away from the ring. This is indeed what is observed. In Fig. 3.36, a sequence of rings about a point mass is shown for an increasing ratio of the central to the ring mass Mc /Mr . The ratio of the inner to outer radius of the ring was held constant at the value A = −0.6 and the total (normalized) mass of the system was taken to be √ Mtot µ = 1. If we consider the sequence of conﬁgurations at the inner massshedding limit and with constant total mass, then we can vary a third parameter such as Mc /Mr . In the limit for which this ratio of masses goes to zero (i.e. when the point mass vanishes), we arrive at the conﬁguration denoted by ‘∗’ in Fig. 3.19. As described there, such a conﬁguration can be found along the sequence bifurcating from the
3.5 Fluid rings with a central black hole
169
Fig. 3.36. Crosssections of Newtonian rings surrounding a point mass with varying ratios of central to ring mass Mc /Mr . The normalized coordinate ζ /o is plotted against /o . For each of these conﬁgurations, the ratio of inner to outer radius of the √ ring was chosen to √ be A = −0.6 and for the normalized total mass we took Mtot µ = (Mc + Mr ) µ = 1 (after Ansorg and Petroff 2005).
Maclaurin spheroid with an eccentricity of = 0.98523 . . . and marks the transition from a spheroidal to a toroidal topology. Presumably, there is no upper limit to the value of Mc /Mr that can be reached. However, this testring limit could only be reached if the ring were not of ﬁnite size, i.e. in the limit Mc /Mr → ∞ it follows that A → −1. The crosssections for conﬁgurations with an inner massshed can be seen in Fig. 3.37. Negative Komar mass of central objects For stationary blackhole–ring and disc–ring systems, it is possible to assign a Komar mass and angular momentum to each of the two objects, see Komar (1959) and Bardeen (1973a). In particular, these expressions are obtained by considering the integrals (1.57) over a spacelike volume V ⊂ containing only the object in question. By virtue of the ﬁeld equations (3.2), these volume integrals can be rewritten in terms of surface integrals over the boundary ∂ V . In this manner it is also possible to assign the Komar quantities to a black hole. For a spacetime containing only a single object, this deﬁnition of the Komar mass coincides with that of the
170
Numerical treatment of the general case
Fig. 3.37. Crosssections of Newtonian rings surrounding a point mass with varying ratios of central to ring mass Mc /Mr . The normalized coordinate ζ /o is plotted against /o . Each of these conﬁgurations possesses √ √ an inner massshed and has a normalized total mass of Mtot = (Mc + Mr ) = 1 (after Ansorg and Petroff 2005).
gravitational mass. For a system with two or more objects, the gravitational mass is equal to the total Komar mass, which is the sum of the individual Komar masses.6 If one deals with the Komar mass, a natural question concerns its positive deﬁniteness. We address this question by analysing the following formulae Mh =
κA + 2 h Jh 4π
(3.68)
and Md = eVd M0 + 2 d Jd ,
(3.69)
valid for central black hole and central disc conﬁgurations respectively. 6 Note that some authors use the term Komar mass exclusively for what we call here the total Komar mass. The
total Komar mass must, of course, obey the positive mass theorem (Schoen and Yau 1979).
3.5 Fluid rings with a central black hole
171
In Equation (3.68), the central black hole’s Komar mass Mh is related to (i) its surface gravity κ, (ii) its horizon area A, (iii) the angular velocity h of the horizon, and (iv) the black hole’s (Komar) angular momentum Jh . For single black holes, Equation (3.68) was given by Smarr (1973), but it holds true even in the presence of a surrounding ring (Bardeen 1973a, see also Carter 1973). A similar expression can be derived for rigidly rotating discs of dust with and without a surrounding ring of matter, cf. (2.237). In Equation (3.69) the disc’s Komar mass Md is given in terms of (i) the constant eVd , (ii) the baryonic mass M0 of the central disc, (iii) its angular velocity d , and (iv) its (Komar) angular momentum Jd . The ﬁrst summands on the right hand sides of formulae (3.68) and (3.69) are always positive. However, each of these terms can become small if we assume the horizon area and the baryonic mass to be ﬁnite and consider the central object to be close to a degenerate black hole. As will be discussed below, we ﬁnd a continuous transition from the central disc to the central black hole conﬁgurations (Ansorg and Petroff 2006), and at the transition point, the central object is a degenerate black hole for which the ﬁrst terms in equations (3.68, 3.69) vanish. For the discussion of the sign of the second summands, a ‘framedragging’ effect of the central object caused by the surrounding ring is important. If the torus is highly relativistic and quickly rotating, it creates a large ergosphere, see Subsection 1.6.2. In this case, a counterrotating central object (i.e. the sign of its angular momentum is opposite to that of the torus) inside the ergosphere is dragged along the direction of the motion of the ring’s ﬂuid elements. As a consequence, the corresponding angular velocity of the central object can assume the same sign as that of the surrounding ring, and thus the second summand becomes negative. Combining the two arguments, it is possible to identify negative Komar masses by considering central objects close to a degenerate black hole and counterrotating with respect to the torus. Note that only highly relativistic and quickly rotating tori will exert a sufﬁciently large framedragging effect to bring this about. Moreover, the speciﬁc rate of counterrotation must be limited since very strong counterrotation would lead to opposite signs of the two angular velocities, h/d and ring , and hence to a positive second summand. In Fig. 3.38 we display sequences of both central black hole and central disc conﬁgurations along which the Komar mass of the central object becomes negative. Along the sequence the ratio of the circumferential inner radius Rcirc = i circ gϕϕ ( = i , ζ = 0) to that of the outer one Ro = gϕϕ ( = o , ζ = 0) was circ = 0.85. Furthermore, the outer massshed held constant at the value Rcirc i /Ro parameter from Equation (3.51) was chosen to be β = 0.3 and as a third parameter d /o = 0.1 was chosen for the disc or rh /o = 0.1 for the black hole.
172
Numerical treatment of the general case
Fig. 3.38. On the lower left, the ratio of the Komar mass of the central object to that of the ring is plotted versus the ring’s redshift zr for a sequence with circ Rcirc i /Ro = 0.85, β = 0.3 and d /o = 0.1 for the disc or rh /o = 0.1 for the black hole (see text for an explanation of the symbols). Knowing that Mr remains positive, one can see that Md/h becomes negative. The lower right shows a similar plot, but containing the disc’s baryonic mass and the square root of the horizon area. Above the plots, the coordinate shape of the ring and central object (solid lines) and their ergospheres (dotted lines) are shown for these sequences (after Ansorg and Petroff 2006).
3.5 Fluid rings with a central black hole
173
The evolution of the coordinate shape of the ring, the central object and their ergospheres can be followed in Fig. 3.38. Looking ﬁrst at the series of pictures on the left, we begin with a fairly ‘Newtonian’ ring and can see that only the disc possesses an ergosphere. As explained above, the ring and the disc must be counterrotating, i.e. their angular momenta have opposite signs. For weakly relativistic rings with a small inﬂuence on the central object, the signs of the central angular momentum and velocity coincide, resulting in a positive Komar mass for the central object. However, as the ring becomes increasingly relativistic and develops an ergosphere, its framedragging effect on the disc becomes more pronounced, causing the disc’s angular velocity to decrease, whence its ergosphere shrinks and ﬁnally vanishes. Going even further, the framedragging ﬁnally forces the disc to corotate with the ring although its angular momentum still has the opposite sign. Relative to the size of the ring, the ergosphere grows very large, which is why we show only a portion of its boundary in the last two pictures in the sequence. From an outside observer’s perspective, the conﬁguration is shrinking toward the centre and the outside metric is beginning to resemble that of the extreme Kerr metric. The ring’s ergosphere continues to grow, ﬁnally engulﬁng the disc. After a good portion of the disc ﬁnds itself inside the ergosphere, the frame dragging becomes signiﬁcant enough that the magnitude of d Jd is sufﬁciently large to result in a negative mass. The series of pictures on the right is the counterpart for a black hole and shows similar behaviour to the disc case. A black hole with nonvanishing h always has an ergosphere surrounding it however. Its sense of rotation must agree with that of the ring before their ergospheres can merge, independent of the sign of Jh . We now devote our attention to the behaviour of the Komar mass in the parametric transition from the central disc of dust to a black hole. A generalization of the proof given in Section 1.8.3 (see also Meinel 2004, 2006) implies that such a transition exists if and only if Vd tends to −∞, which in turn implies that Md = 2 d Jd must hold.7 The equality of (3.69) and (3.68) then requires for nonvanishing black holes that κ = 0. The plot on the lower left of Fig. 3.39 shows that such transitions do indeed exist. Here the Komar mass ratio was chosen as an exemplary parameter and plotted versus a measure of the distance to the transition point representing a degenerate black hole surrounded by a ring. The plot on the lower right suggests a very similar transition to the one known analytically for the rigidly rotating disc of dust without a ring as can be seen by comparing it to Fig. 2.17. The picture sequence in Fig. 3.39 shows the evolution of the coordinate shapes of these conﬁgurations. The results presented reveal clearly that the Komar mass is not an intrinsic property of a gravitational source but rather a feature of an object within a highly 7 The generalization of the arguments in Meinel (2006) requires the assumption that −ξ i u as deﬁned there is i bounded from below. In the presence of a surrounding ring, ηi ui need not have the same sign as d .
174
Numerical treatment of the general case
Fig. 3.39. On the lower left, the ratio of the Komar mass of the central object to that of the ring is plotted versus a measure of the distance to the degenerate black circ hole solution. The sequences are deﬁned by Rcirc i /Ro = 0.85, β = 0 for the outer massshed parameter and V0r = −2.7 for the ring’s corotating surface potential. On the lower right, a plot similar to Fig. 2.17 is shown for the transition. Up above, the coordinate shape of the ring and central object (solid lines) and their ergospheres (dotted lines) are drawn. The framed picture indicates the transition point from the disc to the black hole (after Ansorg and Petroff 2006).
3.5 Fluid rings with a central black hole
175
relativistic spacetime geometry. The discrepancy between the sign of h/d and Jh/d , which is responsible for the negative Komar masses, is a result of the fact that the central body’s motion is determined by the local environment, whereas the angular velocity refers to rotation ‘with respect to inﬁnity’ (see Section 1.1). Since the individual Komar mass is not a good measure of the intrinsic attributes of a local object, it is natural to consider other local mass deﬁnitions. An interesting candidate for such a deﬁnition in the case of black holes is the Christodoulou mass (Christodoulou 1970), which will be discussed shortly. Black holes with degenerate horizon It has been shown by Ansorg and Pﬁster (2008) that the horizon area Ah and the angular momentum Jh of a degenerate black hole (κ = 0) are related by 8πJh  = Ah ,
(3.70)
even in the presence of a surrounding ﬂuid ring. The proof runs as follows: First, a new radial coordinate R is introduced rh2 1 r+ , (3.71) R= 2 r in which, as stated by Bardeen (1973a), pages 251–252, the following functions of the metric potentials are positive and regular with respect to R and cos ϑ in the vicinity of the black hole, even when the degenerate limit is encountered:8 µˆ = r 2 e2µ ,
(3.72)
uˆ = r 2 e−2u , r B. Bˆ = R2 − rh2
(3.73) (3.74)
The coordinate R penetrates the horizon, that is, spatial points with coordinate values R < rh are inside the horizon. Note that, in contrast, for any value r > 0 we obtain R ≥ rh , i.e. the coordinate r is not horizon penetrating. In terms of these functions, the surface gravity can be expressed as: " #−1/2 κ = rh Bˆ µˆ = constant . ˆu Hence a degenerate black hole with κ = 0 is characterized by rh = 0. It can now be shown that the Einstein equations, written for the degenerate limit in terms of the above regular functions in (R, ϑ) and evaluated at the horizon 8 The quantities introduced here are closely related to Bardeen’s expressions: h = 2r , λ = 2R, B = B/2. ˆ R h
176
Numerical treatment of the general case
R = rh = 0, do not contain terms in which Rderivatives are involved. As a consequence, it becomes possible to work out the horizon boundary values of all metric functions explicitly. If these expressions are inserted into the integral formulae for the black hole angular momentum and horizon area, one obtains (3.70), which is valid for axially and equatorially symmetric, stationary conﬁgurations consisting of a degenerate central black hole with surrounding matter. For more details of the proof, see Ansorg and Pﬁster (2008). Moreover, overwhelming evidence from numerical calculations is presented there to support the conjecture that 8πJh  ≤ Ah
(3.75)
holds true for any axially and equatorially symmetric, stationary conﬁguration consisting of a central black hole with surrounding perfect ﬂuid matter, and that, in particular, the equals sign holds if and only if the central black hole is degenerate. A wellknown relation for degenerate Kerr black holes is given by M 2 = J  ,
(3.76)
see Subsection 1.8.2. However, this relation is no longer true if M is taken to be the Komar mass Mh and we allow for matter surrounding the black hole. In fact, we have seen that the Komar mass can vanish for a rotating degenerate black hole with ﬁnite angular momentum Jh . It is interesting to note that (3.76) does hold however, by virtue of (3.70), even when the degenerate black hole is surrounded by additional matter, if one takes M to be the Christodoulou mass ) 4π Jh2 Ah MC = + , (3.77) 16π Ah which plays a fundamental role in the isolated and dynamical horizon formalism, see Ashtekar and Krishnan (2004) for an overview. Moreover, this mass parameter is being used widely in the ﬁeld of dynamical calculations of spacetimes containing black holes, e.g. for describing black holes in the centre of accretion discs (Font 2003, and references therein).
4 Remarks on stability and astrophysical relevance
In general, a relativistic ﬁgure of equilibrium as calculated in the previous chapters is to be expected to exist in nature only if it is in stable equilibrium. Therefore, in addition to other aspects, like realistic equations of state, magnetic ﬁelds and initial conditions, the investigation of stability properties is very important for identifying conﬁgurations that might be astrophysically relevant. A complete stability analysis of relativistic ﬁgures of equilibrium is extremely difﬁcult. Moreover, the stability depends on matter properties like viscosity and thermal conductivity, which are unimportant for the equilibrium state itself and therefore do not need to be speciﬁed in this book. Our intention, as expressed in the preface, is to ‘place emphasis on the rigorous treatment of simple models instead of trying to describe real objects with their many complex facets’ and, consequently, an extensive treatment of stability questions is beyond the scope of this book. Nevertheless, in the following we will discuss some aspects of stability of rotating ﬂuid conﬁgurations in general relativity.
