a

r
X
i
v
:
0
8
0
9
.
1
5
7
1
v
1


[
g
r
-
q
c
]


9

S
e
p

2
0
0
8
Universal properties of the near-horizon optical geometry
G W Gibbons

& C M Warnick

DAMTP, University of Cambridge,

Wilberforce Road, Cambridge
CB3 0WA, UK
DAMTP-2008-80
Abstract
We make use of the fact that the optical geometry near a static non-degenerate Killing
horizon is asymptotically hyperbolic to investigate universal features of black hole physics.
We show how the Gauss-Bonnet theorem allows certain lensing scenarios to be ruled in or
out. We nd rates for the loss of scalar, vector and fermionic hair as objects fall quasi-
statically towards the horizon. In the process we nd the Lienard-Wiechert potential for
hyperbolic space and calculate the force between electrons mediated by neutrinos, extending
the at space result of Feinberg and Sucher. We use the enhanced conformal symmetry of the
Schwarzschild and Reissner-Nordstrom backgrounds to re-derive the electrostatic eld due to
a point charge in a simple fashion.
1 Introduction
There has been over the past few years a very large amount of theoretical work on black holes
addressing problems in quantum gravity, supergravity, string theory and M-theory. Typically one
seeks solutions of the supergravity equations in four or higher dimensions, and while many are
broadly similar to to the well known Kerr-Newman-de-Sitter family in four spacetime dimensions,
there are many dierences of detail and in higher dimensions qualitatively dierent features can
arise. It is desirable therefore to x upon universal properties, true for a broad class of black
holes. For this reason the near horizon geometry of extreme black holes has received a great deal
of attention, since it universally behaves like AdS
2
M
n2
, where M
n2
is typically an n 2
dimensional Einstein space and the symmetry is enhanced from R to SO(2, 1). By contrast the
universal near horizon optical geometry of non-extreme horizons with its enhanced conformal
symmetry has largely been ignored (notable exceptions are [1, 2]). In other words, little has been
done to exploit the fact that near a non-extreme horizon of a static black hole with metric
ds
2
= V
2
dt
2
+
ij
dx
i
dx
j
, (1.1)

[email protected]

[email protected]
1
i = 1, 2, . . . , n 1 the optical metric
a
ij
dx
i
dx
j
= V
2

ij
dx
i
dx
j
, (1.2)
becomes asymptotically hyperbolic, with a conformal boundary whose geometry is that of the
event horizon. In the spherically symmetric case, the limiting optical geometry is precisely that of
hyperbolic space H
n1
= SO(n 1, 1)/SO(n 1) with radius of curvature equal to
1
, where
is the surface gravity.
This is especially ironic because asymptotically hyperbolic geometry has been studied for some
time because of the light it throws on the no-hair properties of asymptotically de-Sitter metrics
and the freezing of perturbations which have crossed the horizon of an inationary universe (see
[3] for a recent discussion and references to earlier work). A much better known case arises in
the AdS/CFT correspondence where the asymptotically hyperbolic geometry of AdS
n
or, in its
Euclidean formulation, H
n
is of interest.
The aim of the present paper is to ll this gap by embarking on an exploration of what can be
learned about the universal qualitative properties of black holes from studying their near horizon
optical geometry using the tools of hyperbolic geometry. We shall principally be concerned with
the two topics
A qualitative study of of null geodesics near a static horizon using the Gauss-Bonnet theorem,
rather in the style of [4] in the case of cosmic strings.
A study of the shedding of hair near static event horizons using propagators in hyperbolic
space.
Of course, in the case of astrophysical black holes the near horizon geometry has long been
studied under the rubric of the Membrane Paradigm [5] and its Rindler like features have been
described. However this work, mainly concentrates on the planar approximation to the horizon
geometry and does not make use of detailed concepts and ideas of hyperbolic geometry. Closer
to what we are interested in is the work of Haba [2] which considers scalar elds near a Killing
horizon using an optical geometry approach and constructs approximate Greens functions in cases
where the horizon is not necessarily spherical. This approach is more in tune with our philosophy
of seeking universal properties. We will focus on spherical horizons and show that the enhanced
symmetry present in this case make approximate propagators much simpler to construct. We
remark that the universal nature of black hole absorption cross-section [6] has recently played an
important role in the understanding of the ratio of shear viscosity to entropy density of conformal
uid in the AdS / CFT correspondence [7].
The paper will be organised as follows: we rst dene the optical metric and explore some
of its properties, including a study of light rays near an event horizon using the Gauss-Bonnet
theorem. We will then present a general argument based on the near horizon limit of the optical
geometry to estimate the rate of loss of hair as bodies fall towards the black hole. Then we will
show how the optical metric allows one to nd the elds due to static electric and scalar charges
in the Schwarzschild and Reissner-Nordstrom backgrounds with very little calculation. This is
a re-derivation of results in the literature in a more coherent and direct way. We will include
equipotential plots for a charged particle approaching a black hole, graphically demonstrating the
no-hair result.
2
2 Optical Metrics
The optical metric may be thought of as the modern incarnation of an idea dating back to Fermat
in the 17th century. Fermat expressed the laws governing reection and refraction of light as what
we would now call an action principle. His principle of least time states that the path taken by
a ray of light is that which minimises the time taken between the two points. This can be used to
derive the more familiar Snells law and other optical laws.
In the case of light rays moving in a static background, with a given choice of time coordinate
t, we may take this at face value and dene the action for light rays in the metric
g = a
2
(x)dt
2
+h
ij
(x)dx
i
dx
j
(2.1)
to be
S =
_
dt =
_
_
a
2
h
ij
dx
i
d
dx
j
d
d, (2.2)
where we use the fact that null rays have ds = 0. Extremizing this action gives the unparameterised
geodesics of the 3 dimensional Riemannian metric:
h
opt.
= a
2
(x)h
ij
(x)dx
i
dx
j
. (2.3)
These unparameterised geodesics are the light rays and the metric h
opt.
is the optical metric. One
may check that these unparameterised geodesics indeed coincide with the projections of the null
geodesics of (2.1) onto the spacelike surfaces t = const. and so the light rays are the paths traced
by photons moving in this static space. The equivalence is clear by considering the metric
g
opt.
= a
2
g = dt
2
+h
opt.
; (2.4)
since the unparameterised null geodesics are conformally invariant objects, the result follows. We
will sometimes refer to the ultra-static metric g
opt.
as the optical metric also, relying on context to
distinguish it from h
opt.
. The optical metric is not necessarily unique as it depends upon a choice
of time coordinate t. For metrics which admit more than one choice of t there can be more than
one optical metric. We shall see this in detail in the case of anti-de Sitter space below.
It is not only statements about the null geodesics which are accessible via the optical metric.
Many of the eld equations of physics both classical and quantum behave well under conformal
transformations and so we can make use of the universal nature of the near horizon optical geometry
to study physics near a black hole (or cosmological) horizon.
2.1 The Optical Metrics of de Sitter and anti-de Sitter
2.1.1 de Sitter
We start with 3 + 1 dimensional de Sitter as the timelike hyperboloid in E
4,1
:
X
2
+Y
2
+Z
2
+W
2
V
2
= 1, ds
2
= dX
2
+dY
2
+dZ
2
+dW
2
dV
2
. (2.5)
We could consider n+1 dimensions, but the generalisations are straightforward. A choice of static
time coordinate t corresponds to a choice of future-directed, timelike, hypersurface orthogonal
3
Killing vector

t
. The Killing vectors of dS are those in E
4,1
which generate rotations and boosts.
A basis for the Killing vectors is given by the hypersurface orthogonal vectors:
M

= X

. (2.6)
Here , are E
4,1
indices. There is no Killing vector which is everywhere timelike, however the
Killing vector:
K = W

V
+V

W
(2.7)
is timelike and future directed in the region W
2
V
2
> 0 W > 0. Furthermore, any other
choice of timelike Killing vector is equivalent to K under a Lorentz transformation. We can nd
a parameterisation of the hyperboloid in this patch, such that K =

t
is a static Killing vector as
follows:
X = r sin sin ,
Y = r sin cos ,
Z = r cos ,
W =

1 r
2
cosh t,
V =

1 r
2
sinh t. (2.8)
On this patch, the metric takes the form
ds
2
= (1 r
2
)
_
dt
2
+
dr
2
(1 r
2
)
2
+
r
2
1 r
2
(d
2
+ sin
2
d
2
)
_
, (2.9)
so that the optical metric may be seen to be the Beltrami metric on Hyperbolic space. In fact
these coordinates cover all of the Beltrami ball and so the optical geometry of the static slicing of
de Sitter is precisely H
3
. The conformal innity of the hyperbolic ball corresponds to the Killing
horizon on the hyperboloid at W
2
V
2
= 0 where K becomes null. It is a general characteristic
of Killing horizons that the optical geometry approaches a constant negative curvature geometry
near the horizon.
2.1.2 Anti-de Sitter
The situation for AdS is somewhat more interesting than that for dS because there exist three
equivalence classes of timelike, future directed, hypersurface orthogonal Killing vectors under the
action of SO(3, 2). To see this we take AdS to be a hyperboloid in E
3,2
:
W
2
V
2
+X
2
+Y
2
+Z
2
= 1, ds
2
= dW
2
dV
2
+dX
2
+dY
2
+dZ
2
. (2.10)
In a similar way to the case of dS, a basis for the Killing vectors is given by:
M

