Kottler Spacetime in Isotropic Static Coordinates

∗
R Solanki

Department of Physics
University of Texas at Dallas
Richardson, TX 75080, USA
arXiv:2103.10002v2 [gr-qc] 19 Nov 2021

November 22, 2021

Abstract
The Kottler spacetime in isotropic coordinates is known where the metric is time-dependent. In
this paper, the Kottler spacetime is given in isotropic static coordinates (i.e., the metric com-
ponents are time-independent). The metric is found in terms of the Jacobian elliptic functions
through coordinate transformations from the Schwarzschild-(anti-)de Sitter metric. In canonical
coordinates, it is known that the unparameterized spatially projected null geodesics of the Kot-
tler and Schwarzschild spacetimes coincide. We show that in isotropic static coordinates, the
refractive indices of Kottler and Schwarzschild are not proportional, yielding spatially projected
null geodesics that are different.

1 Introduction
The McVittie metric [11] is the Kottler spacetime in isotropic coordinates in which the metric
components are time-dependent:

1−µ 2 2
 
2
dt̄ + (1 + µ)4 a(t̄)2 dr̄ 2 + r̄2 (dθ2 + sin2 θdφ2 ) ,
 
ds = − (1.1)
1+µ
where q
H t̄ M Λ
t̄
a(t̄) = e =e 3. and µ = (1.2)
2ar̄
For Λ = 0 (i.e., a(t) = 1), this line element reduces to the Schwarzschild spacetime in isotropic
coordinates and for M = 0, to the Friedmann-Lemaı̂tre-Robertson-Walker (FLRW), i.e., Λ only
universe. Moreover, for µ << 1, the McVittie metric turns to perturbed the FLRW in a Newtonian
gauge [5]. Under the transformation
dξ rH
r = ar̄(1 + µ)2 , t = t̄ + ξ(r) where = p , (1.3)
dr f (r) 1 − 2M/r

the McVittie metric transforms into the Kottler metric in canonical coordinates (see equation (4.1)
below).
In this paper, the Kottler spacetime is given in isotropic coordinates where the metric compo-
nents are time-independent, i.e., in isotropic static coordinates (because g0i = 0). The two cases
∗
E-mail: [email protected]

1
for which the spacetime is static are the region between two horizons when 0 < Λ < 1/9M 2 and
the region beyond the single horizon when Λ < 0. The case with two horizons is presented first,
followed by one horizon. The metric tensor in isotropic coordinates {t, ρ, θ, φ} is obtained by trans-
forming from canonical coordinates {t, r, θ, φ}, and as it turns out, the transformation r = r(ρ)
follows Jacobian elliptic functions. In isotropic coordinates, the spatial metric is conformal to the
Euclidean metric, whereas in canonical coordinates, the circumference of a circle centered at the
origin is 2πr. Throughout the paper, the dot denotes differentiation with respect to the argument.
In order to distinguish between the limit and the variable of integration, prime is used for the
latter. Moreover, Greek indices represent spacetime coordinates (where x0 ≡ t) and Roman indices
represent spatial coordinates.
The metric in isotropic static coordinates has several advantages. In static coordinates, the
propagation of light follows the variational principle of Fermat, and in isotropic coordinates, the
spatial metric shares the same conformal properties as the Euclidean. Moreover, put together, the
light propagation can be mimicked by an optical medium in ordinary geometric optics with an
appropriate index of refraction [13]. For the line element in static coordinates

ds2 ≡ gµν dxµ dxν = g00 dt2 + gij dxi dxj = −g00 (g̃µν dxµ dxν ) , (1.4)

the spatial trajectories of null geodesics (i.e., projection of null geodesics on a constant time hyper-
surface) are given by the geodesics of the optical metric1 (or Fermat metric) [2, 7, 13]
gij i j
ds2opt = − dx dx = nij dxi dxj . (1.5)
g00

One can prove this using the following property of conformal transformations (for a proof using
Fermat’s principle, see for example, page 273 of Ref.[10]). Null geodesics remain invariant under
conformal transformations; however, their parameterization changes from affine to non-affine (see
Appendix D of Ref.[18]). For the metric g̃µν , since g̃00 = −1 and g̃0i = 0, the Christoffel symbols
Γ̃ µ0ν vanish. Therefore, the spatial part of null geodesics2 follow