Stability with respect to axisymmetric perturbations Friedman et al. (1988) have shown that a version of the turningpoint method going back to Poincaré (1885), who investigated the stability of Newtonian equilibrium conﬁgurations (cf. Subsection 3.3.1), can be used to locate points along sequences of relativistic ﬁgures of equilibrium at which secular instability1 with respect to axisymmetric perturbations sets in, see also Thorne (1967). The result can be formulated as follows: Whenever a continuous sequence of equilibrium conﬁgurations labelled by a parameter λ, with a given equation of state = (p) and constant angular momentum J , has an extremum of the gravitational mass M at a point λ = λ0 , this point marks the onset of instability. The unstable part of the 1 A ‘secular’ instability – in contrast to a ‘dynamical’ one – requires some dissipative mechanism, provided for
example by viscosity, in order to take effect.
177
178
Remarks on stability and astrophysical relevance
sequence, near λ0 , can be identiﬁed by the condition d2 M >0 dM02
(4.1)
or, equivalently, dV0 dM > 0. (4.2) dλ dλ (It is assumed that d[(dV0 /dλ)(dM /dλ)]/dλ = 0 at λ = λ0 .) This result can easily be understood by means of the relation (1.59) between variations of M , J and the baryonic mass M0 , specialized to a continuous sequence of equilibrium conﬁgurations, where all quantities depend on the single parameter λ, dJ dM0 dM = + h(0)eV0 . dλ dλ dλ
(4.3)
For a sequence of constant angular momentum J , an extremum in the gravitational mass M at some point λ = λ0 automatically results in an extremum in the baryonic mass M0 at the same point. This means that there exist two nearby points λ1 , λ2 of the sequence (λ1 < λ0 < λ2 ) for some value of M0 near its extremum. The conﬁgurations at these two points thus have the same M0 and J , but they differ in M . Provided that a suitable dissipative mechanism is available, the conﬁguration with larger M is unstable: it can evolve towards the other one as a result of small perturbations. Note that equilibrium ﬁgures are conﬁgurations, which, for prescribed M0 and J , extremize M in a welldeﬁned sense (Hartle and Sharp 1967). Stable equilibria must, of course, be minima of M . An axisymmetric time evolution does not change the total angular momentum J , but viscosity can lead to a redistribution of angular momentum (transfer from one baryonic ﬂuid ring to another). This process is essential for reaching an equilibrium state with its uniform rotation. The equivalence of (4.1) and (4.2) follows immediately from (4.3) and J = constant: dM dV0 dλ d2 M = h(0)eV0 ⇒ = h(0)eV0 2 dM0 dλ dM0 dM0 dV0 dM dM0 −2 = . dλ dλ dλ
(4.4)
This relation also shows that the second derivative d2 M /dM02 diverges at λ = λ0 and changes its sign, leading to a cusp in an M0 M diagram, see Fig. 4.1. Independent of the question of whether the turning point of M (and M0 ) is a maximum or a minimum, the larger M value for given M0 will be found on that
Remarks on stability and astrophysical relevance
179
Fig. 4.1. Schematic M0 M diagram of a J = constant sequence of equilibrium ﬁgures with a given equation of state near a minimum (left picture) or a maximum (right picture) of M and M0 .
Fig. 4.2. Partial sequences of constant angular momentum J for stars governed by the equation of state for a completely degenerate, ideal gas of neutrons. Moving 3/2 from the inside outward, the values of J /Kn are 0, 0.006, 0.012, 0.018, 0.024 and 0.03. For the meaning of the normalization constant Kn see (3.57). The long thicker line at the outside is a portion of the massshedding curve. The short thicker line marks the position of maximal mass for each sequence and is thus the boundary of the secularly unstable region, which is here marked by grey shading. One end of this boundary line, marked by a dot, corresponds to the J = 0 conﬁguration of maximal mass. The projections of this line onto two planes in the M0 − M − V0 space are also shown.
part of the sequence which satisﬁes (4.1). Note that the instability condition can also be written as dM > 0. dV0
(4.5)
180
Remarks on stability and astrophysical relevance
For dV0 /dλ = 0 at λ = λ0 , the parameter V0 can be used itself to label the equilibrium sequence near the turning point. If one determines the turning points of such J = constant sequences in dependence of J , a curve in the twodimensional manifold of solutions (equilibrium ﬁgures for the given equation of state) is obtained, which represents the boundary of the secularly unstable region.2 An example is given in Fig. 4.2. A sufﬁcient condition for dynamical stability with respect to small axisymmetric perturbations is the positivedeﬁniteness of a welldeﬁned energy functional E for all relevant perturbations (Chandrasekhar and Friedman 1972a,b, 1973, Schutz 1972), see also Friedman and Ipser (1992) – since E can only decrease as a consequence of (outgoing) gravitational radiation. (If positive deﬁnite, E plays the role of a Lyapunov functional.) A more practicable method to investigate dynamical stability can be the direct numerical simulation of the time evolution of slightly perturbed equilibrium conﬁgurations with a purely axisymmetric code. However, in general, a sequence of equilibrium ﬁgures becomes dynamically unstable after it has become secularly unstable, see Friedman et al. (1988). Therefore, the simple turningpoint criterion discussed above may be sufﬁcient to ﬁnd the limits of axisymmetric stability. Stability with respect to nonaxisymmetric perturbations A complete stability analysis must, of course, also include the consideration of nonaxisymmetric perturbations. A remarkable result is that all rotating perfectﬂuid equilibrium conﬁgurations are unstable to nonaxisymmetric perturbations by a mechanism that leads to a loss of angular momentum via gravitational radiation. This ‘Chandrasekhar–Friedman–Schutz (CFS) instability’ (Chandrasekhar 1970, Friedman and Schutz 1978) can, however, be damped by viscosity, see Detweiler and Lindblom (1977) and Lindblom and Hiscock (1983). Many more results on axisymmetric as well as nonaxisymmetric stability of relativistic stars can be found in Stergioulas (2003) and references therein. 2 Note that the same curve can also be obtained by considering turning points of M (and J ) for sequences
M0 = constant.
Appendix 1 A detailed look at the massshedding limit
A1.1 The differentiability of functions at the massshedding limit Since solutions to Poissonlike equations on domains containing corners are known to be nonanalytic in general, it will come as no surprise to learn that the functions involved in describing a massshedding conﬁguration are also not analytic. It is of interest, especially for the numerical methods used in this book, to be more precise. Ideally, we would be able to determine the asymptotic behaviour of the functions involved as one approaches a corner. We shall content ourselves in a ﬁrst analysis, however, with determining which derivatives become singular. The behaviour of solutions to Poissonlike equations on domains containing corners has been studied by Wigley (1970), Eisenstat (1974) and discussed in Birkhoff and Lynch (1984). In those analyses, it was always assumed that the domain boundaries are known whereas in our case, the (free) boundary arises from solving a global problem. Although we know that the crosssection of the surface of a massshedding body contains a corner, we do not know much, a priori, about the differentiability of its parametric representation ζb (). Nonetheless, in order to study the behaviour of the Newtonian potential U , let us imagine that the global problem has been solved, that the surface is known, and that a cusp (i.e. mass shedding, cf. page 26) exists at the point = 0 . Inside the body, we shall refer to the potential as Ui and outside as Uo and leave off the index when an expression is valid for either potential. We assume that the known functions Ui , Uo and ζb are C 2 and shall ﬁnd that this leads to a contradiction. Taking the derivative of (1.122) with respect to along the surface yields ∂ 2U ∂ 2 U dζb ∂ 2 U + + 2 ∂ ∂ζ d ∂2 ∂ζ 2
dζb d
2 +
∂U d2 ζb = 2 . ∂ζ d2
As a result of equatorial symmetry, we know that ∂ i+j U =0 for j odd. ∂i ∂ζ j ζ =0 181
(A1.1)
(A1.2)
182
A detailed look at the massshedding limit
Therefore, at the point = 0 , (A1.1) reduces to ∂ 2U ∂ 2 U dζb 2 = 2 , + ∂2 ∂ζ 2 d
(A1.3)
remembering that ζb is assumed to be a C 2 function. Making use of (1.124), the Poisson and Laplace equations at the point = 0 are ∂ 2 Ui ∂ 2 Ui + = 4πµ − 2 , ∂ζ 2 ∂2 ∂ 2 Uo ∂ 2 Uo + = −2 . ∂2 ∂ζ 2
(A1.4a) (A1.4b)
The system of equations made up of (A1.3) and (A1.4) leads to 22 − 4πµ ∂ 2 Ui = , 2 ∂ζ 2 dζb − 1 d
(A1.5a)
∂ 2 Uo 22 = . 2 ∂ζ 2 dζb −1 d
(A1.5b)
Discounting the possibility that = 0, since we do not expect massshedding to occur for nonrotating stars, we ﬁnd dζb 2 = 1. (A1.6) d The transition conditions at the surface of the star require that the ﬁrst derivatives of U be continuous. Choosing to look at the derivative of U with respect to ζ on the boundary at a point 1 not far from 0 , we can write ∂U (1 , ζb (1 )) ∂U (0 , 0) = ∂ζ ∂ζ 2 ∂ U (0 , 0) ∂ 2 U (0 , 0) dζb + + (1 − 0 ) + · · · ∂ζ ∂ d ∂ζ 2 =
∂ 2 U (0 , 0) dζb (1 − 0 ) + · · · . d ∂ζ 2
(A1.7)
Since this equation holds both for Ui and Uo , the ﬁrst derivative is only continuous if ∂ 2 Uo ∂ 2 Ui = ∂ζ 2 ∂ζ 2
(A1.8)
A1.1 The differentiability of functions at the massshedding limit
183
holds on the surface at the point 0 . We see from Equations (A1.5) that this can only be true if µ = 0 at this point. In other words, we have arrived at a contradiction for equations of state such as homogeneous or strange matter that allow for the density to jump discontinuously at the surface. For such conﬁgurations, the functions Ui , Uo and ζb cannot all be C 2 . For other equations of state, we can derive similar results by assuming higher differentiability of the functions. If, for example, we assume now that Ui , Uo and ζb are C 3 functions, and consider equations of state for which µ = 0 at the boundary, then we can proceed to differentiate (1.122) again with respect to . At the point 0 we then ﬁnd ∂ 3U ∂ 3U + 3 ∂3 ∂ ∂ζ 2
dζb d
2 +3
∂ 2 U dζb d2 ζb = 0. ∂ζ 2 d d2
(A1.9)
Differentiating the Laplace and Poisson equations with respect to and eliminating the ﬁrst and second derivatives with respect to using (1.124) and (A1.3) yields ∂ 3 Ui ∂µ d2 ζb ∂ 3 Ui ∂ 2 Ui dζb 1 dζb = 4π + 3 + − (A1.10a) ∂ d2 ∂3 ∂ ∂ζ 2 ∂ζ 2 d d d2 ζb ∂ 3 Uo ∂ 2 Uo dζb 1 dζb ∂ 3 Uo (A1.10b) + 3 2 = 0. + − ∂3 ∂ ∂ζ 2 ∂ζ 2 d d d The solution to the system of equations (A1.9) and (A1.10) implies1 that
dζb d
2 =
1 3
(A1.11)
based on the assumption that Ui and Uo are C 3 functions. Taking the next term in the expansion of (A1.7) at the point 0 and remembering the equality (A1.8), we then ﬁnd ∂ 3 Ui ∂ 3 Uo = . (A1.12) ∂ ∂ζ 2 ∂ ∂ζ 2 This cannot be fulﬁlled along with (A1.10) unless the derivative of µ with respect to vanishes at the equator. This derivative does not vanish for a polytrope with the polytropic index n = 1, for example, meaning that for such stars the potentials and the surface function cannot all be C 3 functions. 1 Strictly speaking, one also has to rule out the possibility that dζ /d = −3r d2 ζ /d2 . The ﬁrst and second e b b derivatives of ζb are both expected to be negative however, as can be veriﬁed for the concrete example of the
Roche model in Equation (A1.18).
184
A detailed look at the massshedding limit
The arguments presented above can be extended to C k functions by taking kth derivatives and ﬁnding contradictions for equations of state for which the (k −2)nd derivative of µ does not vanish at the equator.
A1.2 The massshedding limit in the Roche model If the mass of a star is highly concentrated in its centre, then one can expect that it can be modelled by considering a nongravitating ﬂuid (shell) in the ﬁeld of a point mass. In Newtonian theory, models of this sort are known as ‘Roche models’, based on work done by the French scientist Édouard Roche (1873) (see also Zel’dovich and Novikov 1971, Shapiro and Shibata 2002). As we shall see shortly, the massshedding limit is particularly easy to handle in this model. Taking Equation (1.120) and inserting the potential of a point particle, we ﬁnd that the surface of a ﬂuid rotating rigidly with the angular velocity is deﬁned by the equation M V0 + + 12 2 2 = 0. (A1.13) 2 + ζ2 By considering the point = 0, we can relate the constant V0 to the polar radius rp V0 = −
M . rp
(A1.14)
For massshedding stars, (1.124) leads to the relation M = re3 2 , and the surface equation (A1.13) becomes 1 2 rp = 1. + 2 + ζ 2 2re3
(A1.15)
(A1.16)
Evaluating this expression at the equator ζ = 0 tells us that for massshedding ﬂuids in the Roche model rp 2 (A1.17) = re 3 holds, independent of the equation of state. With this relationship, the curve describing the ﬂuid’s surface can be rewritten as
4re2 − 2 re2 − 2 ζb = , (A1.18) 3re2 − 2
A1.2 The massshedding limit in the Roche model
185
Table A1.1. Physical parameters of Newtonian polytropic stars with the polytropic index n at the massshedding limit. The second column provides a measure of how concentrated the mass is in the centre. The third through ﬁfth columns are to be compared with the corresponding values in the Roche model, see Equations (A1.15), (A1.17) and (A1.19) respectively. n
µ(/re = 0.1, ζ = 0) µc
re3 2
rp re
dζb ( = re ) d
0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.99 0.97 0.94 0.89 0.79 0.63 0.38
0.797 0.906 0.957 0.981 0.992 0.997 0.999
0.4380 0.5570 0.6151 0.6434 0.6569 0.6630 0.6655
−1.433 −1.606 −1.678 −1.709 −1.723 −1.729 −1.731
M
Fig. A1.1. The surface of a massshedding star according to the Roche model as given by (A1.18) is depicted in the /re − ζ /re plane.
where the index ‘b’ was added to conform with the notation in Subsection 1.7.4. It then follows that √ dζb = − 3, (A1.19) d + ζ =0
which means the interior angle of the massshedding cusp is 2π/3. The success of the Roche model in describing polytropic stars for large values of the polytropic index n is demonstrated in Table A1.1. The ﬁrst column in the table
186
A detailed look at the massshedding limit
indicates the value of n used and the second provides a measure of how highly the mass is concentrated in the star’s centre. The remaining three columns demonstrate that the values implied by (A1.15), (A1.17) and (A1.19) are approached for large n, and we see that deviations from the Roche model are on the order of about 0.1% for n = 3.5. Figure A1.1 depicts the shape of the massshedding star according to the Roche model [see (A1.18)]. This curve is indistinguishable from that of a polytropic massshedding star with n = 3.5. The difference between ζb /re for such a polytropic star and ζb /re of the Roche model is greatest at the north pole, the value of which can be found in Table A1.1.
