R. A. Breuer J. Ehlers
R. A. Breuer J. Ehlers
R. A. Breuer J. Ehlers
R. A. Breuer; J. Ehlers
Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences, Vol.
370, No. 1742. (Mar. 28, 1980), pp. 389-406.
Stable URL:
http://links.jstor.org/sici?sici=0080-4630%2819800328%29370%3A1742%3C389%3APOHEWT%3E2.0.CO%3B2-E
Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences is currently published by The
Royal Society.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/rsl.html.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.
http://www.jstor.org
Fri Feb 8 08:35:38 2008
Proc. R. Sac. Lond. A 370, 389-406 (1980)
Printed in Great Britain
1. I N T R O D U C T IAO
NND ASTROPHYSICAL MOTIVATION
Since its discovery the X-ray source Cyg X - 1 has been the first and so far strongest
candidate to be a binary system containing a black hole. Also, Cir X - 1 has been
suggested to have a black hole as one of its binary components, and up to now this
idea is compatible with all known data from Cir X - 1 . The main reasons in support
of a black hole in these systems are (i) the mass estimates derived from the mass
funct'ion of the system together with plausible assumptions on the inclination of the
system relative to the line of sight, and (ii) the fact that the X-ray intensity shows
irregular variations on a millisecond timescale. The latter property strongly
[ 389 I
390 R. A. Breuer and J. Ehlers
suggests that the radiation is emitted within a compact region of less than 100 km in
diameter, possibly during the final spiral motion of accreting matter towards the
black hole which possibly has no magnetic field - as is expected from astrophysical
black holes. They should neutralize quickly in case charge-separating mechanisms
are active in their environments.
These features are independent of the details of the actual accretion process:
whether stellar wind from the super-giant companion drives the accretion (MeszBros
1978),or whether an accretion disk forms as a result of mass transfer in a Roche lobe
overflow, not t o mention the difficulties one encounters with detailed disk models
(Stewart 1976a, b; Drury & Stewart 1978; Thorne & Price 1975 ; Shakura & Sunyaev
1973; Lightmann & Eardley 1974; Shapiro et al. 1976; Livio & Shaviv 1977). How-
ever, the indications for the existence of a black hole in Cyg X - l are somewhat
indirect and could in principle be questioned: neutron stars with unorthodox
equations of state could have higher maximal masses and could thus replace the
black hole; in addition a neutron star with aligned magnetic field or an unmagnetized
neutron star would show no regular pulse structure in its X-rays. What one needs
is more clear-cut, direct observational evidence in favour of a black hole in Cyg X-1.
Such evidence might be provided by observation of the polarization of X-rays
from sources involving neutron stars or black holes (Rees 1975). AS for neutron
stars, X-ray polarimetry could reveal details of the beaming and accretion mech-
anisms, and for black holes it would give evidence for the structure and the position
angle of the accretion disk.
The polarization properties of X-rays from Cyg X-1 seem to be sensitive to strong
general relativistic effects. Satellite measurements claim an upper limit of 3 % linear
polarization a t 2.6 keV in Cyg X-1 (Novick etal. 1977; Chanan et al. 1979).Theoreti-
cally, their polarization properties have been first calculated in the Newtonian
approximation (Angel 1969; Lightman & Shapiro 1976).I n a relativistic calculation
Stark & Connors (1977a, b) find that for rays from the surface of an accretion disk
around Cyg X-I, travelling to the observer through vacuum, the direction of linear
polarization may vary as a function of energy up to 100' owing to general relativistic
effects. Equally, the degree of polarization differs from the Newtonian value. I n
fact, for the X-ray energies of interest between 1 and 100 keV they predict about
half the Newtonian value for the so-called 'one' temperature disk model. The
variation of such effects with the energy of X-rays - though depending crucially
on the particular disk model and assumptions about the emission mechanism -
should allow to differentiate observationally between a fast-rotating black hole or,
alternatively, a slowly rotating black hole or a neutron star. According to Stark &
Connors, for the latter the effect should be smaller by one order of magnitude.
The influence of any magnetic fields on the accretion disk has been ignored by
Stark & Connors. Although one might expect a mostly chaotic magnetic field of the
order of lo7 G (MBszAros et al. 1977),it is quite unclear to what extent a magnetized
corona of the disk could affect the polarization pattern of the X-rays. Theoretically,
a formalism is required here to describe the propagation of electromagnetic waves in
Wave propagation in plasmas in curved space-time. I 391
a magnetized plasma embedded in a strong background gravitational field. To what
extent plasma effects - dispersion, refraction and Faraday rotation - play a role for
Cyg X-1 can be decided only after such a calculation is a t hand and is compared with
the data.