d2 xi dxj dxk
+ Γ̃ ijk = 0, (1.6)
dλ̃2 dλ̃ dλ̃
which is precisely the geodesic equation of the Fermat metric nij . In other words, the spatial
trajectories of null geodesics are the geodesics of the Fermat metric gij /|g00 |, not of spatial metric
gij [10].
For the Schwarzschild and Kottler spacetimes in canonical coordinates, it is known that the
spatial trajectory of null geodesics satisfies the identical second-order ordinary differential equation
[8]. This differential equation, however, represents unparameterized curves (r = r(φ) in this case,
1
In this paper, we consider light rays propagating along the null geodesics of the spacetime metric. Furthermore,
the optical metric nik is of the Riemannian signature.
If the propagation of light is influenced by an isotropic non-dispersive optical medium, then the light rays follow
the null geodesics of another Lorentzian metric (see e.g., [14] and references therein). In literature, this metric of the
Lorentzian signature is also called the optical metric. Therefore, to avoid potential confusion, nik will be referred to
as the Fermat metric henceforth.
2
The temporal part, since Γ̃ 0ik = 0, is given by, dt/dλ̃ = constant. Thus, λ̃ is affine parameter to both, null
geodesics of g̃µν and geodesics of nik , since

dxµ dxν dt dxi dxk

g̃µν = 0 =⇒ = nik = constant.
dλ̃ dλ̃ dλ̃ dλ̃ dλ̃

2
if the equatorial plane is considered). Therefore, the proper length and conformal properties such
as angles are not identical. For example, in three dimensions, the unparameterized geodesics of
both Euclidean and Beltrami metrics3 are straight lines. However, parameterizing these geodesics
by arc length reveals the difference [7]. This property that the unparameterized geodesics of the
Fermat metrics coincide is known as projective equivalence. If two Fermat metrics nij and n̄ij are
projectively equivalent, then the projective curvature tensors W iklm and W̄ iklm are equal. Here,
the Weyl projective curvature tensor, constructed from the Fermat metric nij (of the Riemannian
signature), is given by [6]
1 i
W iklm = Riklm + (δm Rkl − δli Rkm ). (1.7)
2
In general, for an m-dimensional Riemannian manifold, the factor of 1/2 is replaced by 1/(m − 1).4
In this paper, it is shown that for the Schwarzschild and Kottler spacetimes in isotropic coordinates,
the conformal properties of the corresponding Fermat (and spatial) metrics are the same, but
the projective equivalence no longer exist (i.e., opposite example to Ref.[7]).5 In other words, in
isotropic static coordinates, the unparameterized spatially projected null geodesics of the Kottler
spacetime are Λ−dependent, and the angles measured by stationary observers are the same as
3
The line element for one half of the sphere is given by,

r2 dθ2 + sin2 θ dφ2

2 dr2
dl = + .
(1 + r2 /R2 )2 (1 + r2 /R2 )

4
The Riemannian metrics nij (xk ) and n̄ij (xk ) are projectively equivalent if and only if their associated Christoffel
symbols are related by
Γ̄ikl = Γikl + δki al + δli ak , (1.8)
where ak is a covariant vector [6]. For an arbitrary ak , the changes in the Christoffel symbols above are called
projective transformations. In other words, projective transformations preserve unparameterized geodesics, and the
projective curvature tensor W iklm is invariant under these transformations [16]. This tensor vanishes identically in
two dimensions. In dimensions greater than 2, a metric nij (xk ) is projectively flat (i.e., there exists a projective
transformation through which one can obtain n̄ij (xk ) such that R̄iklm = 0) if and only if W iklm vanishes [16].
Basically, unparameterized geodesics are straight lines for a projectively flat metric. Conformal transformations,
ñij (xl ) = exp (2ω) nij (xl ) where ω = ω(xl ), on the other hand, preserve angles in corresponding directions at
corresponding points [6]. Here, the associated Christoffel symbols are related by

Γ̃ikl = Γikl + δki ∂l ω + δli ∂k ω − gkl ∂ i ω , (1.9)

where ∂k ≡ ∂/∂xk . The conformal tensor C iklm is invariant under conformal transformations and vanishes identically
in three dimensions [6]. For the Lorentzian metrics, the unparameterized null geodesics remain invariant under
conformal transformations; although relationship (1.8) is not satisfied, the last term in (1.9) vanishes due to the null
condition when (1.9) is substituted in the null geodesic equation.
5
Consider a metric gµν in static coordinates (i.e. ∂0 gµν = 0 and g0i = 0) with the Fermat metric nik . Under the
coordinate transformations x0µ = x0µ (xν ),

0 ∂xl ∂xm ∂x0i ∂xb ∂xc ∂xd

g00 = g00 , n0ik = nlm , W 0iklm = W abcd ,
∂x0i ∂x0k ∂xa ∂x0k ∂x0l ∂x0m
0
if gµν is also static. If two Fermat metrics nlm and n̄lm are projectively equivalent, then

W abcd = W̄ abcd .
i
However, this does not necessarily mean that W 0iklm and W̄ 0 klm are equal. (If they vanish, on the other hand, the
equality holds in all static coordinates because if a tensor vanishes in one coordinate system, then it vanishes in all
because of the tensor transformation rule. The projective curvature tensor vanishes for a constant curvature space
[7].) This can be explained, for example, through transformations between the isotropic {t, ρ, θ, φ} and canonical
{t, ρ, θ, φ} coordinates. In general, the transformation between the radial coordinates, r = r(ρ) or ρ = ρ(r), is
different for different spacetimes.