Appendix 2 Theta functions: deﬁnitions and relations
In this appendix we provide some basic deﬁnitions and relations for functions that play an important role in Section 2.3. We shall not provide the related mathematical theory nor discuss theta functions in generality here. Starting from the (usual) theta functions, which were introduced by Jacobi, we follow the path taken by Rosenhain and deﬁne hyperelliptic theta functions of two variables, sometimes called ‘ultraelliptic theta functions’ (Rosenhain 1850). Moreover, we list some useful relations between theta functions of each type. Besides the deﬁnition of the wellknown elliptic integrals of the ﬁrst and second kinds and the Jacobian elliptic functions, we include two less wellknown functions, which can be constructed from elliptic integrals, namely Heuman’s lambda function and Jacobi’s zeta function. Furthermore, the relations between the Jacobian theta functions and the Jacobian elliptic functions that are important for our purpose in Subsection 2.3.3 are given. Finally, we list derivatives for some of the above functions that were used in Subsection 2.3.4. Note that the notation in the literature (especially concerning theta functions) is not standardized. Throughout this book we comply strictly with the deﬁnitions presented here. A2.1 Jacobian theta functions ∞ 2 n 1 ϑ1 (v ; B) := (−1) exp 2 (2n + 1) B + (2n + 1)v , n=−∞
ϑ2 (v ; B) :=
∞
exp
2
1 2 (2n + 1)
B + (2n + 1)v ,
n=−∞
ϑ3 (v ; B) :=
∞
(A2.1) exp{n2 B + 2nv },
n=−∞
ϑ4 (v ; B) :=
∞
(−1)n exp{n2 B + 2nv }.
n=−∞
187
188
Theta functions: deﬁnitions and relations
Selected properties of the Jacobian theta functions: iπ ϑ1 (v + iπ; B) = −ϑ1 (v ; B), ϑ1 v + 2 ; B = i ϑ2 (v ; B), ϑ2 v + iπ2 ; B = i ϑ1 (v ; B), ϑ2 (v + iπ; B) = −ϑ2 (v ; B), ϑ3 v + iπ2 ; B = ϑ4 (v ; B), ϑ3 (v + iπ; B) = + ϑ3 (v ; B), ϑ4 v + iπ2 ; B = ϑ3 (v ; B). ϑ4 (v + iπ; B) = + ϑ4 (v ; B),
(A2.2) (A2.3) (A2.4) (A2.5)
ϑ1 (v + B; B) = −e−(2v+B) ϑ1 (v ; B), ϑ2 (v + B; B) = e−(2v+B) ϑ2 (v ; B), ϑ3 (v + B; B) = e−(2v+B) ϑ3 (v ; B),
(A2.6)
ϑ4 (v + B; B) = −e−(2v+B) ϑ4 (v ; B). ϑ1 v + ϑ2 v + ϑ3 v + ϑ4 v +
B 2;B B 2;B B 2;B B ; B 2
= −e−(v+B/4) ϑ4 (v ; B), = e−(v+B/4) ϑ3 (v ; B), = e−(v+B/4) ϑ2 (v ; B),
(A2.7)
= e−(v+B/4) ϑ1 (v ; B).
It is convenient to introduce the following notation (which is systematic but not common!):
πu K(k ) 1 (u, k) := ϑ1 , −π , 2K(k) K(k)
K(k ) πu , −π , 2 (u, k) := ϑ2 2K(k) K(k) (A2.8)
K(k ) πu , −π , 3 (u, k) := ϑ3 2K(k) K(k)
K(k ) πu , −π . 4 (u, k) := ϑ4 2K(k) K(k) Special values for u = 0:
K(k ) 1 (0, k) = ϑ1 0, −π K(k)
K(k ) 2 (0, k) = ϑ2 0, −π K(k)
=
0,
=
2kK(k) , π
(A2.9)
A2.2 Rosenhain’s theta functions
3 (0, k) = ϑ3
K(k ) 0, −π K(k)
4 (0, k) = ϑ4
ϑi,k ϑ1,k
K(k ) 0, −π K(k)
=
=
189
2K(k) , π 2k K(k) . π
A2.2 Rosenhain’s theta functions ≡ ϑi,k (v , w; B11 , B22 , B12 ), (i, k ∈ {1, 2, 3, 4}), ∞ 2 m 1 := (−1) exp 2 (2m + 1) B11 + (2m + 1)v m=−∞
× ϑk w + 12 (2m + 1)B12 ; B22 , ϑ2,k :=
∞
exp
2
1 2 (2m + 1)
B11 + (2m + 1)v
m=−∞
× ϑk w + 12 (2m + 1)B12 ; B22 , ϑ3,k :=
∞
(A2.10)
exp{m2 B11 + 2mv }ϑk (w + mB12 ; B22 ),
m=−∞
ϑ4,k :=
∞
(−1)m exp{m2 B11 + 2mv }ϑk (w + mB12 ; B22 ).
m=−∞
Selected properties: ϑ3,3 (x1 , x2 ; B11 ± iπ, B22 ∓ iπ, B12 ) = ϑ4,4 (x1 , x2 ; B11 , B22 , B12 ), ϑ2,2 (x1 , x2 ; B11 ± iπ, B22 ∓ iπ, B12 ) = ϑ2,2 (x1 , x2 ; B11 , B22 , B12 ), π ϑ4,4 x1 ± i , x2 ; B11 , B22 , B12 = ϑ3,4 (x1 , x2 ; B11 , B22 , B12 ), 2 π ϑ4,4 x1 , x2 ± i ; B11 , B22 , B12 = ϑ4,3 (x1 , x2 ; B11 , B22 , B12 ), 2 π ϑ2,2 x1 ± i , x2 ; B11 , B22 , B12 = ± i ϑ1,2 (x1 , x2 ; B11 , B22 , B12 ), 2 π ϑ2,2 x1 , x2 ± i ; B11 , B22 , B12 = ± i ϑ2,1 (x1 , x2 ; B11 , B22 , B12 ). 2
(A2.11)
(A2.12)
Separation property: ϑi,k (v , w; B11 , B22 , B12 = 0) = ϑi (v ; B11 )ϑk (w; B22 ).
(A2.13)
190
Theta functions: deﬁnitions and relations
A2.3 Elliptic integrals and functions, relations to theta functions Elliptic integrals of the ﬁrst and second kinds:
y
F(ϕ, k) := 0
y
E(ϕ, k) := 0
dt
=
ϕ
dϑ , 1 − k 2 sin2 ϑ
0 (1 − t 2 )(1 − k 2 t 2 )
ϕ 1 − k 2t2 dt = 1 − k 2 sin2 ϑdϑ. 1 − t2 0
The number k is called the modulus whereas k := complementary modulus.
(A2.14)
√ 1 − k 2 is referred to as the
Complete elliptic integrals: K(k) := F
π 2
,k ,
E(k) := E
π 2
,k .
(A2.15)
Legendre’s relation: E(k)K(k ) + E(k )K(k) − K(k)K(k ) =
π . 2
(A2.16)
Jacobian elliptic functions: The elliptic function sn(u, k) was introduced as the inverse function of the elliptic integral of the ﬁrst kind. The functions cn and dn are closely related to sn.
u= 0
y
dt
= (1 − t 2 )(1 − k 2 t 2 )
0
ϕ
dϑ = F(ϕ, k), 1 − k 2 sin2 θ
sn(u, k) := y = sin ϕ, am(u, k) := ϕ, cn(u, k) := 1 − y2 = cos ϕ, 2 2 dn(u, k) := 1 − k y = 1 − k 2 sin2 ϕ.
(A2.17)
(A2.18)
Thus we have the basic identities sn2 (u, k) + cn2 (u, k) = 1, k 2 sn2 (u, k) + dn2 (u, k) = 1, dn2 (u, k) − k 2 cn2 (u, k) = k 2 , k 2 sn2 (u, k) + cn2 (u, k) = dn2 (u, k).
(A2.19)
A2.4 Selected derivatives
191
The following relations to the Jacobian theta functions are valid: K(k ) πu ; −π ϑ K(k) i 1 2K(k) i 1 (u, k) = −√ sn(−iu, k) = − √ , πu kϑ k 4 (u, k) ; −π K(k ) 4
cn(−iu, k) =
k
ϑ2
k ϑ 4
2K(k)
(A2.20)
K(k)
K(k ) πu 2K(k) ; −π K(k) K(k ) πu 2K(k) ; −π K(k)
=
K(k ) πu √ ϑ3 2K(k) ; −π K(k) dn(−iu, k) = k ) πu ϑ4 2K(k) ; −π K(k K(k)
=
k 2 (u, k) , k 4 (u, k)
(A2.21)
√ 3 (u, k) . k 4 (u, k)
(A2.22)
Heuman’s lambda function: 0 (ψ, k) :=
2 E(k)F(ψ, k ) + K(k)E(ψ, k ) − K(k)F(ψ, k ) . π
(A2.23)
Jacobian zeta function: Z(u, k) := E(β, k) −
E(k) F(β, k) K(k)
[with: β ≡ am(u, k)].
(A2.24)
A2.4 Selected derivatives Derivatives of the elliptic integral of the ﬁrst kind: 1 ∂ F(ϕ, k) = , ∂ϕ 1 − k 2 sin2 ϕ ∂ E(ϕ, k) − k 2 F(ϕ, k) − F(ϕ, k) = ∂k kk 2
k sin ϕ cos ϕ . 2 2 2 k 1 − k sin ϕ
Derivatives of the elliptic integral of the second kind: ∂ E(ϕ, k) = 1 − k 2 sin2 ϕ, ∂ϕ E(ϕ, k) − F(ϕ, k) ∂ E(ϕ, k) = . ∂k k
(A2.25)
(A2.26)
In the next equations we use the more concise notation am u ≡ am(u, k), sn u ≡ sn(u, k), . . .
and
E(u) ≡ E (am(u, k), k) .
192
Theta functions: deﬁnitions and relations
Derivatives of the Jacobian elliptic functions with respect to the argument: ∂ am u = dn u, ∂u ∂ sn u = cn u dn u, ∂u ∂ cn u = −sn u dn u, ∂u ∂ dn u = −k 2 sn u cn u. ∂u
(A2.27)
Derivatives of the Jacobian elliptic functions with respect to the modulus: ∂ am u = ∂k ∂ sn u = ∂k ∂ cn u = ∂k ∂ dn u = ∂k
dn u cn u 2 2 , u + k sn u −E(u) + k dn u k k 2 dn u cn u cn u 2 2 , u + k sn u −E(u) + k dn u k k 2 cn u sn u dn u 2 2 , u − k sn u E(u) − k dn u k k 2 ksn u cn u sn u 2 E(u) − k . u − dn u cn u k 2
Derivative of ln ϑ2 with respect to the argument:
∂ K(k ) πF(β, k ) , −π = 0 (β, k) ln ϑ2 ∂w 2K(k) K(k) with w=
πF(β, k ) πu = , 2K(k) 2K(k)
β = am(u, k ).