= X

, (2.11)
4
where , are E
3,2
indices. Under a SO(3, 2) transformation, any Killing vector which is timelike
somewhere on the hyperboloid may be brought into one of three forms, listed below with the region
in which they are timelike:
K
1
= V

W
W

V
, all of AdS,
K
2
= (Z +W)

V
+V
_

Z


W
_
, W +Z > 0,
K
3
= Z

V
+V

Z
, Z
2
V
2
> 0 Z > 0.
(2.12)
We now nd the optical metric in each case:
Case 1 We pick the following parameterisation of the hyperboloid
X = r sin sin ,
Y = r sin cos ,
Z = r cos ,
W =

1 +r
2
cos t,
V =

1 +r
2
sin t, (2.13)
so that the metric is given by:
ds
2
= (1 +r
2
)
_
dt
2
+
dr
2
(1 +r
2
)
2
+
r
2
1 +r
2
(d
2
+ sin
2
d
2
)
_
, (2.14)
and we recognise that the optical metric is the Beltrami metric for S
3
. This covers one half of the
sphere, with the 2-sphere at r = corresponding to an equatorial 2-sphere.
Case 2 We pick a dierent parameterisation for the hyperboloid:
X = x/z,
Y = y/z,
Z = (1 + t
2
x
2
y
2
z
2
)/(2z),
W = (1 t
2
+x
2
+y
2
+z
2
)/(2z),
V = t/z, (2.15)
so that the metric is the Poincare upper half-space metric:
ds
2
=
1
z
2
_
dt
2
+dx
2
+dy
2
+dz
2
_
, z > 0 (2.16)
and the optical geometry is the half-space z > 0 in E
3
.
Case 3 Finally we consider the case where

t
= K
3
. A suitable parameterisation of the static
patch is provided by:
X = tan cos ,
Y = tan sin ,
Z = cosh t cosechsec ,
W = cotanhsec ,
V = sinh t cosechsec . (2.17)
5
The spatial coordinates have ranges 0 < 2, 0 < /2, 0 < . The metric is given by:
ds
2
= cosech
2
sec
2

_
dt
2
+d
2
+ sinh
2
(d
2
+ sin
2
d
2
)
_
. (2.18)
We see that the optical metric is that of hyperbolic space in geodesic polar coordinates. Since
does not range over [0, ) the coordinates only cover half of H
3
and there is a boundary which is
given by the plane = /2 in these coordinates.
We note that all three of the AdS optical metric have a nite boundary. This boundary
corresponds to the conformal innity of the AdS space and is a manifestation of the fact that
AdS is not globally hyperbolic. In the case of dS, the optical metric is complete and has an
asymptotically hyperbolic end which corresponds to the Killing horizon of the static patch. As
we will see below, this behaviour is typical of a Killing horizon, such as the event horizons of
Schwarzschild and Reissner-Nordstrom.
2.2 The Optical Metric of Schwarzschild and Reissner-Nordstrom
Owing to the similarities between the cosmological horizon of de Sitter and the event horizons of
black holes, we might expect the optical geometries to be similar near the horizon. We shall see
that this is indeed the case, and that near the horizon, the geometry of a static black hole has an
asymptotically hyperbolic optical metric.
We start with the Schwarzschild metric
ds
2
=
_
1
2M
r
_
dt
2
+
dr
2
1
2M
r
+r
2
(d
2
+ sin
2
d
2
) (2.19)
and make the coordinate transformation
r = M
_
1 +

_
(2.20)
this takes the asymptotically at end to = 0 and the horizon to = 1 and puts the metric into
the form:
ds
2
=
_
1
1 +
_
_
dt
2
+ 16M
2
_
1 +
2
_
4
_
d
2
(1
2
)
2
+

2
1
2
_
d
2
+ sin
2
d
2
_
_
_
(2.21)
The term inside the braces may be seen to be the metric on H
3
in Beltrami coordinates. In the
limit 1, we thus see that the optical metric tends to a metric of constant negative curvature
as we approach the horizon.
The case of Reissner-Nordstrom is rather similar, although the resulting metric is not so elegant.
In the familiar coordinates, the Reissner-Nordstrom metric is given by
ds
2
=
_
1
2M
r
+
Q
2
r
2
_
dt
2
+
dr
2
1
2M
r
+
Q
2
r
2
+r
2
(d
2
+ sin
2
d
2
) (2.22)
6
In these coordinates, the horizon is at r = M +
_
M
2
Q
2
= M + , where we dene a new
parameter which we will assume to be strictly positive. The case = 0 corresponds to an
extremal black hole which we will not consider here. The coordinate transformation
r = M +

(2.23)
puts the metric into the form
ds
2
=

2
(1
2
)
( +m)
2
_
dt
2
+
( +m)
4

2
_
+m
( +m)
_
4
_
d
2
(1
2
)
2
+

2
1
2
_
d
2
+ sin
2
d
2
_
_
_
(2.24)
We see once again that the optical metric approaches the Beltrami metric on hyperbolic space as
we get close to the horizon. In both cases, the radius of the hyperbolic space is
H
, the inverse
Hawking temperature of the black hole. In fact, this is a general property of a static metric with
a non-degenerate Killing horizon as shown in [1]. In that paper Sachs and Solodukhin show that
near a non-extreme horizon of a static black hole with metric
ds
2
= V
2
dt
2
+
ij
dx
i
dx
j
, (2.25)
i = 1, 2, . . . , n 1 the optical metric
a
ij
dx
i
dx
j
= V
2

ij
dx
i
dx
j
, (2.26)
becomes asymptotically hyperbolic, with a conformal boundary whose geometry is that of the
event horizon. Essentially this is due to the fact that at a non degenerate Killing horizon V
2
must
have a simple zero. We will consider the spherically symmetric case from here on, however we
will try and identify results which we expect to remain the same in the case of more interesting
(compact) horizon topology.
It will prove crucial in our exact calculations later that the optical metrics of both Schwarzschild
and Reissner-Nordstrom take the form:
g
opt.
= dt
2
+H
4
h, (2.27)
where h is the metric on the unit pseudo-sphere, H
3
, and H is a harmonic function on H
3
which
approaches 1 near the conformal boundary of H
3
. This observation is responsible for the fact that
the elds due to static electric and scalar charges in these backgrounds may be found explicitly
[8, 9, 10]. We show below how these elds may be constructed. This special form of the metric
occurs only in the 4-dimensional space-times, so does not, unfortunately, lead to a generalisation
of these results in an obvious way to higher dimensions.
For another viewpoint on the optical geometry of Schwarzschild see [11] where the geometry is
constructed as an embedding in a higher dimensional hyperbolic space.
3 Lensing and The Gauss-Bonnet theorem
In order to discuss null geodesics we could follow the well trodden path of solving the dierential
equations. Instead we will follow the approach of [4, 12] and extract information about geodesics
7
using the Gauss-Bonnet theorem which directly involves the negative curvature of the optical met-
ric. Although here we consider only the Schwarzschild black hole, it is clear that many qualitative
properties may be deduced using only the assumption of negative curvature near the horizon of
the optical metric, provided a totally geodesic 2-surface exists.
Let us now consider geodesics lying in an oriented two-surface . We may apply the Gauss-
Bonnet theorem to obtain useful information [4], in particular about angle sums of geodesic trian-
gles. Let D be domain with Euler number (D) and a not necessarily connected boundary
D, possibly with corners at which the tangent vector of the boundary is discontinuous. If K is
the Gauss curvature of D, such that R
ijkl
= K(f
ik
f
jl
f
il
f
jk
) and k the curvature of D,
i
the
angle through which the tangent turns inwards at the ith corner then
_
D
KdA+
_
D
kdl +

i
= 2(D) . (3.1)
In the case of the Schwarzschild metric, if one considers geodesics in an equatorial plane the
optical metric is
ds
2
=
dr
2
(1
2M
r
)
2
+
r
2
(1
2M
r
)
d
2
. (3.2)
We return here to the standard Schwarzschild coordinates of (2.19). Note that the radial optical
distance is
dr
(1
2M
r
)
= dr

, (3.3)
where r

= r 2M + 2M ln(
r
2M
1) is the Regge-Wheeler tortoise coordinate.
There is a circular geodesic at r = 3M and the horizon r = 2M is at an innite optical distance
inside this at r