Other astrophysical situations, where plasma effects in curved space-time may
influence wave propagation, can occur in galactic nuclei when a massive black hole is
present, and in pulsar magnetospheres of rotating neutron stars, but also in cos-
mology. The cosmological microwave radiation can be affected due to intergalactic
magnetic fields and when the interstellar gas is reionized a t late cosmological times
redshift x 8) (Anile & Breuer 1977).
Generally, one cannot solve Maxwell's equations exactly for waves propagating
in a relativistic medium. Instead one seeks a geometrical optics approximation, not
only in the style of Hamilton's geometrical theory of rays and wave surfaces (Synge
1960),but a refined version allowing also for frequencies, polarization states and the
propagation of amplitudes along rays leading to a radiative transfer equation. The
standard treatments using plane wave solutions to describe wave propagation in a
homogeneous plasma in flat space-time (Ginzburg 1964; Stix 1962) are no longer
applicable when inhomogeneities are present in the plasma or the underlying space-
time is curved. Physically, the reason for this is that the plane wave ansatz loses its
meaning. There are simply no plane waves in the more complicated situations
mentioned. Mathematically, plane waves are Fourier components of the original
field in x - t space. However, Fourier transformation is no longer possible: it pre-
supposes that the quantities to be Fourier-analyzed are defined on a space which has
an Abelian translation group.
However, when the wavelengths are small in comparison with the scale of plasma
inhomogeneities or, say, the Schwarzschild radius of the black hole, intuitively the
plane wave formalism should still be valid locally. Thus the desired formalism has to
satisfy two properties: locally it should resemble a Fourier transformation; globally,
the long-range effects of inhomogeneities of the plasma and the gravitational fields
have to show up in the propagation laws. So far, in our opinion, only several incom-
plete attempts are known towards a systematic derivation of such a formalism in the
framework of general relativity (Madore 1974; Bi6&k & Hadrava 1975; Anile &
Pantano 1977,1979).It is our aim in this and the subsequent paper to provide such
a rigorous treatment for a simple fluid model of the plasma which may be moving in
an arbitrary manner in an inhomogeneous, non-stationary, strong gravitational
field. The high-frequency approximation of electromagnetic waves in a transparent,
isotropic, dispersion-free, moving medium in curved space-time has been treated
by Ehlers (1967). The procedure developed in this paper is a special case of a two-
scale method as given for fluid dynamics in flat space-time by Witham (1965 a, b).
I n the next section we list the basic assumptions and equations for the background
upon which we will perform perturbation theory later on. An existence and unique-
ness theorem is established for the background. I n $ 3 we analyse the system of
differential equations governing linearized perturbations on this background
392 R. A. Breuer and J. Ehlers
resulting in a third-order evolution equation for the potential of the perturbations.
I n § 4 we describe quite generally the formal scheme to obtain oscillatory asymptotic
solutions of such evolution equations using a two-scale method.
I n termsof the de Rham operator q for vector fields, OAa = VgAa - RtAb,(2.2)then
gives
OAa = - enua - J a . (2.13)
Differentiating (2.13) along the electron world lines, we obtain, as a consequence of
the system (2.1-4), the system
This system of nine quasilinear equations for the nine unknowns (A,, ua, n) is, as
recognized by Choquet-Bruhat (1958),a strictly hyperbolic, diagonal Leray system.
I n fact the operators appearing on the left-hand side of (2.14, 15, 16) form a com-
patible set of strictly hyperbolic operators, and the derivatives of the unknowns
occur such as to obey the conditions of Leray's theorem. (For a concise formulation
of this theorem see, for example, Choquet-Bruhat et al. (1977), p. 441. I n the nota-
tion used there, the 'indices' of the unknowns and equations of the system here
considered are as follows :
Hence, according to Leray's theorem, the system (2.14,15, 16) has a unique solution
for any system of sufficiently (finitely) differentiable initial data
(Since the metric is given, these data are equivalent to initial values for the tensors
A,, V, A,, V,, A, on Z, where V,, = V,V,.) Suppose (Aa,ua, n) solves (2.14, 15, 16).
Then, if Ir',, is defined by (2.1I), equations (2.1, 3, 4) are satisfied, and because of
(2.14) and (2.16), the tensor fields (OAa+enua+ J a ) and uaua are covariantly
constant on ua-world lines. Hence, if the initial values satisfy (2.5) and (2.13), then
these equations hold in the whole domain of dependence of the data. Furthermore
(2.6), (2.4), (2.13),andthe identity V, OAa = OV, Aaimply OV, Aa = 0. Therefore,
if V, Aa and a, V, Aa vanish on 2, then the Lorentz condition (2.12) holds in the
domain of dependence, and in that case (2.11) and (2.13) imply (2.2). We have
shown that the system of constraints
u,ua = - 1, (2.5)
a0VaAa = 0 (2.18)
is preserved under the system (2.14, 15, 16) of evolution equations; any set of initial data
(2.17) which satisfies these constraints determines a unique solution of (2.14, 15, 16);
such a solution determines, via (2.11)a solution of the original system (2.1-5). It is also
clear from the preceding reasoning that any solution of (2.1-5) can be obtained in this
way.