(A2.28)
(A2.29)
Appendix 3 Multipole moments of the rotating disc of dust
Here we provide all the quantities that are necessary for calculating the ﬁrst eleven normalized multipole moments P˜ 0 , P˜ 1 , P˜ 2 , . . . , P˜ 10 of the rigidly rotating disc of dust. Their exact deﬁnition and the method of how they can be obtained from the axis potential (2.256) were discussed in detail in Subsection 2.3.4. The moments P˜ n are given as functions of the parameter µ and the more precise structure is P˜ n (µ) = P˜ n b0 (µ), 0 (µ), cj (µ) : j < jmax (n) , (A3.1) where jmax (n) refers to the ﬁnite maximal value of j for a given n. The reader could calculate P˜ 0 , . . . , P˜ 10 using only the deﬁnitions to follow. For clarity, we shall make some remarks concerning their derivation and, for the ﬁrst seven multipole moments, we list the explicit formulae. A3.1 Deﬁnitions and auxiliary coefﬁcients We rewrite the equations for the parameter functions b0 (2.202) and 0 (2.204): 1 ˆ ˆ b0 (µ) = − sn I (µ), h (µ) dn I (µ), h (µ) , h(µ) (A3.2) 1 h2 (µ) 0 ≡ 0 (µ) = cn Iˆ (µ), h (µ) . 1− 2 2 h (µ) The moduli h and h of the Jacobi elliptic functions sn, cn and dn are given by 1 1 µ µ h(µ) = , h (µ) = , (A3.3) 1+
1−
2 2 1 + µ2 1 + µ2 and the main argument of these functions is deﬁned as 4 Iˆ (µ) := 1 + µ2 I0 (µ) , 193
(A3.4)
194
Multipole moments of the rotating disc of dust
where I0 is deﬁned as the ﬁrst of the In (n = 0, 1, 2, . . .) √
1 µ ln( 1 + x2 + x) xn In (µ) := dx . √ √ π 0 µ−x 1 + x2 Furthermore the abbreviations
(A3.5)
τ :=
4
1+
am := am Iˆ (µ), h (µ) , cn := cn Iˆ (µ), h (µ) , scd := sn cn dn ,
1 µ2
and
sn := sn Iˆ (µ), h (µ) , dn := dn Iˆ (µ), h (µ) , E := E(am, h )
(A3.6)
(A3.7)
(A3.8) (A3.9)
are used in order to render the equations more concise. We now provide expressions for the functions cj up to j = 11. These quantities are the coefﬁcients of the function N (µ, y) (2.228) at inﬁnity, which are needed for calculating the ﬁrst eleven multipole moments. The ﬁrst coefﬁcient c1 is evaluated in Subsection 2.3.4. This is the only one which is necessary for calculating the mass and angular momentum (the ﬁrst two moments). All this is explained in detail in the aforementioned subsection. The coefﬁcients cn of the function N (see (2.228) and (2.257)) : √ c1 = 2 E τ + (−(µ (1 + τ 2 ) I0 ) + I1 )/ µ c3 = (−2 E µ3/2 τ + µ3/2 scd (τ − τ 3 ) + µ2 τ 2 (1 + τ 2 ) I0 − 3 µ I1 + 3 I2 )/(3 µ3/2 ) c5 = (−2 E µ5/2 τ (−3 + τ 4 ) + µ5/2 scd τ (−1 + τ 2 ) (4 + cn2 (−1 + τ 2 )) + µ3 (5 − 3 τ 2 − 7 τ 4 + τ 6 ) I0 + 5 µ2 (1 + τ 4 ) I1 − 20 µ I2 + 10 I3 )/(10 µ5/2 ) c7 = (4 E µ7/2 τ (−5 + 3 τ 4 ) − µ7/2 scd τ (−1 + τ 2 ) (6 cn2 (−1 + τ 2 ) + cn4 (−1 + τ 2 )2 − 3 (−5 + τ 4 )) − 2 µ4 (14 − 5 τ 2 − 17 τ 4 + 3 τ 6 + τ 8 )I0 + 14 µ3 (1 − 3 τ 4 ) I1 + 14 µ2 (5 + τ 4 ) I2 − 84 µ I3 + 28 I4 )/(28 µ7/2 )
A3.1 Deﬁnitions and auxiliary coefﬁcients
195
c9 = (2 E µ9/2 τ (35 − 30 τ 4 + 3 τ 8 ) + µ9/2 scd τ (−1 + τ 2 ) (56 − 24 τ 4 + 8 cn4 (−1 + τ 2 )2 + cn6 (−1 + τ 2 )3 − 4 cn2 (7 − 7 τ 2 − τ 4 + τ 6 )) + µ5 (117 − 35 τ 2 − 146 τ 4 + 30 τ 6 + 21 τ 8 − 3 τ 10 ) I0 − 9 µ4 (17 − 26 τ 4 + τ 8 ) I1 − 144 µ3 (1 + τ 4 ) I2 + 36 µ2 (11 + τ 4 ) I3 − 288 µ I4 + 72 I5 )/(72 µ9/2 ) c11 = (µ11/2 τ (−4 E (63 − 70 τ 4 + 15 τ 8 ) − scd (−1 + τ 2 ) (10 cn6 (−1 + τ 2 )3 + cn8 (−1 + τ 2 )4 − 5 cn4 (−1 + τ 2 )2 (−9 + τ 4 ) − 40 cn2 (3 − 3 τ 2 − τ 4 + τ 6 ) + 10 (21 − 14 τ 4 + τ 8 ))) + 2 µ6 (−220 + 63 τ 2 + 299 τ 4 − 70 τ 6 − 74 τ 8 + 15 τ 10 + 3 τ 12 ) I0 + 22 µ5 (37 − 50 τ 4 + 5 τ 8 ) I1 − 22 µ4 (1 − 42 τ 4 + τ 8 ) I2 − 440 µ3 (3 + τ 4 )I3 + 88 µ2 (19 + τ 4 ) I4 − 880 µ I5 + 176 I6 )/(176 µ11/2 ). The coefﬁcients m ˜ n of the function g (see (2.254) and (2.255)) : m ˜ 0 = −b0 − 0 c1 m ˜ 1 = −i (b0 + 2 0 c1 ) m ˜ 2 = b0 + (2 + b20 ) 0 c1 + b0 20 c12 + (30 (c13 − 12 c3 ))/3 m ˜ 3 = (i/3) (3 b30 + 12 b20 0 c1 + 3 b0 20 (4 + 3 c12 ) + 2 30 (12 c1 + c13 − 12 c3 )) m ˜ 4 = b0 − 2 b30 + (2 − 7 b20 ) 0 c1 − b0 20 (8 + (5 + b20 ) c12 ) − (2 30 (24 c1 + (1 + 2 b20 ) c13 − 6 (2 + b20 ) c3 ))/3 + b0 40 ((−2 c14 )/3 + 8 c1 c3 ) + 50 ((−2 c15 )/15 + 4 c12 c3 − 16 c5 ) m ˜ 5 = (i/15) (15 b0 − 30 b30 − 30 (−1 + 3 b20 + b40 ) 0 c1 − 30 b0 20 (2 (1 + c12 ) + b20 (2 + 3 c12 )) − 40 30 ((3 + 6 b20 ) c1 + 2 b20 c13 − 6 b20 c3 ) + 30 b0 40 (−10 + 2 τ 4 − 6 c12 − c14 + 12 c1 c3 ) − 4 50 (−30 (−5 + τ 4 ) c1 + 10 c13 + c15 − 30 c12 c3 − 120 (c3 − c5 )))
196
Multipole moments of the rotating disc of dust
m ˜ 6 = b50 + b20 (−1 + 8 b20 ) 0 c1 + b0 20 (−8 − c12 + 16 b20 (1 + c12 )) + (30 (24 (−2 + 7 b20 ) c1 + (−2 + 34 b20 + 3 b40 ) c13 + 12 (2 − 7 b20 ) c3 ))/3 + (b0 40 (168 − 24 τ 4 + 120 c12 + 5 (2 + b20 ) c14 − 24 (5 + b20 ) c1 c3 ))/3 + (50 (−240 (−7 + τ 4 ) c1 + 80 c13 + (4 + 17 b20 ) c15 − 120 (1 + 2 b20 ) c12 c3 + 240 (−4 c3 + (2 + b20 ) c5 )))/15 + b0 60 ((17 c16 )/45 − (32 c13 c3 )/3 + 16 c32 + 32 c1 c5 ) + 70 ((17 c17 )/315 − (8 c14 c3 )/3 + 16 c1 c32 + 16 c12 c5 − 64 c7 ) m ˜ 7 = (i/315) (315 b0 − 945 b30 + 945 b50 + 630 (1 − 5 b20 + 8 b40 ) 0 c1 + 315 b0 20 (−16 − 7 c12 + 3 b40 c12 + b20 (28 + 25 c12 )) + 210 30 (12 (−4 + 11 b20 + b40 ) c1 + (−1 + 21 b20 + 13 b40 ) c13 − 12 (−1 + 3 b20 + b40 ) c3 ) − 105 b0 40 (4 (−39 + 3 τ 4 − 45 c12 − 2 c14 + 24 c1 c3 ) + 3 b20 (−20 + 4 τ 4 − 24 c12 − 9 c14 + 48 c1 c3 )) + 84 50 (−30 (−13 + τ 4 + 2 b20 (−5 + τ 4 ))c1 + 10(1 + 8 b20 ) c13 + 17 b20 c15 − 240 b20 c12 c3 − 120 (c3 + 2 b20 c3 − 2 b20 c5 )) + 21 b0 60 (900 c12 + 120 c14 + 17 c16 − 180 τ 4 (4 + c12 ) − 480 c13 c3 + 240 (7 + 3 c32 ) − 1440 c1 (c3 − c5 )) + 2 70 (−420 (−5 + τ 4 ) c13 + 168 c15 + 17 c17 − 840 c14 c3 − 5040 c1 (−7 + 3 τ 4 − c32 ) − 5040 c12 (c3 − c5 ) + 5040 ((−5 + τ 4 ) c3 + 4 (c5 − c7 )))) m ˜ 8 = −b0 + 4 b30 − 4 b50 − (2 − 12 b20 + 16 b40 + 3 b60 ) 0 c1 − b0 20 (−8 (1 + c12 ) + b40 (8 + 17 c12 ) + b20 (16 + 19 c12 )) − (4 30 (6 (−2 + 5 b20 + 8 b40 ) c1 + b20 (5 + 23 b20 )c13 + 3 b20 (1 − 8 b20 )c3 ))/3 − (b0 40 (−120 + 24 τ 4 + 72 c12 − 2 c14 + 3 b40 c14 + 24 c1 c3 + b20 (336 − 48 τ 4 + 384 c12 + 74 c14 − 384 c1 c3 )))/3 − (50 (−120 (7 b20 (−7 + τ 4 ) − 2 (−5 + τ 4 )) c1 + 80 (−1 + 17 b20 ) c13 + (−4 + 149 b20 + 30 b40 ) c15 − 60 (−2 + 34 b20 + 3 b40 ) c12 c3
A3.1 Deﬁnitions and auxiliary coefﬁcients
197
− 240 (−2 + 7 b20 ) (2 c3 − c5 )))/15 − (b0 60 (12600 c12 + 1200 c14 + (85 + 77 b20 ) c16 − 360 τ 4 (16 + 5 c12 ) − 1200 (2 + b20 ) c13 c3 + 720 (24 + (5 + b20 ) c32 ) − 1440 c1 (10 c3 − (5 + b20 ) c5 )))/45 − (2 70 (−840 (−7 + τ 4 ) c13 + 336 c15 + (17 + 124 b20 ) c17 − 210 (4 + 17 b20 ) c14 c3 − 5040 c1 (−24 + 8 τ 4 − (1 + 2 b20 ) c32 ) − 5040 c12 (2 c3 − (1 + 2 b20 ) c5 ) − 10080 (−((−7 + τ 4 ) c3 ) − 4 c5 + (2 + b20 ) c7 )))/315 + b0 80 ((−62 c18 )/315 + (136 c15 c3 )/15 − 64 c12 c32 − (128 c13 c5 )/3 + 128 c3 c5 + 128 c1 c7 ) + 90 ((−62 c19 )/2835 + (68 c16 c3 )/45 − (64 c13 c32 )/3 − (32 c14 c5 )/3 + 128 c1 c3 c5 + 64 c12 c7 + (64 (c33 − 12 c9 ))/3) m ˜ 9 = (i/2835) (2835 b30 − 2835 b50 − 2835 b70 − 5670 b20 (−1 − b20 + 7 b40 ) 0 c1 − 2835 b0 20 (12 − c12 + 12 b40 (3 + 4 c12 ) − b20 (32 + 11 c12 )) − 1890 30 (12 (3 − 14 b20 + 25 b40 ) c1 + (1 − 17 b20 + 98 b40 + 6 b60 ) c13 − 12 (1 − 5 b20 + 8 b40 ) c3 ) − 945 b0 40 (36 b40 c1 (c1 + c13 − 2 c3 ) + b20 (804 − 84 τ 4 + 972 c12 + 125 c14 − 600 c1 c3 ) + 2 (−240 + 24 τ 4 − 120 c12 − 7 c14 + 84 c1 c3 )) − 378 50 (−60 (40 − 4 τ 4 + b40 (−5 + τ 4 ) + b20 (−107 + 11 τ 4 )) c1 + 20 (−4 + 71 b20 + 13 b40 ) c13 + (−2 + 96 b20 + 107 b40 ) c15 − 60 (−1 + 21 b20 + 13 b40 ) c12 c3 − 240 ((−4 + 11 b20 + b40 ) c3 − (−1 + 3 b20 + b40 ) c5 )) + 126 b0 60 (2 (−6615 c12 − 450 c14 − 17 c16 + 45 τ 4 (28 + 15 c12 ) + 480 c13 c3 − 180 (31 + 4 c32 ) + 360 c1 (15 c3 − 4 c5 )) + 3 b20 (−900 c12 − 270 c14 − 67 c16 + 180 τ 4 (2 + c12 ) + 1080 c13 c3 − 120 (7 + 6 c32 ) + 1440 c1 (c3 − c5 )))
198
Multipole moments of the rotating disc of dust