= . The Gauss curvature

K =
2M
r
3
(1
3M
2r
) (3.4)
is everywhere negative. It falls to zero like
2M
r
3
at innity but near the horizon the Gauss-curvature
approaches the negative constant
1
(4M)
2
. This is precisely as we expect to nd given the results
of the previous section.
The fact that the Gauss curvature is negative looks on the face of it rather paradoxical, since
one usually thinks of gravitational elds as focussing a bundle of light rays. However, as Lodge
perhaps dimly realised [13] a spherical vacuum gravitational eld does not quite act in that way.
The equation of geodesic deviation governing the separation of two neighbouring light rays in
the equatorial plane is
d
2

dt
2
+K = 0 . (3.5)
Thus neighbouring light rays actually diverge. The focussing eect of a gravitational lens is not,
as we shall see shortly, a local but rather a global, indeed even topological, eect.
One might wonder whether the full 3-dimensional curvature
3
R
ijkl
of the optical metric has all
of its sectional curvatures negative, but this cannot be. The sectional curvature of a surface is
related to the full curvature tensor by
3
R
ijkl
= K(f
ik
f
jl
f
1l
f
jk
) K
ik
K
jl
+K
il
g
jk
, (3.6)
8
(iii)

(i)

(ii)

r = 3m
r = 3m
r = 3m
r = 3m

(iv)
(v)
(vi)

Figure 1: The geodesic polygons described in (i)-(vi)

where K
ij
is the second fundamental form or extrinsic curvature of the surface. For a totally
geodesic surface K
ij
= 0, and the two sectional curvatures agree. One such totally geodesic
surface is the equatorial plane for which, as we have seen, K is negative. Another totally geodesic
submanifold is the sphere at r = 3M for which K is obviously positive.
The negativity of the Gauss curvature of the optical metric in the equatorial plane is a fairly
universal property of black hole metrics. To see this we note that if
ds
2
= d
2
+l
2
()d
2
, (3.7)
then
K =
1
l
d
2
l
d
2
. (3.8)
Any metric with the same qualitative features as the Schwarzschild metric, as long as it has
a positive mass, will have K negative. Indeed this fact might be made the basis of excluding
negative mass objects observationally.
A simple calculation shows the integral over the region outside the circular geodesic at r = 3M
is _
r3M
KdA = 2 . (3.9)
Let us now apply the Gauss-Bonnet theorem to various cases.
(i) Geodesic triangle not containing the the region inside r = 3M. In this case () = 1. If
, , are the necessarily positive internal angles, we nd that the angle sum is less that ,
+ + = +
_

KdA < . (3.10)

(ii) Geodesic di-gon S not containing the the region inside r = 3M. In this case (S) = 1. If
and are the internal angles,
+ =
_
S
KdA < 0 . (3.11)
9
In other words two such geodesics cannot intersect twice if the hole is not inside the di-gon. Neither,
in these circumstances, can a geodesic intersect itself because
(iii) Geodesic loop T not containing the the region inside r = 3M. In this case (T) = 1 and one
nds that if if is the internal angles,the
= +
_
T
KdA < , (3.12)
which is plainly impossible.
This might seem counter-intuitive in the light of ones usual intuition about light bending, but
this feeling is dispelled by considering cases in which the domain D has two boundary components,
the second, inner, one being the circular geodesic at r = 3M. The domain with the circle removed
has the topology of an annulus and thus its Euler number vanishes.
(iv) Geodesic triangle with hole
o
enclosing the geodesic circle at r = 3M and with the region
the region inside r = 3M removed.
If , , are the internal angles, we nd that the angle sum is greater than ,
+ + = 3 +
_

0
KdA . (3.13)
Similarly
(v) Geodesic di-gon S
0
with the the region inside r = 3M removed. In this case (S
0
) = 0 and
one nds that if If and are the internal angles, then
+ = 2 +
_
S
0
KdA > 0 . (3.14)
In other words two such geodesics may intersect twice if the hole is inside the di-gon. Moreover,
in these circumstances, a geodesic can intersect itself because:
(vi) Geodesic loop T
0
containing the the region inside r = 3M. In this case (T
0
) = 1 and if If
is the internal angle, we nd that
= +
_
T
0
KdA, (3.15)
which is plainly possible.
Similar results may be obtained by considering geodesics inside r = 3M, but now domain must
not contain the horizon, otherwise
_
D
KdA will diverge. Near the horizon the geometry is that of
Lobachevsky space with constant curvature
1
4M
.
(vii) Deection. We consider a geodesic line with no self-intersection which at large distances, is
radial. The angle between the asymptotes is , with the convention that it is positive if the light
ray is bent towards the hole. The geodesic decomposes the region inside two circles, one of very
large radius and the other at r = 3M into two domains D

whose common boundary component

consists of the geodesic, which intersects the circle at innity at right angles. We chose D
+
to
enclose the hole so it has an inner boundary component at r = 3M and a portion of the circle at
innity through which the angle has range . Clearly D
+
is topologically an annulus and so
10
D
+ D

r = 3m
Figure 2: Light bending by a Schwarzschild black-hole
it has vanishing Euler number, (D
+
) = 0. The other domain has Euler number (D

) = 1, and
ranges through + . The Gauss-Bonnet formula applied to D

acquires a contribution from

the two corners and the circle at innity. The result is
=
_
D

KdA > 0. (3.16)

For a geodesic whose distance of closest approach is very large, we may estimate this integral
by approximating the geodesic as the straight line r =
b
sin
. The impact parameter to this lowest
non-trivial order coincides with the distance of nearest approach and equals b. To the necessary
accuracy
KdA
2M
r
3
rdrd . (3.17)
The domain of integration D

is, with sucient accuracy over r

b
sin
, 0 2. A simple
calculation gives the classic result
=
4M
b
. (3.18)
Note the same method works for any static metric, not just Schwarzschild (for an application
to gravitational lensing see [12]) and shows that the Gauss-Bonnet method does not just give
qualitative results, but it can be made into a quantitative tool.
4 No-hair properties from the optical metric
We now shift our attention to a dierent aspect of black hole physics, the so called no-hair
property. A stationary black hole has only three measurable quantities associated with it: mass,
angular momentum and electric charge. Thus no matter how complicated a system we start with,
once it has undergone gravitational collapse to form a black hole, we are left with only these
three pieces of information. This presents only a minor philosophical problem if we are prepared
to accept that the information contained in the initial system is somehow trapped irretrievably
behind the horizon. Once one includes Hawkings observation that black holes may radiate and
indeed evaporate over time, the question of where the information goes becomes more vexed, giving
rise to the so called information loss paradox.
11
We shall be interested with discovering how, at the classical level, information is lost as a body
falls into a black hole. We will work with an approximation where the in-falling body is supposed
to have a negligible eect on the background and so we may consider physics in a xed black-hole
geometry. This amounts to a linearisation of the problem, but allows analytic progress to be made.
At the linearised level, the no-hair property may be translated mathematically into the notion
that the black hole exterior cannot support any external elds which are regular both at the horizon
and spacelike innity. An interesting question is what happens to the elds around some compact
body as it falls into the black hole. This corresponds to asking what happens to a propagator as
its pole approaches the horizon. This gives information both about the classical scenario, but also
about the outcome of scattering experiments performed as the body falls towards the hole [14].
We are interested in nding a fairly general approach to study how hair is lost as bodies
carrying charges fall into a black hole. We will argue that this is a property of the geometry close
to the black hole horizon, and so we can consider the problem in this region where the geometry
simplies. We translate the question of nding propagators to a problem in the optical metric and
then show how we can estimate the rate of information loss in this geometry. As we noted above,
near the horizon this geometry approaches that of hyperbolic space irrespective of the details of
the black hole under consideration.
4.1 Physics in R
t
H
3
There have been investigations of physics in spaces of constant negative curvature for some time
and with varying motivations. Callan and Wilczek initiated a study of quantum mechanics on H
4
in [15] in order to geometrically regulate the infra-red divergences of Euclidean eld theory. In [16]
Atiyah and Sutclie considered Skyrmions in H
3
as a means of nding approximate Skyrmions in
E
3
for the case where the pion mass is non-zero. Field theories on H
3
are also thermodynamically
interesting as one might expect, anticipating the Hawking radiation of horizons. A study of some
thermodynamic properties, especially Bose-Einstein condensation is shown in [17]. There have also
been studies of electrostatics and magnetostatics in hyperbolic space, with particular reference to
the Gauss linking formula [18].
Although the references above provide a reasonably comprehensive discussion of physics in
R
t
H
3
, in the interests of a self-contained exposition we will discuss some aspects here. Using
the near horizon limit of the optical metric, this corresponds after a conformal transformation to
physics in the neighbourhood of a non-extremal black hole horizon. The fact that analytic progress
is possible may be traced to the fact that the metric is conformally equivalent to the inside of the
future light-cone of the origin in Minkowski space as follows.
Throughout this section we represent a point in H
3
as a point on the unit pseudosphere H
3
=
X X = 1, X
0
> 0 in E
3,1
. This makes the equations manifestly SO(3, 1) invariant and easy to
translate between dierent coordinate systems. The point (t, X) in R
t
H
3
is mapped to a point
on the interior of the future light cone of the origin in E
3,1
according to:
: R
t
H
3
x E
3,1
, x x = 1, x
0
> 0
(t, X) x = Xe
t
. (4.1)
If g is the metric on R
t
H
3
and is the standard metric on E
3,1
, then one nds that:

= e
2t
g, (4.2)
12
so we have exhibited the conformal equivalence of these two spaces. This means that given any
conformally invariant equation whose propagator may be found in Minkowski space, one may nd
the propagator for R
t
H
3
.
4.1.1 Massless Wave Equation
Although the massless wave equation is conformally invariant, and so the propagator may be
constructed from the known at space propagator, it is more convenient to directly solve in this
case. We seek to solve the equation:
(
g

1
6
R
g
)G(t, X; , Y ) =
_

2
t
+
h
+ 1
_
G =
(4)
g
((t, X), (, Y )). (4.3)
One might think that the appearance of the curvature term above gives rise to an eective mass,
however it is important to include this term in the massless wave equation to ensure, for example,
that disturbances propagate along the light-cone as one would expect. Following standard treat-
ments, one Fourier transforms in time and takes = 0 without loss of generality. We then need
to solve the Helmholtz equation
_

h
+ 1 +k
2
_

G =
h
(X, Y ). (4.4)
This has the general solution, found by using geodesic polar coordinates on H
3
:

G(X, Y ) =
Ae
ik
+Be
ik
4 sinh
(4.5)
where = D(X, Y ) and A + B = 1. The fact that this (and other Greens functions on H
3
)
depends only on D(X, Y ) is due to the 2-point homogeneity of the space. Undoing the Fourier
transform one nds:
G(t, X; , Y ) = A
(t D(X, Y ))
sinh D(X, Y )
+B
(t +D(X, Y ))
sinh D(X, Y )
(4.6)
The choices for A and B determine what combination of the advanced and retarded propagator
we have. Note that this propagator is periodic with period 2i in the time coordinate.
4.1.2 Lienard-Wiechert Potential
The Lienard-Wiechert Potential describes the electromagnetic eld due to a charge q moving in
Minkowski space along some path r(s) E
3,1
where s is any parameter. The potential at a point
x is constructed as follows: rst nd a solution s
r
to the equation:
(x r(s)) (x r(s)) = 0 (4.7)
which should correspond to the intersection of the path of the charge with the past light cone for
a retarded propagator. The Maxwell eld is then determined by the one-form
A =
q
4
0
r dx
(x r) r

s=sr
. (4.8)
13
Note that this is invariant under re-parameterisations of the path of the particle r(s). In calculating
F = dA one should be wary since A depends on x both explicitly and also implicitly through s
r
.
Dierentiating (4.7) one nds that:
ds
r
=
(x r) dx
(x r) r
. (4.9)
The standard calculations may now be performed and the Maxwell eld calculated.
In order to nd the eld due to a point charge q moving along the curve (s, R(s)) R
t
H
3
we will use the conformal invariance of the Maxwell equations. Using the conformal map we
may pull back the Lienard-Wiechert potential from Minkowski space. One nds that the lightcone
condition may be re-written:
t s
r
= D(X, R(s
r
)) (4.10)
and the potential is given by:
A =
q
4
0
1

R X
_
(R X)
2
1
_
(R X +

R X)dt +R dX +

R dX
_

s=sr
. (4.11)
Once again the dependence on (X, t) is subtle, but one may calculate the Maxwell eld by using:
ds
r
=
_
(X R)
2
1dt +R dX
_
(X R)
2
1

R X
, (4.12)
which follows from dierentiating (4.10).
Calculating the eld strength in the limit where the source charge is at a large distance from
the observer but

R

R and

R

R remain bounded, we nd that the eld decays like e

with the
separation of charge and observer.
We may interpret this in terms of black hole optical geometry which approaches H
3
R
t
near
the horizon. In this case, as we shall see later one must add a static spherically symmetric eld
to enforce the condition that the black hole is uncharged. We have shown that even including the
corrections to the electromagnetic eld due to the motion of the charge, the eld due to a particle
falling into a black hole tends to a monopole charge as the particle approaches the horizon.
4.1.3 Spinors on hyperbolic space
Having dealt with spin 0 and spin 1 elds on R
t
H
3
, the logical next step is to discuss spinors and
the Dirac operator on this space. Following Dirac [19], we will write the Dirac equation in terms
of objects in the embedding space, E
3,1
as this will allow us to maintain SO(3, 1) covariance.
We will make use of the following observation: the Dirac algebra for E
3,1
may be represented
in the form:

0
= i
3
I
2
,

i
=
2

i
, (4.13)
where
i
are the Pauli matrices which form a 2-component representation of the Dirac algebra
for E
3
. We could, if we so chose, construct this 2-component representation by considering Weyl
14
spinors,

on an auxiliary E
3,1
restricted to a constant time hyperplane, . These 2-spinors
would then transform under Poincare transformations of this auxiliary E
3,1
xing the hyperplane
and would have a natural L
2
inner product respecting these transformations given by
(
1
,
2
) =
_

1
T
2
=
_

2
. (4.14)
We use the notation

= (I
2
,
i
),

= (I
2
,
i
). A Dirac spinor on the original E
3,1
space is then
the product of two 2-component spinors, the rst transforming under SO(1, 1) which generates
the boosts of the Lorentz group and the second under the rotational SO(3). In fact the second
spinors transform under the whole E(3) symmetry of E
3
but the translations act by the identity.
The reason we take this somewhat circuitous approach to constructing the Dirac spinors is that it
will allow us to construct spinors for R
t
H
3
respecting the SO(3, 1) invariance of H
3
.
We will now make use of the fact that the metric in the forward light cone of the origin of
Minkowski space may be written in the form
ds
2
= dt
2
+t
2
h, (4.15)
with h the metric on H
3
. This is sometimes referred to as the Milne universe. This Minkowski
space will play the role of the auxiliary space above. Using the standard approach to construct
the Dirac operator from the spin connection, one nds that acting on Weyl spinors,
,D =
0
_

t
+
3
2t
_
+
1
t
,D
(h)
. (4.16)
Now any vector in the forward light cone of the origin may be written
W = tX, with X X = 1, t > 0, X
0
> 0 (4.17)
we may re-write the Dirac operator as
,D =

W
= X
_
X

W
+
3
2
1
[W[
_
+
1
[W[
_
[W[

W
XX

W

3
2
X
_
, (4.18)
comparing this with (4.16) we conclude that
,D
(h)
= XX
3
2
X,
= X
_
1
2

(X

)
3
2
_
. (4.19)
It may be checked that X anti-commutes with the RHS of this expression, so it is in fact
convenient to take
,D
(h)
=
1
2

(X

)
3
2
= M
3
2
, (4.20)
15
the operator M is that introduced by Dirac in [19]. The relation to the standard construction for
Dirac operators in a curved space is developed in [20] Following the discussion above, there is a
natural L
2
inner product on the space of Weyl spinors on H
3
given by
(
1
,
2
)
H
3 =
_
H
3
[X]
1
X
2
(4.21)
Here [X] is the Riemannian volume form of H
3
. This inner product is positive denite, because
X X = 1 and X
0
> 0. Importantly, it is also SO(3, 1) invariant by construction and so respects
all of the symmetries of hyperbolic space.
We can exhibit a set of plane wave eigenfunctions of the Dirac operator by considering the
analogous plane wave eigenfunctions for the Laplace operator on H
3
exhibited by Moschella and
Schaeer [21]. Let be a constant 2-spinor (in the sense that

= 0), then the spinor

(X) =

(2)
3
2
( X)

3
2
i
(4.22)
satises
,D
(h)

(X) = i

(X). (4.23)
Furthermore these functions tend pointwise to the standard plane wave basis for eigenfunctions
on E
3
as the radius of curvature of H
3
tends to innity. Obviously and dene the same
function, up to scale, for any C

so that the space of eigenfunctions is R

+
CP
1
. Using the
results of Moschella and Schaeer it should be possible to establish the following normalisation
and completeness results
(

)
H
3
= (

)(

) (4.24)
and
_
CP
1
[]
_

0
d

(X)