Note that the initial data set (2.17) contains 17 functions. Since there are seven
constraints, ten functions can be chosen arbitrarily. However, the Lorentz-gauge is
preserved under the restricted gauge transformations A,+ A, + 8, A with [7A = 0,
and such a gauge function A is determined by the initial data A and a. A on Z. This
reduces the number of physically essential initial data from ten to eight, in accord-
ance with the previous counting.
To complete the proof of statement (I)of the theorem, we suppose that (Ir",,, ua, n)
Wave propagation in plasmas in curved space-time. I 395
are initial data on 2 which obey the constraints (2.9, 10, 5). Then, by (2.9), there
exist on Z1 functions A, such that FAp = 2BrA Apl;these Ah are determined up to
+
changes A,+ A , aAA.Choose A, arbitrarily on 2, and define the initial data
3, A, and a, A, by a, A, = Fa, + ahA, and V, Aa = 0, respectively. This provides, on
2, A, and a, A, or, equivalently, A, and V, A,, and by construction these data
satisfy (2.11) and (2.12). Next, compute the initial values aooAafrom (2.13). This
then fixes V,, A, on 2. A straight-forward calculation shows that the constraint
(2.lo), combined with the 0-component of (2.13), implies a,(V, Aa) = 0. Conse-
quently initial data of the original system (2.7, 8, 3, 4) satisfying the corresponding
constraints (2.9, 10, 5) determine uniquely up to 2 gauge-initial data ( A and A,), an
initial data set (2.17)obeying the constraints (2.5, 12, 13, 18). The solution belonging
to these data then furnishes a solution of (2.1-5) with the prescribed initial values
(Fa,, ua, n). Uniqueness follows from the uniqueness of the Cauchy-problem for
(2.14-16) combined with the fact, established above, that the initial data (2.17)are
determined by those of the original system exactly up to 'initial gauge transfor-
mations '.
To prove statement (11)of the theorem we first verify that the system (2.5, 12, 13,
18) of constraints is linearization stable. I n fact, let (Aa, ua, n ) satisfy those con-
straints, and let (Aa,Qa,&) obey the linearized constraints
u , =~0, ~ (2.19)
ao(vaAa)= o.
Define a 1-parameter family (3,
Ua, n), of fields on Z by putting
(11)If (Fa,, ...) and (Fab, ...) obey equations (2.1-6) and (3.1-5),
respectively, then (F,, ... ) approximates a solution of the full system (2.1-6)
provided s is suficiently small.
For the linear system (3.1-5) the decomposition into evolution equations and con-
straints can be carried out, in contrast to the situation for the full equations (2.1-5),
in an intrinsic, physically and mathematically preferred way since uaV, is apreferred
time-derivative.
The purpose of this section is to reduce the rather complex system (3.1-5) to a
simpler one which is better suited for calculations, in particular for the treatment
of high-frequency waves. We shall eliminate redundant equations and unknowns
and, in particular, get rid of constraints.
Equation (3.4) is implied by (3.2) and can therefore be omitted from the system.
Next, we insert (2.22) into (3.3) and use (3.5) to get
ubVbt2a + (w; + 0;) .tib= (elm)(F;.tib+ @gub). (3.6)
The equation obtained by transvecting this equation with ua is implied by (2.3) and
(3.5) whence (3.6), or rather (3.2), can be replaced by its 'spatial part',
The pair (3.I, 10) of equations for Pa,is equivalent to the system (3.1-5) modulo the
background equations (2.1-5); i.e. any solution of (3.1-5) satisfies (3.1, 10); and if
Pa, obeys (3.1, 10) and /iia,fiare defined by (3.8, 9), then (3.1-5) holds.
The reduced system (3.I, 10) consists of seven partial differential equations for the
six components of the Faraday tensor pab. Only six of these equations are evolution
equations; the space-part of (3.1)is a constraint. To get rid of this constraint and to
reduce further the number of equations and unknowns we introduce a potential
A, for Pa,,
which can be done without loss of generality. The remaining gauge freedom,
Aa-+A^,+VaA, is restricted by the condition uaVaA = 0 which says that A is
constant on the unperturbed electron world lines. The gauge-freedom associated
with (3.12) is, therefore, less than that associated with the Lorentz gauge. Landau
gauge initial data are determined, for a given field Pab,up to only one arbitrary
function on the initial hypersurface.