− 72 70 (−105 (−13 + τ 4 + 8 b20 (−5 + τ 4 )) c13 + 42 (1 + 17 b20 ) c15 + 124 b20 c17 − 3570 b20 c14 c3 − 1260 c1 (−31 + 7 τ 4 + 2 b20 (−7 + 3 τ 4 − 4 c32 )) − 1260 c12 (c3 + 8 b20 c3 − 8 b20 c5 ) + 1260 ((−13 + τ 4 + 2 b20 (−5 + τ 4 )) c3 + 4 (c5 + 2 b20 c5 − 2 b20 c7 ))) − 54 b0 80 (1680 τ 8 + 2100 c14 + 238 c16 + 31 c18 − 1428 c15 c3 − 420 τ 4 (56 + 18 c12 + c14 − 12 c1 c3 ) + 2520 c12 (7 + 4 c32 ) − 6720 c13 (c3 − c5 ) + 5040 (7 + 2 c32 − 4 c3 c5 ) − 5040 c1 (5 c3 − 4 c5 + 4 c7 )) − 4 90 (−756 (−5 + τ 4 ) c15 + 306 c17 + 31 c19 − 2142 c16 c3 − 7560 c13 (−7 + 3 τ 4 − 4 c32 ) − 15120 c14 (c3 − c5 ) + 45360 c1 (21 − 14 τ 4 + τ 8 + 2 c32 − 4 c3 c5 ) + 22680 c12 ((−5 + τ 4 ) c3 + 4 (c5 − c7 )) + 30240 (3 (−7 + 3 τ 4 ) c3 − c33 − 3 ((−5 + τ 4 ) c5 + 4 (c7 − c9 ))))) m ˜ 10 = −b0 + 4 b30 − 6 b50 + 4 b70 + (−2 + 13 b20 − 32 b40 + 32 b60 ) 0 c1 + b0 20 (24 + 9 c12 + 6 b60 c12 − 10 b20 (8 + 5 c12 ) + 8 b40 (9 + 10 c12 )) + (30 (24 (6 − 31 b20 + 40 b40 + 3 b60 ) c1 + (2 − 84 b20 + 253 b40 + 90 b60 ) c13 − 12 (2 − 12 b20 + 16 b40 + 3 b60 ) c3 ))/3 − (b0 40 (−8 (−57 + 3 τ 4 − 63 c12 − 2 c14 + 24 c1 c3 ) + b40 (−168 + 24 τ 4 − 408 c12 − 163 c14 + 408 c1 c3 ) + b20 (−816 + 48 τ 4 − 1296 c12 − 119 c14 + 456 c1 c3 )))/3 + (50 (−120 (38 − 2 τ 4 + 8 b40 (−7 + τ 4 ) + b20 (−97 + 5 τ 4 )) c1 + 80 (−1 + 37 b20 + 46 b40 ) c13 + b20 (103 + 736 b20 + 15 b40 ) c15 − 240 b20 (5 + 23 b20 ) c12 c3 − 240 (2 (−2 + 5 b20 + 8 b40 ) c3 + b20 (1 − 8 b20 ) c5 )))/15 + (b0 60 (−4320 + 22680 c12 + 720 c14 − 17 c16 − 1080 τ 4 (−4 + c12 ) − 8640 c1 c3 + 480 c13 c3 − 720 c32 + 15 b40 (7 c16 − 48 c13 c3 )
A3.1 Deﬁnitions and auxiliary coefﬁcients
− 8 b20 (−5040 c12 − 1110 c14 − 139 c16 + 720 τ 4 (2 + c12 ) + 2220 c13 c3 − 1440 (3 + c32 ) + 2880 c1 (2 c3 − c5 )) − 1440 c1 c5 ))/45 + (2 70 (−840 (5 − τ 4 + 17 b20 (−7 + τ 4 )) c13 + 84 (−4 + 149 b20 ) c15 + (−17 + 1099 b20 + 378 b40 ) c17 − 210 (−4 + 149 b20 + 30 b40 ) c14 c3 + 2520 c1 (3 b40 c32 + 2 (−6 + 6 τ 4 − c32 ) + b20 (168 − 56 τ 4 + 34 c32 )) − 2520 c12 ((−4 + 68 b20 ) c3 + (2 − 34 b20 − 3 b40 ) c5 ) + 5040 ((7 b20 (−7 + τ 4 ) − 2 (−5 + τ 4 )) c3 + 2 (−2 + 7 b20 ) (2 c5 − c7 ))))/315 + (2 b0 80 (12600 τ 8 + 29400 c14 + 2380 c16 + 5 (31 + 44 b20 ) c18 − 84 (85 + 77 b20 ) c15 c3 − 4200 τ 4 (54 + 24 c12 + c14 − 12 c1 c3 ) + 25200 c12 (12 + (2 + b20 ) c32 ) − 16800 c13 (4 c3 − (2 + b20 ) c5 ) + 2520 (165 + 40 c32 − 8 (5 + b20 ) c3 c5 ) − 10080 c1 (35 c3 + 2 (−10 c5 + (5 + b20 ) c7 ))))/315 + (2 90 (−3024 (−7 + τ 4 ) c15 + 1224 c17 + (62 + 691 b20 ) c19 − 252 (17 + 124 b20 ) c16 c3 − 15120 c13 (−24 + 8 τ 4 − (4 + 17 b20 ) c32 ) − 7560 c14 (8 c3 − (4 + 17 b20 ) c5 ) + 45360 c1 (165 − 90 τ 4 + 5 τ 8 + 8 c32 − 8 (1 + 2 b20 ) c3 c5 ) + 90720 c12 ((−7 + τ 4 ) c3 − 2 (−2 c5 + c7 + 2 b20 c7 )) − 60480 (−24 (−3 + τ 4 ) c3 + (1 + 2 b20 ) c33 − 6 (−((−7 + τ 4 ) c5 ) − 4 c7 + (2 + b20 ) c9 ))))/2835 10 7 4 2 5 + (2 b0 10 0 (691 c1 − 44640 c1 c3 + 642600 c1 c3 + 257040 c1 c5
− 3628800 c12 c3 c5 − 1209600 c13 c7 + 1814400 (c52 + 2 c3 c7 ) − 1209600 c1 (c33 − 3 c9 )))/14175 11 8 5 2 + 11 0 ((1382 c1 )/155925 − (248 c1 c3 )/315 + (272 c1 c3 )/15
+ (272 c16 c5 )/45 − (512 c13 c3 c5 )/3 − (128 c14 c7 )/3 + 256 c1 (c52 + 2 c3 c7 ) − (256 c12 (c33 − 3 c9 ))/3 + 256 (c32 c5 − 4 c11 )).
199
200
Multipole moments of the rotating disc of dust
A3.2 Normalized multipole moments With the normalized coefﬁcients m ˜ n , we are now able to calculate the ﬁrst six ˜ ˜ ˜ 4, M ˜ 6, M ˜ 8 and M ˜ 10 and the ﬁrst ﬁve normalized normalized mass moments M0 , M2 , M rotational moments J˜1 , J˜3 , J˜5 , J˜7 and J˜9 . The scheme derived in Fodor et al. (1989) and adapted to our notation and normalization reads as follows. ˜ n (m Scheme for M ˜ i ) and J˜n (m ˜ i ) up to n = 10 : ˜0 = m ˜0, M ˜1, J˜1 = m ˜2 = −m ˜2, M ˜3, J˜3 = − m 1 ˜4 = m ˜ 4 − (m ˜0 −m ˜ 21 ) m ˜0, ˜ 2m M 7 1 1 2 ˜J5 = m ˜ 2m ˜ 3m ˜ 5 − (m ˜0 −m ˜ 1) m ˜ 1 − (m ˜0 −m ˜ 2m ˜ 1) m ˜0 , 21 3 1 5 2 ˜6 = − m ˜ 2m (m ˜ 2m ˜ 6 + (m ˜0 −m ˜ 21 ) m ˜0m ˜0 − ˜0 −m ˜ 21 ) m ˜2 M 33 231 8 4 ˜ 2m ˜ 1) m ˜ 1 − (m ˜1 −m ˜ 22 ) m ˜0 ˜ 3m ˜ 3m − (m ˜0 −m 33 33 6 ˜ 4m − (m ˜0 −m ˜ 3m ˜ 1) m ˜0 , 11 ˜7 − ... , J˜7 = − m ˜8 = m ˜8 − ... , M ˜9 − ... , J˜9 = m ˜ 10 = − m ˜ 10 − . . . . M ˜ 8, M ˜ 10 can be found in Fodor et al. (1989). The complete formulae for J˜7 , J˜9 and M ˜ 0, M ˜ 2, M ˜ 4, M ˜ 6 and J˜1 , J˜3 , J˜5 . Next we list the resulting expressions for M The ﬁrst seven multipole moments: ˜ 0 = −b0 − 0 c1 M J˜1 = −b0 − 2 0 c1 ˜ 2 = (−3 b0 − 6 0 c1 − 3 b20 0 c1 − 3 b0 20 c12 − 30 c13 + 12 30 c3 )/3 M
A3.2 Normalized multipole moments
201
J˜3 = −b30 − 4 b20 0 c1 − b0 20 (4 + 3 c12 ) − (2 30 (12 c1 + c13 − 12 c3 ))/3 ˜ 4 = b0 − 2 b30 − ((−14 + 48 b20 + b40 ) 0 c1 )/7 M − (2 b0 20 (28 + (16 + 5 b20 ) c12 ))/7 − (2 30 (168 c1 + (4 + 19 b20 ) c13 − 12 (7 + 4 b20 ) c3 ))/21 − (b0 40 (19 c14 − 192 c1 c3 ))/21 − (50 (19 c15 − 480 c12 c3 + 1680 c5 ))/105 J˜5 = (63 b0 − 105 b30 − 21 b50 − 6 (−21 + 46 b20 + 38 b40 ) 0 c1 − 6 b0 20 (42 + 16 c12 + b20 (56 + 89 c12 )) − 24 30 (7 (3 + 8 b20 ) c1 + 3 (−1 + 6 b20 ) c13 − 45 b20 c3 ) + b0 40 (−1260 + 252 τ 4 − 1176 c12 − 145 c14 + 1560 c1 c3 ) + 63 50 (8 (−5 + τ 4 ) c1 − (16 c13 )/3 − (74 c15 )/315 + (160 c12 c3 )/21 + 32 (c3 − c5 )))/63 ˜ 6 = (2310 b30 − 5775 b50 − 15 b20 (−852 + 2462 b20 + 7 b40 ) 0 c1 M − 15 b0 20 (110 b40 c12 − 3 (616 + 331 c12 ) + b20 (4312 + 4348 c12 )) − 5 30 (336 (−33 + 134 b20 ) c1 + (−1278 + 8092 b20 + 1711 b40 ) c13 − 12 (−462 + 1594 b20 + 133 b40 ) c3 ) − 5 b0 40 (38808 − 5544 τ 4 + 33432 c12 + 5 (331 + 494 b20 ) c14 − 48 (503 + 223 b20 ) c1 c3 ) + 3 50 (18480 (−7 + τ 4 ) c1 − 8400 c13 − 7 (−26 + 389 b20 ) c15 + 120 (35 + 253 b20 ) c12 c3 − 1680 (−44 c3 + (22 + 17 b20 ) c5 )) − b0 60 (2723 c16 − 60720 c13 c3 + 43200 c32 + 171360 c1 c5 ) − 70 (389 c17 − 15180 c14 c3 + 43200 c1 c32 + 85680 c12 c5 − 221760 c7 ))/3465. ˜ 8, M ˜ 10 and J˜7 , ˜ 1, . . . , m ˜ 10 are sufﬁcient for calculating M The coefﬁcients m ˜ 0, m ˜J9 . Because of the length of the corresponding expressions, we shall not list them here. They were used, of course, for producing the plots in Fig. 2.13. To derive the moments, we made use of a Mathematica program.
202
Multipole moments of the rotating disc of dust
A3.3 Multipole moments in the extreme relativistic limit Let us conclude with a remark regarding the extreme relativistic limit µ → µ0 . We know from Subsection 2.3.5 that the ‘exterior’ metric becomes that of the extreme Kerr solution, and therefore the multipole moments tend to ˜ 2l (µ0 ) = (−1)l m M ˜ 2l (µ0 ) = 1 , ˜ 2l+1 (µ0 ) = 1 . J˜2l+1 (µ0 ) = (−1)l m
(A3.10)
Based on the expressions given in this appendix, we can see this explicitly for the ﬁrst eleven moments. In the limit, we have 0 → 0 and b0 → −1. For the coefﬁcients m ˜ 0, m ˜ 10 , we immediately see that ˜ 1, . . . , m m ˜ 2l (µ0 ) = (−1)l , m ˜ 2l+1 (µ0 ) = (−1)l i ,
(A3.11)
which can in fact be shown for all l. The structure for the ﬁrst eleven multipole moments is ˜ 2l = (−1)l m M ˜ 2l + As(l) m ˜ js(l) − m ˜ is(l) −1 m ˜ js(l) +1 , ˜ is(l) m s(l)
˜ 2l+1 + J˜2l+1 = (−1)l m
˜ it(l) m Bs(l) m ˜ jt(l) − m ˜ it(l) −1 m ˜ jt(l) +1 .
(A3.12)
t(l)
Using (A3.11), one can verify m ˜ im ˜j − m ˜ i−1 m ˜ j+1 (µ0 ) = 0 , from which it follows that (A3.10) holds.
(A3.13)
Appendix 4 The disc solution as a Bäcklund limit
In this appendix, we discuss an alternative representation of the solution for the rigidly rotating disc of dust. The underlying mathematical structure of this formulation is given through the socalled Bäcklund transformation, which is a technique that enables one to construct explicit solutions to the linear matrix problem (2.41) and the corresponding Ernst potentials f . These solutions take a particularly simple form, since they can be written as quotients of determinants in which only elementary functions and functions that can be calculated from a ‘seed solution’ f0 appear (see below for examples). The Kerr solution for a rotating black hole in vacuum, Equation (2.358), can be considered as a particular example of a Bäcklund transform, see e.g. Neugebauer (1980a). Moreover, the method allows for the construction of regular Ernst potentials, which correspond to disclike sources of the gravitational ﬁeld. In particular, it is possible to identify the rigidly rotating disc of dust as a welldeﬁned limit of these solutions. After the introduction of disclike solutions, generated by Bäcklund transformations, depending on a set of parameters as well as a real analytic function, an appropriate generalization is given which allows the Ernst potentials to be written in terms of two free functions. For the rigidly rotating disc of dust, these functions take on a simple explicit form.
A4.1 Disclike solutions of the Bäcklund type The expression f = f (/0 , ζ /0 ; {Xν }q ; g) = f0
D− D+
(A4.1)
satisﬁes the Ernst equation, where {X1 , . . . , Xq } =: {Xν }q is a set of complex parameters, g a real analytic function deﬁned on the interval [0, 1], and D± and f0
203
204
The disc solution as a Bäcklund limit
are given by 1 1 ±1 α1 λ1 1 λ21 D± = ±1 α1 λ3 1 .. .. . . 2q 1 λ1 and
1
1
α1∗ λ∗1
α2 λ2
α2∗ λ∗2
· · · αq λq
(λ∗1 )2
λ22
(λ∗2 )2
···
α1∗ (λ∗1 )3
α2 λ32
α2∗ (λ∗2 )3
.. .
.. .
.. .
..
.
.. .