(X

) X

=
H
3(X, X

)I
2
(4.25)
however this has so far proved dicult. We shall proceed therefore on the assumption that this is
the case. For a discussion of integration over CP
1
and the measure [], see the appendix. One
may at least show that the second result is true as the hyperbolic radius tends to innity.
We can now construct Dirac spinors on R
t
H
3
by taking the tensor product of a Weyl spinor
on H
3
with an SO(1, 1) spinor. The Dirac operator is given by:
,D = i
3
I
2

t
+
2

_
M
3
2
_
. (4.26)
The Dirac conjugate is given by
=

i
3
X (4.27)
we note that
0
= i
3
I
2
so the Dirac conjugate does not take its standard form. This is related
to the fact that we chose to make the Dirac operator on H
3
more symmetric by multiplying by
X. We note nally that we may identify

5
=
1
I
2
(4.28)
16
as the chirality matrix which satises
_

5
, ,D
_
= 0 and
_

5
_
2
= I
4
. (4.29)
We may nally construct a complete set of plane wave solutions to the Dirac equation
,D = 0 (4.30)
as follows:

0
sz
(X, t) =

(2)
3
2
e
sit

(
z
X
z
)

3
2
+si

z
(4.31)
where

=
1

2
_
1

_
, and
z
=
1
_
1 +[z[
2
_
1
z
_
. (4.32)
We have introduced the quantum numbers s = which distinguishes the positive and negative
energy solutions, z C which parameterizes CP
1
under the usual stereographic projection and
= , the chirality. With this choice of parameterisation of CP
1
the appropriate measure in the
completeness relation (4.25) is
[] =
2idzd z
_
1 +[z[
2
_
2
(4.33)
which we recognise as the measure on S
2
under stereographic projection.
We will now use these results to calculate the force between electrons (or indeed other leptons)
mediated by neutrino exchange.
4.1.4 Neutrino mediated forces
In at space there is a long range lepton-lepton force mediated by the exchange of a pair of
neutrinos. The potential, as shown by Feinberg and Sucher [22], is
V (r) =
G
2
W
4
3
r
5
, (4.34)
where G
W
is the weak-interaction coupling constant. Their calculation was based on calculating the
one-loop scattering of one electron by another mediated by a pair. A simpler means of nding
this potential, as described by Hartle [23], is to treat the neutrino eld as quantum mechanical
and the electrons as classical both in their role as a source for the neutrino eld and as particles
acted on by that eld. Hartle shows that in this limit, the neutrino eld obeys the modied Dirac
equation
_
i,D
G
W

2
N(1 +
5
)
_
(x) = 0 (4.35)
where N

is the classical electron number current. The equation of motion of an electron in the
classical limit is
m
Du

d
=
G
W

2
u

) (4.36)
17
where u

is the electrons four-velocity and the potential B

is given in terms of the neutrino eld

by:
B

(x) =

(x)

(1 +
5
)(x)
_

0
(x)

(1 +
5
)
0
(x)
_
. (4.37)
(x) is the neutrino eld with the weak interactions turned on and
0
(x) is the same eld with
the interactions turned o. Both expectation values are taken in vacuum with no free neutrinos.
The normalisation of the neutrino eld is xed by the canonical anti-commutation relations which
relate the anti-commutators of elds on a spacelike hypersurface . The only non-vanishing bracket
is
_
(x),

(x

) T
_
=

(x, x

) (4.38)
where T is the future directed unit normal to . Note that the standard relation would be
between and

, however as noted above, we have chosen to make the Dirac operator simpler
at the expense of taking a non-standard Dirac conjugate. For at space sliced along constant t
hyperplanes, (4.38) reduces to the standard relation.
Let us suppose that there is a complete set of solutions to the modied Dirac equation (4.35)
with the same quantum numbers as for the source free Dirac equation, so that we may write

sz
(t, X) = e
sit

sz
(X) (4.39)
and we will assume the completeness relation

s
_
C
2idzd z
_
1 +[z[
2
_
2
_

0
d
sz
(X)
sz
(X

) X

=
H
3(X, X

)I
2
(4.40)
(note that we dont sum over here). We may therefore expand the neutrino eld in the form
(t, X) =

_
C
2idzd z
_
1 +[z[
2
_
2
_

0
d
_
e
it

+z
b
z
+e
it

z
d

z
_
. (4.41)
The canonical anti-commutation relations for the neutrino eld imply the following non-vanishing
relations for the creation operators b and d
_
b
z
, b

_
=
_
d
z
, d

_
=
_
1 +[z[
2
_
2
2i
(z z

)(

(4.42)
while all other brackets vanish. We suppose the existence of a vacuum state [0) such that b
z
[0) =
d
z
[0) = 0. Using the anti-commutation relations we nd that the electric part of the neutrino
mediated vector potential, V = B
0
takes the form
B
0
(X) = 2
_
C
2idzd z
_
1 +[z[
2
_
2
_

0
d
_

z+
(X) X
z+
(X)

0
z+
(X) X
0
z+
(X)
_
. (4.43)
The fact that the coupling has a 1+
5
factor ensures that only positive chirality modes contribute.
18
As we are only interested in eects at the lowest order in G
W
, we may consider an expansion
of the spinors in terms of G
W
. Expanding to rst order

sz
=
0
sz
+G
W

1
sz
(4.44)
we nd that provided we assume that N
t
=
H
3(X, X

) is the only non-zero component of the

electron current, the modied Dirac equation (4.35) implies that
_
M
3
2
+is
_

0
sz
= 0
_
M
3
2
+is
_

1
sz
= i

2
H
3(X, X

)
0
sz
. (4.45)
We know from the previous section that the normalised zeroth order spinors take the form

0
sz
=

(2)
3
2
(
z
X
z
)

3
2
+is

z
. (4.46)
In order to solve the second equation we note that
M
2
2M =
2
H
3 , (4.47)
where the Laplacian here is the scalar Laplacian acting componentwise to the right. This may be
veried by following the argument of Dirac [19] with the appropriate signature and dimension. We
see that
_
M
3
2
+is
_ _
M
1
2
is
_
=
_

2
H
3 + 1 +
_
+
is
2
_
2
_
I
2
(4.48)
we recognise the right hand side of the equation as the conformal wave equation with a complex
wavenumber, for which we have already found the Greens function. Thus we may solve the second
of equations (4.45) by

1
sz
(X) =
__
M
1
2
is
_
(X, X

0
sz
(X

), (4.49)
where
(X, X

) =
i

2
4
e
is
e
/2
sinh
, with cosh = X X

. (4.50)
There is a choice of sign here corresponding to picking the retarded propagator. We note that after
Fourier transforming back to the time domain the propagator will be anti-periodic in time with
period 2i. We also note that there appears to be a breaking of the symmetry one might expect
under . This chiral symmetry breaking is a subtle consequence of the negative curvature
and is discussed in [15, 24].
Putting this all together, we nd
B
0
(X)
G
W
= 4Re
_
C
2idzd z
_
1 +[z[
2
_
2
_

0
d
0
z+
(X) X
_
M
1
2
+i
_
()
0
z+
(X

). (4.51)
19
Inserting (4.46) for
0
, the integrand reduces to the form

2
(2)
3
_
(
z
X
z
)

3
2
+i
(
z
X

z
)

1
2
i

()
sinh
+(
z
X
z
)

1
2
+i
(
z
X

z
)

3
2
i
_
(i
1
2
)() coth

()
_
_
. (4.52)
We will rst perform the integrals over CP
1
which are of the form:
I
a
=
_
C
2idzd z
_
1 +[z[
2
_
2
(
z
X
z
)
a
(
z
X

z
)
a2
. (4.53)
One may verify that
(
z

z
) = (1, n
z
) , (4.54)
where n
z
is the pull back of z to the unit sphere in R
3
under the standard stereographic projection
map. I
a
is Lorentz invariant (see Appendix), so we may assume without loss of generality that
X = (cosh , 0, 0, sinh ),
X

= (1, 0, 0, 0), (4.55)

and we may integrate over S
2
using standard spherical polar coordinates so that
I
a
=
_
sin dd
(cosh + cos sinh )
a
=
4 sinh(1 a)
(1 a) sinh
. (4.56)
Note that this is symmetric under a 2 a as it must be since we could have chosen X and X

the other way around. Integrating (4.52) over CP

1
then, we have after some simplication
i

2
2
8
3
_
1
sinh
2

+
2
(1 + 4
2
) sinh
4

_
+

2
2
8
3
(1 + 4
2
)
e
2i
sinh
4

(2 sinh i cosh ) , (4.57)

so that
B
0
()
G
W
=
4

2
4
3
Re
_

0
d
_

2
1 + 4
2
e
2i
sinh
4

(2 sinh i cosh )
_
. (4.58)
This integral is manifestly divergent for large , however the potential we are interested in is a low
energy eect, so we may introduce a large momentum cut-o by sending i which will
make the integral converge for large and taking the 0 limit after calculating the integral.
Doing this, we nd that the integral may be performed exactly and we nd that the neutrino eld
gives rise to an eective potential:
V () =
G
W