If (3.11) is inserted into (2.10) and the gauge condition (3.12) is used to simplify
some terms, the equation
4. OSCILLATORYA S Y M P T O T I C S O L U T I O N S AS A P P R O X I M A T I O N S TO
HIGH-FREQUENCY SOLUTIONS O F LINEAR, HYPERBOLIC SYSTEMS
where the eiconal S is a C-valued function and the amplitudes 8,V , are Cm-valued
on Rn.
t For this section compare, for example, Courant & Hilbert (1962); Duistennaat (1974)
Lax (1957) ; the treatment given here is a straightforward generalization of those given in
these references.
Wave propagation in plasmas in curved space-time. I 401
We first describe an algorithm for determining S and the V,. We write d: =
(a,, ...,a,) for the gradient operator a d put
1: = dS. (4.4)
By means of the operator identity
e(-i/4 S de(i/4S = d + (i/,c) 1 t45)
the equation D(e('/dSV) = 0 is transformed into L. V = 0 where the coefficients of
L(c) = e(-ilc)sD(s)e('lc)Scontain the wave covector 1. Applicationof (4.5)to (4.1)gives
P
L(s) = C €fLj, (4.6)
j=O
with an as yet unrestricted complex scalar amplitude v,. This is all the information
obtained in the first step. It suffices to determine all coefficients of the operator (4.6).
Wave propagation i n plasmas i n curved space-time. I 403
I n first order in s, (4.8) demands
Using (4.14,16,7, 15),recalling the definition of T and assuming h =/= 0, we get from
the last equation the transport equation
Taa, v, +fv, = 0, (4.18)
wheref is known from the first step. (Both T aand f depend on the choice of N and R
in a neighbourhood of the ray considered. I n applications one chooses R such that it
represents unit amplitude or unit intensity so that v, measures the magnitude of the
wave.) This linear, homogeneous equation determines the amplitude v, along each
ray via initial conditions and, by (4.16),the leading contribution V, to the amplitude
V. The initial conditions for v, determine the projile of the wave. If the initial value
of v, has compact support, V,will be non-zero only in a spatially compact ray bundle;
this case corresponds to a pulse-wave.
Once V, has thus been determined, (4.17)is solvable for F since N . (iL,.V,) = 0. If
R1 denotes one particular solution of
L,.R,+iL,.V, = 0,
the general solution of (4.17) is
T(= R1+vlR (v,s@).
After the second step one therefore knows e(il"s V, everywhere, and one knows the
form e(il"s([vo- isv,] R- isR,) of U in (4.3) to first order, but not yet the value of v,.
If the coefficients of D(s) and the initial values of X,v,, v,, .. . are C m or if one takes
derivatives in a distributional sense, this procedure can be continued indefinitely.
The higher-order amplitudes v,, v,, .. . are determined by initial values and inhomo-
geneous transport equations of the form
They all propagate along the rays. This justifies the name transport vector of T .
After (n + 1)steps one knows
. " , ,
as well as the form of Ucn),but not v,.
Note that it is the set (A&',A&'-,, ...,A:) of the principal parts of the terms of the
original operator D in (4.1) which determines the dispersion relation, the eiconal
equation and the ray-congruences along which the amplitude are transported. The
lower-order parts of D determine the coefficientsf, f,.,gj of the transport equations.
According to the construction of the series (4.3) one has
to 2 define a one-parameter family of initial data for (4.2).Let U(&)be the family of
(exact) solutions of (4.2) belonging to these initial data, and let
with zero initial data on 2 in terms of the magnitude of the source Q(s) gives an
estimate of the error Ecn)(s).If, as the form of (4.1) and simple examples suggest,
(4.25) implies
] W(e)] < e-p-]&(6)1-xO.const, (4.26)
where xO is the 'time distance' from the initial hypersurface xO = 0, then (4.24)
implies for a compact domain
E y e ) = 0(cn+l-P). (4.27)
Moreover, then
( U - U(n-p)l = ( U - U(n)+ O(&n+l-p)1 = O(&n+l-~
1. (4.28)
That is, if the operator D(s) is such that (4.25) implies (4.26),then
The authors are grateful for helpful discussions with the Relativity Group at
Munich, in particular with D. Christodoulou. This work was partly supported by
the Deutsche Forschungsgemeinschaft.
Wave propagation in plasmas in curved space-time. I 405
LINKED CITATIONS
- Page 1 of 1 -
This article references the following linked citations. If you are trying to access articles from an
off-campus location, you may be required to first logon via your library web site to access JSTOR. Please
visit your library's website or contact a librarian to learn about options for remote access to JSTOR.
References