(λ∗1 )2q
λ2
(λ∗2 )2q
···
λq
2q
f0 = exp −
1
1 λ2q
· · · αq λ3q
2q
∗ ∗ αq λq (λ∗q )2 αq∗ (λ∗q )3 .. . ∗ 2q (λq ) (A4.2) 1
(−1)q g(x2 )dx W1 (ix)
−1
with
···
1
(X − ζ /0 )2 + (/0 )2 ((W1 ) < 0) , 0 Xν − i¯z λν = , λ∗ν λν = 1 (z = + iζ ), 0 Xν + iz 1 q 2 (−1) g(x )dx λν , (0 Xν + iz) αν = − tanh 20 (ix − Xν )W1 (ix)
(A4.3)
W1 (X ) =
αν∗ α ν = 1.
−1
Through the additional requirement that for each parameter Xν there must also be a parameter Xµ with Xν = −X µ , reﬂectional symmetry, f (, −ζ ) = f (, ζ ), is ensured.1 Moreover, the parameters Xν are assumed to lie outside the imaginary interval [−i, i]. The above Ernst potential f is obtained by a multiple Bäcklund transformation applied to the real seed solution f0 , see Neugebauer (1980a). The particular ansatz chosen for the seed solution f0 guarantees a resulting Ernst potential that corresponds to a disclike source of the gravitational ﬁeld. Furthermore, f does not possess singularities at (, ζ ) = 0 ([Xν ], −[Xν ]). This is due to the fact that αν λν is a function of λ2ν , and this means that f does not behave like a square root function near the critical points (, ζ ) = 0 ([Xν ], −[Xν ]), but rather like a rational function. In addition, one has to make sure that no zeros in the denominator of (A4.1) occur. The real function g that enters the Ernst potential 1 Hence, the set {iX } consists of real parameters and/or pairs of complex conjugate parameters. ν q
A4.2 Generalization of the Bäcklund type solutions by a limiting process
205
is assumed to be analytic on [0, 1] in order to guarantee analytic behaviour of the surface energymomentum distribution. The additional requirement g(1) = 0
(A4.4)
ensures regularity at the rim of the disc. A4.2 Generalization of the Bäcklund type solutions by a limiting process The set {Xν }q of complex parameters can be translated into an analytic function ξ : [0, 1] → R such that the corresponding Ernst potential depends on two real analytic functions deﬁned on [0, 1]: f = f (/0 , ζ /0 ; ξ ; g). This concept proves to be sufﬁciently general to describe arbitrarily rotating discs. In this manner it becomes possible to describe the solution of the rigidly rotating disc of dust as a welldeﬁned limit of the Bäcklund type solutions. The following equalities for the above solutions f = f ({Xν }q ; g) will help in introducing the aforementioned, analytic function ξ , see Ansorg (2001): f [{X1 , . . . , Xq−2 , Xq−1 , Xq }; g] = f [{X1 , . . . , Xq−2 }; g] if Xq−1 = −Xq ∈ R f [{X1 , . . . , Xq−2 , Xq−1 , Xq }; g] = f [{X1 , . . . , Xq−2 }; g] if Xq−1 = X q lim f [{X1 , . . . , Xq−1 , it}; g] = f [{X1 , . . . , Xq−1 }; g]
t→∞
if t ∈ R lim f [{X1 , . . . , Xq−2 , Xq−1 , Xq }; g] = f [{X1 , . . . , Xq−2 }; g]
Xq →∞
if Xq−1 = −X q . The desired function ξ = ξ({Xν }q ) is supposed to be invariant under the above modiﬁcations of the set {Xν }q that do not affect the Ernst potential. This requirement is met by the real analytic function q
i Xν − x 1 ξ(x2 ; {Xν }q ) = ln , x ∈ [−1, 1], (A4.5) x i Xν + x ν=1
206
The disc solution as a Bäcklund limit
which can be proved by considering that for each parameter Xν there is also a parameter Xµ with Xν = −X µ , and that, moreover, the parameters Xν do not lie on the imaginary interval [−i, i]. The set X of all functions ξ = ξ(x2 ; {Xν }q ), q ∈ N, which are deﬁned by (A4.5) forms a dense subset of the set A of all real analytic functions on [0, 1]. Now, for a given function g, each ξ ∈ X is mapped by (A4.1) onto a uniquely deﬁned Ernst potential f ∈ E : g : X −→ E , g (ξ ) = f ({Xν }q ; g), (A4.6) where the set {Xν }q results from ξ by (A4.5).2 The mapping g can be extended to form a continuous function deﬁned on A.3 It then follows that, given the two real functions g and ξ , deﬁned and analytic on the interval [0, 1], the Ernst potential f (ξ ; g) = lim f ({Xν(q) }q ; g) q→∞
exists and is independent of the particular choice of the sequence (q) {{Xν }q }∞ q=q0 which serves to represent ξ by q
i Xν(q) − x 1 ξ(x ) = lim ln (q) x q→∞ ν=1 i Xν + x 2
for
x ∈ [−1, 1].
It can be shown that, in this formulation, the solution for the rigidly rotating disc of dust assumes the form f = f (ξ ; g) with the functions ξ and g given by ξ(x2 ) =
x2 − C1 (µ)x + C2 (µ) 1 , ln 2 2x x + C1 (µ)x + C2 (µ) C1 (µ) = 2[1 + C2 (µ)],
C2 (µ) =
1 1 + µ2 , µ
1 g(x2 ) = − arcsinh[µ(1 − x2 )]. π They depend parametrically on µ, 0 < µ < µ0 = 4.62966184 . . . , which was introduced in (2.79). Note that a rather technical detail is the determination of an appropriate set {Xν }q to give a satisfactory approximation of ξ in terms of (A4.5). There are many ways to do this and we here provide a single, concrete example. 2 Here, E denotes the set of all Ernst potentials corresponding to disclike sources. 3 Detailed mathematical aspects are discussed in Ansorg (2001) and Ansorg et al. (2002b).
A4.2 Generalization of the Bäcklund type solutions by a limiting process
207
For a given function ξ , one can use Equation (A4.5) to write q
Pq (−x) i Xν − x 2 = exp x ξ(x ) ≈ i Xν + x Pq (x)
(A4.7)
ν=1
with Pq (x) =
q−1
bν xν + xq .
ν=0
The coefﬁcients bν can be determined by equating left and right hand sides of (A4.7) at the q zeros xµ2 ∈ [0, 1] of the Chebyshev polynomial Tq (2x2 − 1) and solving the corresponding linear system: q q
exp xµ ξ(xµ2 ) bν xµν = bν (−xµ )ν . ν=0
ν=0
The zeros of Pq determine the Xν . In the limit q → ∞, we thus obtain an exact representation of ξ in [0, 1].
References
Abramowicz, M. A., Calvani, M. and Nobili, L. (1983). Runaway instability in accretion disks orbiting black holes. Nature, 302, 597. Abramowicz, M. A., Curir, A., SchwarzenbergCzerny, A. and Wilson, R. E. (1984). Selfgravity and the global structure of accretion discs. Mon. Not. R. Astron. Soc., 208, 279. Ansorg, M. (1998). Timelike geodesic motions within the general relativistic gravitational ﬁeld of the rigidly rotating disk of dust. J. Math. Phys., 39, 5984. Ansorg, M. (2001). Differentially rotating disks of dust: Arbitrary rotation law. Gen. Rel. Grav., 33, 309. Ansorg, M., Fischer, T., Kleinwächter, A., Meinel, R., Petroff, D. and Schöbel, K. (2004). Equilibrium conﬁgurations of homogeneous ﬂuids in general relativity. Mon. Not. R. Astron. Soc., 355, 682. Ansorg, M., Kleinwächter, A. and Meinel, R. (2002a). Highly accurate calculation of rotating neutron stars. Astron. Astrophys., 381, L49. Ansorg, M., Kleinwächter, A. and Meinel, R. (2003a). Highly accurate calculation of rotating neutron stars: Detailed description of the numerical methods. Astron. Astrophys., 405, 711. Ansorg, M., Kleinwächter, A. and Meinel, R. (2003b). Relativistic Dyson rings and their black hole limit. Astrophys. J. Lett., 582, L87. Ansorg, M., Kleinwächter, A. and Meinel, R. (2003c). Uniformly rotating axisymmetric ﬂuid conﬁgurations bifurcating from highly ﬂattened Maclaurin spheroids. Mon. Not. R. Astron. Soc., 339, 515. Ansorg, M., Kleinwächter, A., Meinel, R. and Neugebauer, G. (2002b). Dirichlet boundary value problems of the Ernst equation. Phys. Rev. D, 65, 044006. Ansorg, M. and Petroff, D. (2005). Black holes surrounded by uniformly rotating rings. Phys. Rev. D, 72, 024019. Ansorg, M. and Petroff, D. (2006). Negative Komar mass of single objects in regular, asymptotically ﬂat spacetimes. Class. Quantum Grav., 23, L81. Ansorg, M. and Pﬁster, H. (2008). A universal constraint between charge and rotation rate for degenerate black holes surrounded by matter. Class. Quantum Grav., 25, 035009. Arnold, D. N. (2001). A Concise Introduction to Numerical Analysis (Minneapolis, Institute for Mathematics and its Applications, University of Minnesota). URL www.ima.umn.edu/ arnold/597.0001/nabook.pdf.
208
References
209
Ashtekar, A. and Krishnan, B. (2004). Isolated and dynamical horizons and their applications. Living Reviews in Relativity, 7. Cited on May 12, 2007, URL www.livingreviews.org/lrr200410. Bardeen, J. M. (1970). A variational principle for rotating stars in General Relativity. Astrophys. J., 162, 71. Bardeen, J. M. (1971). A reexamination of the postNewtonian Maclaurin spheroids. Astrophys. J., 167, 425. Bardeen, J. M. (1973a). Rapidly rotating stars, disks, and black holes. In C. DeWitt and B. S. DeWitt (eds.), Black Holes, Les astres occlus (New York, Gordon and Breach Science Publishers), pp. 241–289. Bardeen, J. M. (1973b). Timelike and null geodesics in the Kerr metric. In C. DeWitt and B. S. DeWitt (eds.), Black Holes, Les astres occlus (New York, Gordon and Breach Science Publishers), pp. 215–240. Bardeen, J. M. and Horowitz, G. T. (1999). Extreme Kerr throat geometry: A vacuum analog of AdS2 × S2 . Phys. Rev. D, 60, 104030. Bardeen, J. M., Press, W. H. and Teukolsky, S. A. (1972). Rotating black holes: Locally nonrotating frames, energy extraction, and scalar synchrotron radiation. Astrophys. J., 178, 347. Bardeen, J. M. and Wagoner, R. V. (1971). Relativistic disks. I. Uniform rotation. Astrophys. J., 167, 359. Beig, R. and Chrus´ciel, P. (2006). Stationary black holes. In J.P. Françoise, G. L. Naber and Tsou S. T. (eds.), Encyclopedia of Mathematical Physics, vol. 5 (Oxford, Academic Press/Elsevier), pp. 38–44. Eprint: grqc/0502041. Belinski, V. A. and Zakharov, V. E. (1978). Integration of the Einstein equations by the inverse scattering problem technique and the calculation of the exact soliton solutions. Zh. Eskp. Teor. Fiz. Pis’ma, 75, 195. Belokolos, E. D., Bobenko, A. I., Enol’skii, V. Z., Its, A. R. and Matveev, V. B. (1994). AlgebroGeometric Approach to Nonlinear Integrable Equations. Nonlinear Dynamics (New York, Springer). Birkhoff, G. and Lynch, R. E. (1984). Numerical Solution of Elliptic Problems. Studies in Applied Mathematics (Philadelphia, SIAM). Bodo, G. and Curir, A. (1992). Models of selfgravitating accretion disks. Astron. Astrophys., 253, 318. Bonazzola, S., Gourgoulhon, E. and Marck, J.A. (1998). Numerical approach for high precision 3D relativistic star models. Phys. Rev. D, 58, 104020. Bonazzola, S., Gourgoulhon, E., Salgado, M. and Marck, J.A. (1993). Axisymmetric rotating relativistic bodies: A new numerical approach for ‘exact’ solutions. Astron. Astrophys., 278, 421. Bonazzola, S. and Schneider, J. (1974). An exact study of rigidly and rapidly rotating stars in general relativity with application to the Crab pulsar. Astrophys. J., 191, 273. Boyer, R. H. and Lindquist, R. W. (1967). Maximal analytic extension of the Kerr metric. J. Math. Phys., 8, 265. Buchdahl, H. A. (1959). General relativistic ﬂuid spheres. Phys. Rev., 116, 1027. Butterworth, E. M. and Ipser, J. R. (1976). On the structure and stability of rapidly rotating ﬂuid bodies in general relativity. I. The numerical method for computing structure and its application to uniformly rotating homogeneous bodies. Astrophys. J., 204, 200. Carter, B. (1973). Black hole equilibrium states. In C. DeWitt and B. S. DeWitt (eds.), Black Holes, Les astres occlus (New York, Gordon and Breach Science Publishers), pp. 57–214.