2
B
t
() =
G
2
W
8
3

2
sinh
4

_
cosh + sinh
2
Shi
_
, (4.59)
where Shi is the sinh integral:
Shi =
_

0
sinh t
t
dt . (4.60)
20
This equation is valid for any large with respect to the length scale dened by G
W
. We may
take the limit where the hyperbolic radius of the space tends to innity and we nd that
V ()
G
2
W
4
3

5
(4.61)
which is the result of Feinberg and Sucher for at space.
One might be concerned by the fact that there appears to be an asymmetry between the right
and left handed neutrinos implicit in (4.50) however it is possible to perform the same calculation
under the assumption that left handed neutrinos couple to electrons and the answer found is
precisely the same.
4.2 Thermodynamics
We noted above that the scalar propagator was periodic and the fermion propagator anti-periodic
in imaginary time. Reinserting dimensions, the period is given by 2iR where R is the radius of
the hyperbolic space. One expects that thermal propagators for a eld at temperature T should
have an imaginary period equal to 1/T, the inverse of the temperature. As we remarked above,
for the near horizon optical geometry of a horizon with surface gravity , R =
1
so we nd that
elds in the neighbourhood of a horizon are thermalized at a temperature
T
H
=

2
, (4.62)
precisely the Hawking temperature of the horizon. Notice that we nowhere had to Euclideanize
the time direction of the manifold in order to derive this result.
4.3 Approximate calculations near the horizon
Let us consider trying to construct the propagator for some physical eld in a black hole background
in the limit where the pole of the propagator approaches the horizon. We assume that since we
are considering a small perturbation to the background that the equations are linear and moreover
that they may be converted by a conformal transformation to equations with respect to the optical
metric. Further, since the black-hole background is assumed to be static, we may Fourier transform
with respect to t so that the equations may be expressed in terms of the optical geometry of
t = const. slices. We divide these slices into two regions:
Region I is a region surrounding the horizon, such that in this region the optical metric has
constant negative curvature to order .
Region II is the complement of region I and contains the asymptotically at end.
Region I may be thought of as the exterior of a ball in H
3
in the case where the horizon has
spherical topology. One would expect that if the topology diers for that of the sphere, then this
will not have an eect on the propagator in the limit where the pole approaches the horizon, so
we assume that Region I is indeed of this form. The conformal innity of the hyperbolic space
corresponds to the horizon of the black hole. We wish to solve the problem:
L(x) = L
hopt.
(x) +L

(x) = (x, x
0
) , (4.63)
21
where x
0
is close to the horizon and we have split the linear operator L into a geometric operator
constructed from h
opt.
and another part L

which is assumed to be small in Region I. This may

require shrinking Region I. We do not assume here that is a scalar the same considerations
will apply for elds of any spin.
In region I, the problem simplies to nding a propagator in hyperbolic space. This is a
simplication because H
3
is maximally symmetric so one may make use of isometries to move the
pole of the propagator around. We dene G
I
(x, x
0
) to satisfy the equation on hyperbolic space:
L
h
G
I
(x, x
0
) = (x, x
0
) (4.64)
subject to suitable boundary conditions as x approaches conformal innity (i.e. the horizon).
In region II, we are solving the homogeneous problem
L
hopt.
G
II
(x, x
0
) + L

G
II
(x, x
0
) = 0 (4.65)
such that G
II
(x, x
0
) agrees with G
I
(x, x
0
) at the boundary between regions I and II and decays
suitably as x approaches the asymptotically at innity. The approximate propagator we construct
is then given by:
G(x, x
0
) =
_

_
G
I
(x, x
0
) +K(x) if x in Region I
G
II
(x, X
0
) +K(x) if x in Region II
, (4.66)
K here is any solution of the homogeneous problem on the whole of the exterior of the black hole
which satises appropriate boundary conditions both at spacelike innity and at the horizon. It
is these solutions which carry any hair which the black hole may have. The charges carried by
the black hole as a result of K do not follow from regularity at innity or the horizon but must be
determined by, for example, integral conservation laws.
We are now ready to describe the limit as the pole of the propagator approaches the horizon.
In Region I we see this as the pole of a propagator in H
3
moving towards conformal innity. As the
boundary of Region I is a xed compact surface in H
3
, the elds on the boundary decay. Typically
this decay is exponentially quickly in the hyperbolic distance of the pole from some xed point.
Thus G
II
will also decay at this rate, by linearity. Thus only K can remain in the limit as the pole
approaches the horizon, with all other terms decaying. If the black hole cannot support a regular
external eld K, then the elds must all approach zero as the pole of the propagator approaches
the horizon. In the case of spherical symmetry it is useful to take the boundary of Region I to be
a sphere as it its then possible to decompose all the functions into spherical harmonics and the
decay rates for each multipole moment can be calculated separately.
We thus have a method to calculate the rates of decay of propagators as their poles approach
the horizon in terms of the distance in the optical metric. We may relate the optical distance along
a radial geodesic starting from some xed point, , to the proper distance to the horizon along
that geodesic in the physical metric, , by:
Ce

(4.67)
in the region near the horizon. This allows us to re-express the decay rates in terms of proper
distance to the horizon in the physical metric. We will give constructions below for some simple
propagators in hyperbolic space which are useful when constructing the approximate propagators
in region I.
22
4.3.1 Example - Massive scalar eld
In the case of a massive scalar eld satisfying the Klein-Gordon equation, we wish to nd a
propagator which satises:

g
m
2
= (x, x
0
). (4.68)
Using conformal transformations dened in section (4.4.2), solving this is equivalent to solving the
equation
1
H
5
(
h
+ ) +
1
H
_
k
2
m
2

2
_
=
1
H
6

h
(x, x
0
), (4.69)
where h refers, as always, to the metric on H
3
and H 1, 0 as we approach the horizon.
This is of the form supposed above and the exterior supports no solutions to the homogeneous
Klein-Gordon equation except the zero solution. The propagator in region I is given by:
e
ik
sinh
, where = D(x, x
0
), (4.70)
with D(p, q) the distance in the optical metric between p and q. Thus as x
0
goes to conformal
innity, the propagator falls o like e

. In terms of the proper distance of the pole of the

propagator from the horizon , we nd that the propagator vanishes like
1
. This is in agreement
with Teitelboim [14]. We will verify this analysis below by showing that for the k = 0 mode we
may solve the problem exactly throughout the exterior.
4.3.2 Example - Proca equation
The generalisation of Maxwells equations for electromagnetism to the case where the photon is
not taken to be massless is given by the Proca equations. In terms of a one-form A the vacuum
equations may be written:
d dA+m
2
A = 0 . (4.71)
In this case, we may quickly estimate the rate at which the information is lost as a charged particle
falls quasi-statically into a (Schwarzschild or Reissner-Nordstrom) back hole if the electromagnetic
eld is mediated by a massive vector boson. We make the ansatz:
A =
(x)
H(x)
dt, (4.72)
and nd that for a point charge at x
0
the function should satisfy:
1
H
5

h
m
2

2
H
=
1
H
6

h
(x, x
0
), (4.73)
where h, H and are as in the last section. This is once again of the form conjectured above
and for m ,= 0 there are no solutions to the vacuum equations regular throughout the exterior of
the black hole. The m = 0 case corresponds to a Maxwell eld and is treated below, however we
expect a signicant dierence as for this case the equations are gauge invariant.
The propagator in region I is given by:
1
e
2
1
, where = D(x, x
0
), (4.74)
23
where D(p, q) is as above. As the pole recedes to innity the propagator decays like e
2
corre-
sponding to a fall o as the square of the proper distance to the horizon,
2
. This decay at twice
the rate of the massive scalar boson case is also in agreement with Teitelboim.
4.3.3 Example - Forces from Neutrino pair exchange
As noted above there is a force in at space between leptons mediated by neutrinos which might
in principle be used to measure the lepton number of a black hole. As a black hole should not
have a measurable lepton number associated with it, we will now consider the problem of neutrino
mediated forces in the vicinity of an event horizon. We expect such forces to vanish as a lepton
approaches the horizon. We will require the following short Lemma which may be proven by
considering the behaviour of the spin connection under conformal transformations.
Lemma 4.1. Suppose g =
2
g are two conformally related n-dimensional metrics and

=
(n1)/2

is a Dirac spinor, then

,D

=
(n+1)/2
,D. (4.75)
Thus solutions of Diracs equation for g may be rescaled to solutions for g. Also
_

=
_

(d
n1
)(
1n
2
)
2

=
_

, (4.76)
so that orthonormality is preserved. It is not true however that if we start with a complete set
then after rescaling we have a complete set. We will now restrict to the case where d = 4 and the
conformal transformation depends only on the spatial coordinates X so that the metric remains
static.
By considering how the equation (4.35) transforms under such a conformal transformation then
assuming the rescaled solutions to Diracs equation are complete we may calculate the interaction
potential for the neutrino mediated force. Suppose that V (X, X

) represents the potential at X,

due to an electron at X

, with the metric g. Then the potential

V (X, X

) for the metric g is given

by:

V (X, X

) =
3
(X)
2
(X

)V (X, X

). (4.77)
There is clearly an asymmetry between the source and the test particle, however this is due the
the redshift eect which means that an energy measured at dierent spatial points will vary.
As an example, we may consider conformally rescaling R
t
H
3
to a metric on a static patch
of the de Sitter space. In this case we pick an arbitrary point X
0
and we may write the de Sitter
metric as
g =
1
(X X
0
)
2
g, (4.78)
where g is the metric on R
t
H
3
. Suppose we take a patch with the observer at the origin, then
the potential measured by this observer due to an electron at X

is given by:
V () =
G
2
W
cosh
2

8
3

2
sinh
4

_
cosh + sinh
2
Shi
_
, (4.79)
24
with cosh = X X

. As the electron approaches the horizon, and the potential is

extinguished like e

/
3
, thus demonstrating the no-hair property of the de Sitter cosmological
horizon for neutrino mediated forces.
Unfortunately this method does not work completely for the Schwarzschild event horizon be-
cause the conformally rescaled solutions of the Dirac equation do not form a complete basis,
essentially because neutrinos may start either at spatial innity or at the horizon, this point is
made by Teitelboim and Hartle [14, 23]. Accordingly, we do not reproduce precisely the extinction
rate of Hartle, who nds the potential vanishes like e

/, but we do nd the correct exponential

rate.
4.4 Exact Calculations for Schwarzschild and Reissner-Nordstrom
4.4.1 Electric Charge
We would like to nd the eld due to a static electric charge in the Schwarzschild or Reissner-
Nordstrom background. In order to do this, we make use of the fact that Maxwells equations in
vacuo:
dF = 0, d
g
F = 0, (4.80)
are conformally invariant in 4 space-time dimensions. Thus if we can nd the eld due to a point
particle with respect to the optical metric g
opt.
then we know the eld with respect to the physical
metric. The reason for this is that in 4 spacetime dimensions, for any 2-form with conformal
weight 0, we have:

2
g
=
g
. (4.81)
Thus the Maxwell action
S =
_
M
F F (4.82)
does not change under a conformal transformation. It is also helpful to note at this stage that the
charge contained inside a 2-surface :
Q

=
_

(F) (4.83)
is conformally invariant, where : / is the inclusion map and may refer to any represen-
tative of the conformal class of g, by (4.81).
In order to solve the rst Maxwell equation, we as usual introduce a one-form potential A and
make a static ansatz:
F = dA, A = (x)dt . (4.84)
The reason for this static ansatz is primarily the fact that the resulting equations are analytically
tractable. It will give a good approximation to the eld of a freely falling particle, provided the
particle is not moving very quickly. Alternatively, one may imagine a thought experiment where a
charge is lowered from innity towards the black hole horizon and measurements of the elds are
made as the charge approaches the black hole.
The second Maxwell equation becomes the familiar Laplace equation for with respect to the
optical metric:
d
hopt.
d = 0. (4.85)
25
We will make use of the fact noted above that
h
opt.
= H
4
h, (4.86)
where h is the hyperbolic metric of H
3
with radius 1 and H is harmonic on H
3
. We are therefore
able to relate the Laplacian of h
opt.
to the Laplacian of h. We will require the following lemma:
Lemma 4.2. Suppose h
1
and h
2
are two three dimensional metrics with related Laplace operators
1
and
2
. Further suppose that they are conformally related:
h
1
= H
4
h
2
. (4.87)
Then if = H
1
the following relation holds:
1
=
1
H
5
2

H
6
2
H. (4.88)
Proof. Use the standard formula =
1

x
i

gg
ij
x
j
and make the substitutions above, then
collect terms to nd (4.88).
We wish to nd the Greens function for the Laplacian on

h = h
opt.
. This satises the following
equation:

G(x, x
0
) =
(

h)
(x, x
0
), (4.89)
Where

is taken to act on the x coordinates. The Dirac delta function is dened by the require-
ment:
_
U

(g)
(x, x
0
)dvol
g
=
_

_
1 if x
0
|
0 if x
0
/ |
(4.90)
for an open subset | /. Applying the above lemma and making use of the fact that
h
H = 0
we have that

=
1
H
5

h
, (4.91)
with and related as above. Inserting this into the denition of the Greens function we have

G(x, x
0
) = H(x)
1
G(x, x
0
), where G obeys:

h
G(x, x
0
) = H(x)
5

h)
(x, x
0
) = H(x
0
)
1

(h)
(x, x
0
), (4.92)
where in this last step we use properties of the Dirac delta function. It would appear that we have
simply replaced one Greens function problem with another, however the great advantage is that
we now seek Greens functions on H
3
which is maximally symmetric so if we can nd the Greens
function for x
0
= 0 we can generate all Greens functions by SO(3, 1) transformations.
Suppose V
O
(x) satises

h
V
O
(x) =
(h)
(x, O), V
O

1
4
as D(x, O) , (4.93)
26
where O is a xed point in H
3
and D(x, O) is the hyperbolic distance from x to O. In Beltrami
coordinates V
O
(x) = (4 [x[)
1
. By SO(3, 1) invariance, for any other point x
0
we can nd a
isometry T : H
3
H
3
which satises
T

x
0
h = h, T
x
0
(x
0
) = O. (4.94)
We then dene
V
x
0
= T

x
0
V
O
, i.e. V
x
0
(x) = V
O
(T
x
0
(x)). (4.95)
The map T is not uniquely dened, but since V
O
is spherically symmetric any two maps T satisfying
(4.94) give the same function V
x
0
. This new function satises

h
V
x
0
(x) =
(h)
(x, x
0
). (4.96)
Putting together (4.92) and (4.96) we nd that the Greens function for the Laplacian of h
opt.
has
the form

G(x, x
0
) =
1
H(x)H(x
0
)
(V
O
(T
x
0
(x)) +A) +B. (4.97)
We have now to specify boundary conditions. The constant B is unphysical, and if we choose
B = 0, the potential will vanish at the asymptotically at end. The constant A arises because
if V
x
0
satises (4.96) then so does V
x
0
+ A. This constant gives rise to a non-trivial eld which
corresponds to the black hole carrying a (linearised) charge. Transforming to Eddington-Finkelstein
coordinates shows that the function

G is regular at the horizon for all values of A, so we must look
elsewhere for our nal boundary condition. This comes from the fact that Gauss law should be
satised. If one considers a surface which encloses the black hole, but not the point x
0
then Q

should vanish. Enforcing this condition xes A. In the case where the harmonic function H takes
the Reissner-Nordstrom form:

H(x) = 4V
O
(x) +m (4.98)
Gauss law requires A =
m
4
and we nally have:

G(x, x
0
) =
1
H(x)H(x
0
)
_
V
O
(T
x
0
(x)) +
m
4
_
. (4.99)
This construction is valid for both Schwarzschild and Reissner-Nordstrom and gives the linear
perturbation to the electromagnetic eld due to a static point charge located in the spacetime.
The potential has been found in terms of geometric objects of hyperbolic space, and so is valid for
any coordinate system on H
3
.
We can see from this equation how the information associated with the precise location of
the charged particle is lost as it is lowered towards the black hole. The only term which is not
spherically symmetric about O in (4.99) is the V
O
term. As the point x
0
recedes from O towards
the black hole horizon which is at the conformal innity of H
3
, this term approaches a constant
exponentially quickly in D(x
0
, O). The potential tends to the spherically symmetric eld associated
with the black hole carrying a charge and deviations from this eld fall exponentially with D(x
0
, O).
We plot below the isopotentials for a point charge in Schwarzschild, taking isotropic coordinates
so that the spatial sections are conformally at and the lines of force are normal to the isopotentials.
Isotropic coordinates for Schwarzschild correspond to Poincare coordinates on the hyperbolic space
from which the optical metric is constructed. The black region in the plots corresponds to the
interior of the black hole event horizon and we have returned the asymptotically at end to innity.
27
Figure 3: Plots showing the equipotentials as a point charge is lowered into a Schwarzschild black
hole in isotropic coordinates. The horizon is located at the boundary of the black disc and the
point charge is red. The blue contour is the equipotential of the horizon.
28
4.4.2 Scalar Charge
We will now show how to treat exactly a static massless scalar eld in the Schwarzschild or Reissner-
Nordstrom backgrounds. We will once again make use of the optical metric and the relationship
between this metric and the hyperbolic metric. The main result we shall require is summarised as
Lemma 4.3. If g is a scalar at static metric of the form:
g =
2
g
opt.
=
2
(dt
2
+H
4
h), (4.100)
with h the metric on H
3
with radius 1 and H harmonic on H
3
then must satisfy

h
(H) + H = 0. (4.101)
Further, if =
1
H
1
, then

g
=
3
_

1
H

t
2
+
1
H
5
(
h
+ )
_
. (4.102)
Proof. This follows from standard identities for conformal transformations.
We would like to calculate the eld G(x, x
0
) at x due to a unit static scalar charge at x
0
. For
a general moving point charge, G satises:

g
G(x, x

()) =
_

(g)
(x, x

())d , (4.103)
where is the proper time along the worldline x

() of the charged particle. Assuming that this

particle is static, we nd using the Lemma above that G(x, x
0
) =
1
(x)H
1
(x)(x, x
0
) where
satises

h
(x, x
0
) + (x, x
0
) = H
1
(x
0
)
(h)
(x, x
0
). (4.104)
It is convenient once again to make use of the SO(3, 1) invariance of hyperbolic space in order to
solve this equation. We rst seek solutions to the simpler equation