210
References
Chandrasekhar, S. (1939). An Introduction to the Study of Steller Structure (Chicago, University of Chicago Press). Chandrasekhar, S. (1967). The postNewtonian effects of General Relativity on the equilibrium of uniformly rotating bodies II. The deformed ﬁgures of the Maclaurin spheroids. Astrophys. J., 147, 334. Chandrasekhar, S. (1968). The virial equations of the fourth order. Astrophys. J., 152, 293. Chandrasekhar, S. (1969). Ellipsoidal Figures of Equilibrium (New Haven, Yale University Press). A revised edition was published by Dover, New York, in 1987. Chandrasekhar, S. (1970). Solutions of two problems in the theory of gravitational radiation. Phys. Rev. Lett., 24, 611. Chandrasekhar, S. and Friedman, J. L. (1972a). On the stability of axisymmetric systems to axisymmetric perturbations in general relativity. I. The equations governing nonstationary, and perturbed systems. Astrophys. J., 175, 379. Chandrasekhar, S. and Friedman, J. L. (1972b). On the stability of axisymmetric systems to axisymmetric perturbations in general relativity. II. A criterion for the onset of instability in uniformly rotating conﬁgurations and the frequency of the fundamental mode in case of slow rotation. Astrophys. J., 176, 745. Chandrasekhar, S. and Friedman, J. L. (1973). On the stability of axisymmetric systems to axisymmetric perturbations in general relativity. IV. Allowance for gravitational radiation in an oddparity mode. Astrophys. J., 181, 481. Christodoulou, D. (1970). Reversible and irreversible transformations in blackhole physics. Phys. Rev. Lett., 25, 1596. Chrus´ciel, P. T., Greuel, G.M., Meinel, R. and Szybka, S. J. (2006). The Ernst equation and ergosurfaces. Class. Quantum Grav., 23, 4399. Contopoulos, G. (1994). Order and chaos. In G. Contopoulos, N. K. Spyrou and L. Vlahos (eds.), Galactic Dynamics and NBody Simulations, Lecture Notes in Physics, vol. 433 (Berlin, Springer Verlag), pp. 33–100. Cook, G. B., Shapiro, S. L. and Teukolsky, S. A. (1992). Spinup of a rapidly rotating star by angular momentum loss: Effects of general relativity. Astrophys. J., 398, 203. Detweiler, S. L. and Lindblom, L. (1977). On the evolution of the homogeneous ellipsoidal ﬁgures. Astrophys. J., 213, 193. Dyson, F. W. (1892). The potential of an anchor ring. Phil. Trans. R. Soc. London, Ser. A, 184, 43. Dyson, F. W. (1893). The potential of an anchor ring – part II. Phil. Trans. R. Soc. London, Ser. A, 184, 1041. Eisenstat, S. C. (1974). On the rate of convergence of the Bergman–Vekua method for the numerical solution of elliptic boundary problems. SIAM J. Numer. Anal., 11, 654. Eriguchi, Y. and Hachisu, I. (1982). New equilibrium sequences bifurcating from Maclaurin sequence. Prog. Theor. Phys., 67, 844. Eriguchi, Y. and Sugimoto, D. (1981). Another equilibrium sequence of selfgravitating and rotating incompressible ﬂuid. Prog. Theor. Phys., 65, 1870. Ernst, F. (1968). New formulation of the axially symmetric gravitational ﬁeld problem. Phys. Rev, 167, 1175. Ernst, F. (1977). A new family of solutions of the Einstein ﬁeld equations. J. Math. Phys., 18, 233. Fischer, T., Horatschek, S. and Ansorg, M. (2005). Uniformly rotating rings in general relativity. Mon. Not. R. Astron. Soc., 364, 943. Fodor, G., Hoenselaers, C. and Perjés, Z. (1989). Multipole moments of axisymmetric systems in relativity. J. Math. Phys., 30, 2252.
References
211
Font, J. A. (2003). Numerical hydrodynamics in general relativity. Living Reviews in Relativity, 6. Cited on May 12, 2007, URL www.livingreviews.org/lrr200304. Friedman, J. L. and Ipser, J. R. (1992). Rapidly rotating relativistic stars. Phil. Trans. R. Soc. Lond. A, 340, 391. Friedman, J. L., Ipser, J. R. and Parker, L. (1986). Rapidly rotating neutron star models. Astrophys. J., 304, 115. Erratum (1990): Astrophys. J., 351, 705. Friedman, J. L., Ipser, J. R. and Parker, L. (1989). Implications of a halfmillisecond pulsar. Phys. Rev. Lett., 62, 3015. Friedman, J. L., Ipser, J. R. and Sorkin, R. D. (1988). Turningpoint method for axisymmetric stability of rotating relativistic stars. Astrophys. J., 325, 722. Friedman, J. L. and Schutz, B. F. (1978). Secular instability of rotating Newtonian stars. Astrophys. J., 222, 281. Geroch, R. (1970). Multipole moments. II. Curved space. J. Math. Phys., 11, 2580. Göpel, A. (1847). Entwurf einer Theorie der Abel’schen Transcendenten erster Ordnung. Crelle’s J. für Math., 35, 227. Gourgoulhon, E., Haensel, P., Livine, R., Paluch, E., Bonazzola, S. and Marck, J.A. (1999). Fast rotation of strange stars. Astron. Astrophys., 349, 851. Hachisu, I. and Eriguchi, Y. (1984). Bifurcation points on the Maclaurin sequence. Publ. Astron. Soc. Japan, 36, 497. Hansen, R. O. (1974). Multipole moments of stationary spacetimes. J. Math. Phys., 15, 46. Harrison, B. K. (1978). Bäcklund transformation for the Ernst equation of general relativity. Phys. Rev. Lett., 41, 1197. Hartle, J. B. and Sharp, D. H. (1967). Variational principle for the equilibrium of a relativistic, rotating star. Astrophys. J., 147, 317. Hawking, S. and Ellis, G. (1973). The Large Scale Structure of SpaceTime (Cambridge, Cambridge University Press). Heusler, M. (1996). Black Hole Uniqueness Theorems (Cambridge, Cambridge University Press). Hoenselaers, C., Kinnersley, W. and Xanthopoulos, B. C. (1979). Generation of asymptotically ﬂat, stationary spacetimes with any number of parameters. Phys. Rev. Lett., 42, 481. Horatschek, S. and Petroff, D. (2008). Uniformly rotating homogeneous rings in Newtonian gravity. Eprint: arXiv: 0802.0078. Kerr, R. P. (1963). Gravitational ﬁeld of a spinning mass as an example of algebraically special metrics. Phys. Rev. Lett., 11, 237. Kippenhahn, R. and Weigert, A. (1990). Stellar Structure and Evolution (Berlin, SpringerVerlag). Kleinwächter, A. (2000). Properties of the Neugebauer–Meinel solution. Ann. Phys. (Leipzig), 9, 99. Special Issue. Kleinwächter, A. (2001). Discussion of the theta formula for the Ernst potential of the rigidly rotating disk of dust. In A. Macias, J. L. CervantesCota and C. Lämmerzahl (eds.), Exact Solutions and Scalar Fields in Gravity (Kluwer Press, New York), pp. 39–51. Kleinwächter, A., Meinel, R. and Neugebauer, G. (1995). The multipole moments of the rigidly rotating disk of dust in general relativity. Phys. Lett. A, 200, 82. Komar, A. (1959). Covariant conservation laws in general relativity. Phys. Rev., 113, 934.
212
References
Komatsu, H., Eriguchi, Y. and Hachisu, I. (1989a). Rapidly rotating general relativistic stars – I. Numerical method and its application to uniformly rotating polytropes. Mon. Not. R. Astron. Soc., 237, 355. Komatsu, H., Eriguchi, Y. and Hachisu, I. (1989b). Rapidly rotating general relativistic stars – II. Differentially rotating polytropes. Mon. Not. R. Astron. Soc., 239, 153. Kordas, P. (1995). Reﬂectionsymmetric, asymptotically ﬂat solutions of the vacuum axistationary Einstein equations. Class. Quantum Grav., 12, 2037. Kowalewsky, S. (1885). Zusätze und Bemerkungen zu Laplace’s Untersuchung über die Gestalt der Saturnringe. Astronomische Nachrichten, 111, 37. Kramer, D. and Neugebauer, G. (1968). Zu axialsymmetrischen stationären Lösungen der Einsteinschen Feldgleichungen für das Vakuum. Commun. Math. Phys., 10, 132. Krazer, A. (1903). Lehrbuch der Thetafunktionen (Leipzig, Teubner). Kundt, W. and Trümper, M. (1966). Orthogonal decomposition of axisymmetric stationary spacetimes. Z. Phys., 192, 419. Labranche, H., Petroff, D. and Ansorg, M. (2007). The quasistationary transition of strange matter rings to a black hole. Gen. Rel. Grav., 39, 129. Landau, L. D. (1932). On the theory of stars. Phys. Z. Sowjetunion, 1, 285. Landau, L. D. and Lifshitz, E. M. (1980). Statistical Physics: Part 1 (Oxford, Pergamon Press). Lanza, A. (1992). Selfgravitating thin disks around rapidly rotating black holes. Astrophys. J., 389, 141. Lattimer, J. M., Prakash, M., Masak, D. and Yahil, A. (1990). Rapidly rotating pulsars and the equation of state. Astrophys. J., 355, 241. Lewis, T. (1932). Some special solutions of the equations of axially symmetric gravitational ﬁelds. Proc. Roy. Soc. Lond. A, 136, 176. Lichtenstein, L. (1933). Gleichgewichtsﬁguren rotierender Flüssigkeiten (Berlin, Springer). Lindblom, L. (1992). On the symmetries of equilibrium stellar models. Phil. Trans. R. Soc. Lond. A, 340, 353. Lindblom, L. and Hiscock, W. A. (1983). On the stability of rotating stellar models in general relativity theory. Astrophys. J., 267, 384. Maclaurin, C. (1742). A Treatise on Fluxions (Edinburgh, Ruddimans). Maison, D. (1978). Are the stationary, axially symmetric Einstein equations completely integrable? Phys. Rev. Lett., 41, 521. MasoodulAlam, A. K. M. (2007). Proof that static stellar models are spherical. Gen. Rel. Grav., 39, 55. Meinel, R. (2002). Black holes: A physical route to the Kerr metric. Ann. Phys. (Leipzig), 11, 509. Meinel, R. (2004). Quasistationary collapse to the extreme Kerr black hole. Ann. Phys. (Leipzig), 13, 600. Meinel, R. (2006). On the black hole limit of rotating ﬂuid bodies in equilibrium. Class. Quantum Grav., 23, 1359. Meinel, R. and Kleinwächter, A. (1995). Dragging effects near a rigidly rotating disk of dust. In J. B. Barbour and H. Pﬁster (eds.), Mach’s Principle: From Newton’s Bucket to Quantum Gravity (Boston, Birkhäuser), pp. 339–346. Meinel, R. and Neugebauer, G. (1995). Asymptotically ﬂat solutions to the Ernst equation with reﬂection symmetry. Class. Quantum Grav., 12, 2045. Montero, P. J., Rezzolla, L. and Yoshida, S. (2004). Oscillations of vertically integrated relativistic tori – II. Axisymmetric modes in a Kerr spacetime. Mon. Not. R. Astron. Soc., 354, 1040.
References
213
Neugebauer, G. (1979). Bäcklund transformations of axially symmetric stationary gravitational ﬁelds. J. Phys. A, 12, L67. Neugebauer, G. (1980a). A general integral of the axially symmetric stationary Einstein equations. J. Phys. A, 13, L19. Neugebauer, G. (1980b). Recursive calculation of axially symmetric stationary Einstein ﬁelds. J. Phys. A, 13, 1737. Neugebauer, G. (1988). Thermodynamics of rotating bodies. In Z. Perjés (ed.), Relativity Today (Singapore, World Scientiﬁc), pp. 134–146. Neugebauer, G. (1996). Gravitostatics and rotating bodies. In G. S. Hall and J. R. Pulham (eds.), General Relativity (London, SUSSP Publications and Institute of Physics Publishing), pp. 61–81. Neugebauer, G. (2000). Rotating bodies as boundary value problems. Ann. Phys. (Leipzig), 9, 342. Neugebauer, G. and Herold, H. (1992). Gravitational ﬁelds of rapidly rotating neutron stars: Theoretical foundation. In J. Ehlers and G. Schäfer (eds.), Relativistic Gravity Research: With Emphasis on Experiments and Observations, Lecture Notes in Physics, vol. 410 (Berlin, Springer), pp. 305–318. Neugebauer, G., Kleinwächter, A. and Meinel, R. (1996). Relativistically rotating dust. Helv. Phys. Acta, 69, 472. Neugebauer, G. and Kramer, D. (1983). Einstein–Maxwell solitons. J. Phys. A, 16, 1927. Neugebauer, G. and Meinel, R. (1993). The Einsteinian gravitational ﬁeld of a rigidly rotating disk of dust. Astrophys. J., 414, L97. Neugebauer, G. and Meinel, R. (1994). General relativistic gravitational ﬁeld of a rigidly rotating disk of dust: Axis potential, disk metric and surface mass density. Phys. Rev. Lett., 73, 2166. Neugebauer, G. and Meinel, R. (1995). General relativistic gravitational ﬁeld of a rigidly rotating disk of dust: Solution in terms of ultraelliptic functions. Phys. Rev. Lett., 75, 3046. Neugebauer, G. and Meinel, R. (2003). Progress in relativistic gravitational theory using the inverse scattering method. J. Math. Phys., 44, 3407. Nishida, S. and Eriguchi, Y. (1994). A general relativistic toroid around a black hole. Astrophys. J., 427, 429. Novikov, S., Manakov, S. V., Pitaevskii, L. P. and Zakharov, V. E. (1984). Theory of Solitons: The Inverse Scattering Method (New York, Consultants Bureau). Nozawa, T., Stergioulas, N., Gourgoulhon, E. and Eriguchi, Y. (1998). Construction of highly accurate models of rotating neutron stars – comparison of three different numerical schemes. Astron. Astrophys. Supp., 132, 431. Oppenheimer, J. R. and Volkoff, G. M. (1939). On massive neutron cores. Phys. Rev., 55, 374. Ostriker, J. (1964). The equilibrium of selfgravitating rings. Astrophys. J., 140, 1067. Papapetrou, A. (1966). Champs gravitationnels stationnaires à symmétrie axiale. Ann. Inst. H. Poincaré A, 4, 83. Peshier, A., Kämpfer, B. and Soff, G. (2000). The equation of state of deconﬁned matter at ﬁnite chemical potential in a quasiparticle description. Phys. Rev. C, 61, 045203. Petroff, D. (2003). PostNewtonian Maclaurin spheroids to arbitrary order. Phys. Rev. D, 68, 104029. Petroff, D. and Horatschek, S. (2008). Uniformly rotating polytropic rings in Newtonian gravity. Eprint: arXiv: 0802.0081.
214
References
Petroff, D. and Meinel, R. (2001). PostNewtonian approximation of the rigidly rotating disk of dust to arbitrary order. Phys. Rev. D, 63, 064012. Poincaré, H. (1885). Sur l’équilibre d’une masse ﬂuide animée d’un mouvement de rotation. Acta mathematica, 7, 259. Rezzolla, L., Yoshida, S. and Zanotti, O. (2003). Oscillations of vertically integrated relativistic tori – I. Axisymmetric modes in a Schwarzschild spacetime. Mon. Not. R. Astron. Soc., 344, 978. Robinson, D. C. (1975). Uniqueness of the Kerr black hole. Phys. Rev. Lett., 34, 905. Roche, É. (1873). Essai sur la constitution et l’origine du système solaire. Mém. de la section des sciences, Acad. des sciences et lettres de Montpellier, 1, 235. Rosenhain, G. (1850). Auszug mehrerer Schreiben des Herrn Dr. Rosenhain an Herrn Prof. C.G.J. Jacobi über die hyperelliptischen Transcendenten. Crelle’s J. für Math., 40, 319. Schöbel, K. and Ansorg, M. (2003). Maximal mass of uniformly rotating homogeneous stars in Einsteinian gravity. Astron. Astrophys., 405, 405. Schoen, R. and Yau, S.T. (1979). Positivity of the total mass of a general spacetime. Phys. Rev. Lett., 43, 1457. Schutz, B. F. (1972). Linear pulsations and stability of differentially rotating stellar models. II. Generalrelativistic analysis. Astrophys. J. Supp., 24, 343. Schwarzschild, K. (1916). Über das Gravitationsfeld einer Kugel aus inkompressibler Flüssigkeit nach der Einsteinschen Theorie. Sitzungsberichte der Königlich Preussischen Akademie der Wissenschaften, 1, 424. Sen, N. R. (1934). On the equilibrium of an incompressible sphere. Mon. Not. R. Astron. Soc., 94, 550. Shapiro, S. L. and Shibata, M. (2002). Collapse of a rotating supermassive star to a supermassive black hole: Analytic determination of the black hole mass and spin. Astrophys. J., 577, 904. Shapiro, S. L. and Teukolsky, S. A. (1983). Black Holes, White Dwarfs, and Neutron Stars. The Physics of Compact Objects (New York, John Wiley). Shibata, M., Taniguchi, K. and Uryu, K. (2003). Merger of binary neutron stars of unequal mass in full general relativity. Phys. Rev. D, 68, 084020. Smarr, L. (1973). Mass formula for Kerr black holes. Phys. Rev. Lett., 30, 71. Stahl, H. (1896). Theorie der Abel’schen Funktionen (Leipzig, Teubner). Stephani, H. (2004). Relativity (Cambridge, Cambridge University Press). Stephani, H., Kramer, D., MacCallum, M., Hoenselaers, C. and Herlt, E. (2003). Exact Solutions of Einstein’s Field Equations, 2nd edn. (Cambridge, Cambridge University Press). Stergioulas, N. (2003). Rotating stars in relativity. Living Reviews in Relativity, 6. Cited on 11 April 2007, URL www.livingreviews.org/lrr20033. Stergioulas, N. and Friedman, J. L. (1995). Comparing models of rapidly rotating relativistic stars constructed by two numerical methods. Astrophys. J., 444, 306. Stoner, E. C. (1932). The minimum pressure of a degenerate electron gas. Mon. Not. R. Astron. Soc., 92, 651. Thorne, K. S. (1967). The general relativistic theory of stellar structure and dynamics. In C. DeWitt, E. Schatzman and P. Véron (eds.), High Energy Astrophysics, Volume 3 (New York, Gordon and Breach), pp. 259–441. Thorne, K. S. (1980). Multipole expansions of gravitational radiation. Rev. Mod. Phys., 52, 299. Tolman, R. C. (1934). Relativity, Thermodynamics and Cosmology (Oxford, Oxford University Press).
References
215
Tolman, R. C. (1939). Static solutions of Einstein’s ﬁeld equations for spheres of ﬂuid. Phys. Rev., 55, 364. Tooper, R. F. (1964). General relativistic polytropic ﬂuid spheres. Astrophys. J., 140, 434. Tooper, R. F. (1965). Adiabatic ﬂuid spheres in general relativity. Astrophys. J., 142, 1541. Trümper, M. (1967). Einsteinsche Feldgleichungen für das axialsymmetrische, stationäre Gravitationsfeld im Innern einer starr rotierenden Flüssigkeit. Z. Naturforsch., 22a, 1347. Wald, R. M. (1984). General Relativity (Chicago, University of Chicago Press). Wigley, N. M. (1970). Mixed boundary value problems in plane domains with corners. Math. Z., 115, 33. Will, C. M. (1974). Perturbation of a slowly rotating black hole by a stationary axisymmetric ring of matter. I. Equilibrium conﬁgurations. Astrophys. J., 191, 521. Will, C. M. (1975). Perturbation of a slowly rotating black hole by a stationary axisymmetric ring of matter. II. Penrose processes, circular orbits and differential mass formulæ. Astrophys. J., 196, 41. Wilson, J. R. (1972). Models of differentially rotating stars. Astrophys. J., 176, 195. Wolf, T. (1998). Structural equations for Killing tensors of arbitrary rank. Computer Phys. Commun., 115, 316. Wong, C. Y. (1974). Toroidal ﬁgures of equilibrium. Astrophys. J., 190, 675. Zanotti, O., Font, J. A., Rezzolla, L. and Montero, P. J. (2005). Dynamics of oscillating relativistic tori around Kerr black holes. Mon. Not. R. Astron. Soc., 356, 1371. Zel’dovich, Y. B. and Novikov, I. D. (1971). Relativistic Astrophysics, vol. 1 (Chicago, University of Chicago Press).
Index
4acceleration, 24 4velocity, 5, 24
conﬂuent, 43 Buchdahl limit, 18, 39, 141
Abelian differential, 55 accretion discs, 176 angular momentum, 2, 13 disc of dust, 79 Kerr, 27, 112 Maclaurin disc, 38 Maclaurin spheroid, 36 angular velocity, 5 Maclaurin disc, 38 of locally nonrotating observer, 91 of the horizon, 26, 28, 109, 112 asymptotic behaviour, 13, 36 asymptotic ﬂatness, 2 axial symmetry, 3 axis of symmetry, 4, 42
centrifugal force, 25, 38 centrifugal potential, 17 CFS instability, 180 Chebyshev coefﬁcients, 122 Chebyshev expansion, 116 Chebyshev polynomials, 116 chemical potential, 13 Christoffel symbols, 24 circular orbits, 28, 91, 97 circumferential radius, 149 corotating potentials, 5, 8, 17 corotating system, 5, 10, 15, 45, 80, 109 collapse, 19, 166 collocation point, 121, 135 compactiﬁcation, 116, 129, 132 conformal transformation, 9 coordinate mappings, 127 covariant derivative, 6 cylindrical coordinates, 5
Bäcklund transformation, 112, 204 Bernoulli–l’Hospital rule, 112 bifurcation points, 139 binding energy, 157, 160 disc of dust, 87, 95 black hole, 26, 31, 108 degenerate, 176 extreme Kerr, 29, 31, 146, 152 Kerr, 27 Schwarzschild, 29 surrounded by a ﬂuid ring, 166 black hole limit, 31, 147 disc of dust, 104 rings, 165 black hole uniqueness, 33, 112 boundary conditions horizon, 27, 167 boundary of the ﬂuid body, 6 boundary value problem, 40 black hole, 109 disc, 22 Boyer–Lindquist coordinates, 27 branch cut, 43 branch points, 43
Dirac delta distribution, 20 direct orbits, 28, 91 disc limit, 19, 153 disc of dust, 25, 38, 40, 153 dust limit, 12, 25 Dyson rings, 144 Einstein’s ﬁeld equations, 7 vacuum, 10 elliptic coordinates oblate, 34 elliptic functions, 70, 71, 86, 190 elliptic integrals, 86, 190 embedding diagram, 30 energydensity, 6, 10 internal, 10 energymomentum tensor disc, 24 dust, 12 perfect ﬂuid, 6 enthalpy, 6, 13, 115
216
Index equation of state, 6, 10 barotropic, 153 completely degenerate, ideal gas of neutrons, 11, 161 homogeneous ﬂuid, 10, 137 polytropic, 11, 154 strange quark matter, 12, 162 equator, 25 equatorial plane, 28 ergosphere, 14, 141, 144 disc of dust, 88 Kerr black hole, 28 ergosurface, 14 Ernst equation, 2, 10, 42, 109 corotating, 23 Ernst potential, 10, 82, 204 Kerr solution, 112 Euler equation, 17, 36 extreme Kerr solution, 112
isolated body, 1 isotropic coordinates, 39
far ﬁeld, 1 Fermi gas completely degenerate, ideal, 11, 161 Fermi–Dirac statistics specialrelativistic, 11 ﬁeld equations, 7, 114 frame dragging, 171 free boundary value problem, 128
Laplace equation, 16, 36, 182 twodimensional, 9, 21 Legendre functions, 124, 139 Legendre polynomials, 139 Lense–Thirring effect, 2 Lewis–Papapetrou metric, 4 line element, 4, 114 far ﬁeld, 1 line integration, 23 linear problem, 41 local inertial system, 2 locally nonrotating observers, 8, 91 Lyapunov functional, 179
Gauss’ theorem, 20 general relativity, 17 geodesic motion, 12, 25, 28, 91, 96 Gibbs phenomenon, 119, 120 global problem, 10 gravitational energy, 37 Maclaurin disc, 38 gravitational radiation, 3 gravitomagnetic effects, 2, 166 gravitomagnetic potential, 4 Hamiltonian system, 96 Heuman’s lambda function, 58, 187 holomorphic function, 47 homogeneous ﬂuids, 10 horizon, 26, 31, 109, 167 area, 175 degenerate, 33, 112, 168 dynamical, 176 isolated, 176 hyperelliptic functions, 54 hypersurface null, 26, 32 spacelike, 13 timelike, 22 inﬁnity, 2, 5, 7 spatial, 5, 42 integrability condition, 9, 42 inverse method, 41, 108 direct problem, 42
Jacobi’s inversion problem, 54 Jacobi’s zeta function, 58, 86, 187 Jacobian matrix, 123 jump matrix, 49 Kerr metric, 27, 108 Kerr solution, 27, 112 extreme, 107, 113 Killing vector, 4, 14, 26 kinetic energy Maclaurin disc, 38 of rotation, 36 Komar mass, 169 individual, 170 total, 170 Kronecker symbol, 117
Maclaurin disc, 37 Maclaurin sequence, 140 Maclaurin spheroid, 34 marginally bound orbit, 28 marginally stable orbit, 28 mass baryonic, 13, 40, 79 Christodoulou, 175, 176 disc of dust, 79 gravitational, 2, 13, 39, 79 Kerr, 27, 112 Komar, 169 Maclaurin spheroid, 36 massdensity, 6, 10 baryonic, 6, 10 massshed parameter, 141 massshedding limit, 25, 141, 146, 150, 181 matching conditions, 2 metric, 114 asymptotic behaviour, 14 axisymmetric perfect ﬂuid body in stationary rotation, 4 disc of dust, 57 far ﬁeld, 1 Lewis–Papapetrou, 4 Minkowski, 2, 5, 16 MIT bag constant, 12
217
218 MIT bag model, 12, 163 moment of inertia, 37 multipole moments, 82, 107, 193 nearhorizon geometry, 108 neutron gas, 161 neutron stars, 3, 11 Newton–Raphson scheme, 123 Newtonian limit, 16 disc of dust, 52, 93 Schwarzschild spheres, 40 Newtonian potential, 16, 34, 181 generalized, 4, 24 nonrotating limit, 17 normal vector, 13, 22, 26, 33 orthogonal transitivity, 4, 33 partial derivatives, 8 perfect ﬂuid, 3, 6 photon orbit, 28 Poisson equation, 16, 36, 182 Poisson integral, 34 polytrope, 11, 157 polytropic constant, 11, 154 polytropic exponent, 11 polytropic index, 11, 154 positive mass theorem, 170 postNewtonian expansion disc limit, 19 disc of dust, 95 Maclaurin spheroids, 139 pressure, 6 pseudospectral method, 115 radius ratio, 135 redshift, 7 reﬂectional symmetry, 20, 25, 83, 204 retrograde orbits, 28, 91 Riemann matrix, 56 Riemann surface, 43, 54 Riemann–Hilbert problem, 49, 51 rigid rotation, 3, 5 rings, 145 black hole limit, 165 Roche model, 184 Rosenhain’s theta functions, 67, 76, 189 rotation rigid, 3, 5 with respect to inﬁnity, 2, 5, 175 with respect to the ‘ﬁxed stars’, 2 with respect to the local inertial system, 2 Schwarzschild coordinates, 18 Schwarzschild metric, 39 Schwarzschild solution, 112, 141 interior, 39 seed solution, 203 soliton theory, 2, 41 spectral approximation, 119
Index spectral coefﬁcients, 117 spectral expansion, 116 spectral parameter, 41 spectral resolution, 119 speed of light, 92 speed of sound, 11 spherical symmetry, 18 spheroidal conﬁgurations, 129 stability, 3, 177 dynamical, 177, 179 secular, 137, 177 static model, 17 stationarity, 3 local, 14 stationary limit, 15 strange matter, 12, 162 strange quark matter, 6, 12 subdomains, 128 superluminal motion, 15 surface of the ﬂuid, 7 shape, 10, 34 surface condition, 6, 31, 35 Newtonian, 17 surface energydensity, 22 surface gravity, 112, 167 surface layer, 20 surface massdensity, 79 Maclaurin disc, 38 symmetry axis, 4 temperature, 3, 6 thermodynamic equilibrium, 3 theta functions, 54, 57, 187 elliptic, 58, 70 hyperelliptic, 187 moduli, 59 ultraelliptic, 58, 187 throat geometry, 30, 33, 108 Tolman condition, 3, 13 Tolman–Oppenheimer–Volkoff equation, 18 topology, 146, 169 toroidal conﬁgurations, 130 twobody limit, 153 twobody systems stationary, 166 variational principle, 13 velocity of rotation, 8, 16, 91 viscosity, 3, 177 volume element, 13 Weierstrass function, 105 Weyl coordinates canonical, 9, 21, 109 Weyl–Lewis–Papapetrou coordinates, 112 white dwarfs, 11 zero angular momentum observers, 8, 167