O
(x) +
O
(x) =
(h)
(x, O) (4.105)
subject to the condition that and d are bounded in the metric induced by h as D(x, O) .
One nds that the solution is related to the metric functions for Reissner-Nordstrom according to:

O
(x) =
1
4

(x)H(x). (4.106)
Unlike in the scalar charged case, there is no arbitrary constant. Using the identities above, we
nd that the eld due to a static unit scalar charge at a point x
0
is given by:
G(x, x
0
) =

O
(T
x
0
(x))
H(x)H(x
0
)(x)
, (4.107)
where T
0
is dened as in the previous section.
We may once again consider the behaviour of G(x, x
0
) as the scalar charge approaches the
horizon. We nd that G and its derivatives fall o like e
D(x
0
,O)
as D(x
0
, O) . Unlike the
case of an electric charge, there is no residual monopole term, so the black hole does not become
charged. Thus, we see precisely how massless scalar hair is shed as a point scalar charge is lowered
into a black hole and it it as predicted by our approximate argument given above.
29
5 Conclusion
We have seen how it is possible to make use of the universal asymptotics of the optical metric
near a Killing horizon to study physical problems in this region. We have presented a method
of studying null geodesics based on the Gauss-Bonnet theorem which directly links the negative
curvature of the optical geometry to physical lensing scenarios. We have re-derived classic results
about the loss of hair as objects fall into a black hole in a simplied manner and by making use
of the universality of the near horizon optical metric, extended these results to apply beyond the
Schwarzschild case where they were rst investigated.
A Integration on CP
1
In section 4.1 we found that the space of solutions to Diracs equation on R
t
H
3
could be identied
with R
+
CP
1
where the CP
1
arose by identifying Weyl spinors which were complex multiples of
one another. In subsequent calculations it was necessary to integrate over this space of solutions
in a Lorentz invariant fashion. The aim of this appendix is to explain how this is possible.
There are two key observations to be made. Firstly it should be noted that the space of Weyl
spinors caries a natural 2-form dened by:
[] = 2i

. (A.1)
This is Lorentz invariant by construction.
Secondly we may represent CP
1
as a smooth 2-dimensional surface in C
2
, where we assume
that for almost every [] CP
1
there is exactly one point such that [] = [ ]. Since we are
interested in integrating over CP
1
it doesnt matter if this fails to be true for some set of measure
zero. Suppose now that we chose a dierent surface

. In order that this fulls the requirements

to represent CP
1
there must exist some smooth function : C
2
C such that for almost every
point , ()

. In other words we may, by extending the domain of if necessary dene

a local dieomorphism
: U C
2
U

C
2
() (A.2)
such that () =

up to a set of measure zero. One may verify that

= [[
4
. (A.3)
Thus if we have a function f : C
2
C which is a scalar under Lorentz transformations and which
satises f() = [[
4
f() then the integral
_

f (A.4)
is independent of which surface in C
2
we use to represent CP
1
. Suppose now that f = f(, X
i
)
where X
i
are some vectors in E
3,1
and such that
f(
s

,
v

X
i
) = f(, X
i
), (A.5)
30
where
s
,
v
are the spinor and vector representations of the Lorentz transformation respectively.
If we pick a surface which represents CP
1
, we may dene a function
I(X
i
) =
_

f(, X
i
) =
_

f(
s

,
v

X
i
). (A.6)
If

is the function on C
2
dened by left multiplication by
s

then we may use the Lorentz

invariance of the measure to write
I(X
i
) =
_

(f(,
v

X
i
)) =
_

()
f(,
v

X
i
)
=
_

f(,
v

X
i
) = I(
v

X
i
), (A.7)
where we have made use of the Lorentz invariance of f and , together with the independence of
the integral on the choice of representative of CP
1
. Thus the integral is a Lorentz scalar a fact
which we make use of in section 4.1.4 to calculate the integral (4.53)
As an example, we may take = (1, z)
t
/(1 +[z[
2
)
1/2
: z C which covers all of CP
1
except
one point. We nd then that
[

=
2i
_
1 +[z[
2
_
2
dz d z, (A.8)
the standard measure on the sphere under stereographic projection. We use this fact in calculating
the neutrino mediated force between electrons.
References
[1] I. Sachs and S. N. Solodukhin, Horizon holography, Phys. Rev. D 64 (2001) 124023
[arXiv:hep-th/0107173].
[2] Z. Haba, Green functions and dimensional reduction of quantum elds on product manifolds,
Class. Quant. Grav. 25 (2008) 075005 [arXiv:0709.3227 [hep-th]].
[3] G. W. Gibbons and S. N. Solodukhin, The Geometry of Large Causal Diamonds and the
No Hair Property of Asymptotically de-Sitter Spacetimes, Phys. Lett. B 652 (2007) 103
[arXiv:0706.0603 [hep-th]].
[4] G. W. Gibbons, No glory in cosmic string theory, Phys. Lett. B 308 (1993) 237.
[5] K. S. . Thorne, R. H. . Price and D. A. . Macdonald, Blach Holes: the membrane paradigm,
New Haven, USA: Yale Univ. Pr. (1986)
[6] S. R. Das, G. W. Gibbons and S. D. Mathur, Phys. Rev. Lett. 78 (1997) 417
[arXiv:hep-th/9609052].
[7] G. Policastro, D. T. Son and A. O. Starinets, Phys. Rev. Lett. 87 (2001) 081601
[arXiv:hep-th/0104066].
31
[8] E. T. Copson, Proc. R. Soc. A 118 (1928) 184.
[9] B. Linet, Electrostatics and magnetostatics in the Schwarzschild metric, J. Phys. A 9 (1976)
1081.
[10] B. Leaute and B. Linet, Electrostatics in a Reissner-Nordstrom space-time, Phys. Lett. A
58 (1976) 5.
[11] M. A. Abramowicz, I. Bengtsson, V. Karas and K. Rosquist, Poincare ball embeddings of
the optical geometry, Class. Quant. Grav. 19 (2002) 3963 [arXiv:gr-qc/0206027].
[12] G. W. Gibbons and M. C. Werner, Applications of the Gauss-Bonnet theorem to gravitational
lensing, arXiv:0807.0854 [gr-qc].
[13] O Lodge Nature 104 (1919) 354
[14] C. Teitelboim, Nonmeasurability of the quantum numbers of a black hole, Phys. Rev. D 5
(1972) 2941.
[15] C. G. . Callan and F. Wilczek, Infrared behaviour at negative curvature, Nucl. Phys. B 340
(1990) 366.
[16] M. Atiyah and P. Sutclie, Skyrmions, instantons, mass and curvature, Phys. Lett. B 605,
106 (2005) [arXiv:hep-th/0411052].
[17] G. Cognola and L. Vanzo, Bose-Einstein condensation of scalar elds on hyperbolic mani-
folds, Phys. Rev. D 47, 4575 (1993) [arXiv:hep-th/9210003].
[18] Dennis DeTurck and Herman Gluck, J. Math. Phys. 49, 023504 (2008), Electrodynamics and
the Gauss linking integral on the 3-sphere and in hyperbolic 3-space, DOI:10.1063/1.2827467
[19] P. A. M. Dirac, The electron wave equation in de-Sitter space Ann. Math. 36 (1935) 657
[20] D. V. Paramonov, N. N. Paramonova and N. S. Shavokhina, The Dirac equation in the
Lobachevsky space time, JINR-E2-2000-79;
[21] U. Moschella and R. Schaeer, Quantum Theory on Lobatchevski Spaces, Class. Quant.
Grav. 24 (2007) 3571 [arXiv:0709.2795 [hep-th]].
[22] G. Feinberg and J. Sucher, Phys. Rev. 166, 1638 (1968)
[23] J. B. Hartle, Can A Schwarzschild Black Hole Exert Long Range Neutrino Forces? In *J
R Klauder, Magic Without Magic*, San Francisco 1972, 259-275
[24] E. V. Gorbar, Dynamical symmetry breaking in spaces with constant negative curvature,
Phys. Rev. D 61, 024013 (2000) [arXiv:hep-th/9904180].
32