3
coordinate angles (locally). Thus, the isotropic static form of the Kottler spacetime is more suitable
for studying gravitational lensing using quasi-Newtonian approximations (e.g., the lens equation
relating source and image positions; see [15] and references therein).
The paper is organized as follows. In Section 2, the equivalence between the spatially projected
null geodesics of the Kottler and Schwarzschild is extended from unparameterized to affinely param-
eterized curves. The analogy between the index of refraction (in standard geometric optics) and the
projective equivalence (in isotropic coordinates) is explored in Section 3. The problem of finding a
static isotropic metric is systematically developed in Section 4. In Section 5, for 0 < Λ < 1/9M 2 ,
the metric is calculated in terms of Jacobian elliptic functions. Similarly, for Λ < 0, the solu-
tion is presented in section 6. In both these sections, the index of refraction is plotted and the
Schwarzschild limit is discussed.

2 Remark on Parameterization
The unparameterized geodesics of projectively equivalent Fermat metics coincide, but affine (e.g.,
arc-length) parameterization differs [6, 7]. However, the spatial trajectory of affinely parameterized
null geodesics (of the corresponding spacetime metrics) can be identical; an example is the Kottler
and Schwarzschild spacetimes in canonical coordinates. Here, the null geodesics of the Kottler
metric (4.1) in the equatorial plane can be derived, for example, using the same approach as for
the Schwarzschild metric (see Section 6.3 of Ref.[18]):

Λr2
 
2M
1− − ṫ = E, (2.1)
r 3
r2 φ̇ = J, (2.2)
J2
   
2 2 Λ 2 2M
ṙ = E + J − 2 1 − , (2.3)
3 r r

where E and J are the constants of motion, energy and angular momentum of the photons respec-
tively. Moreover, ṫ = dt/dλ where λ is affine parameter and equation (2.3) is the null condition
in terms of the constants of motion. If the affine parameter is rescaled such that ẋµ is the wave
four-vector, then E becomes the frequency of light rays [10]. As can be seen, r(λ) and φ(λ) if the
frequency is modified to r
Λ
Ē = E 2 + J 2 , (2.4)
3
are the same as those in Schwarzschild. In other words, the spatial trajectory of affinely parame-
terized null geodesics of Kottler and Schwarzschild spacetimes (in canonical coordinates) coincide
but the frequency of light rays differs.

3 Index of Refraction and Projective Equivalence

In ordinary optics, the index of refraction represents the ratio between the speed of light in vacuum
and in the optical medium. Therefore, for an isotropic optical medium with an index of refraction
n(ρ) the time it takes for a light ray to traverse spatial distance (say between spatial points 1 and
2) is given by,
Z 2
t= n(ρ)dl (3.1)
1

4
where dl2 = dρ2 + ρ2 dθ2 + sin2 θ dφ2 and n(ρ)dl is the optical length. Moreover, according to

Fermat’s variational principle, the light ray follows spatial trajectory (between points 1 and 2) for
which the optical length is stationary, i.e., δt = 0 [9]. This is equivalent to finding the geodesics of
a curved space (Riemannian) with the line element,
2
= n2 (ρ) dρ2 + ρ2 dΩ2 ,


dlopt (3.2)

where dΩ2 = dθ2 + sin2 θ dφ2 .

The spherically symmetric static spacetime in isotropic coordinates can be written as

ds2 = F (ρ) −dt2 + n2 (ρ)(dρ2 + ρ2 dΩ2 ) .

 
(3.3)

The associated Fermat metric, from equation (1.5), is given by equation (3.2). Therefore, the
metric function n(ρ) is analogous to the index of refraction of a spherically symmetric optical
medium in ordinary optics (see [13] and references within). The metric of equation (3.3) follows
the same relationship (3.1), since the spacetime interval vanishes for the null path. And for the
metric conformal to it (conformal factor F (ρ)), t is the proper time and n(ρ)dl is the proper length.
The projective equivalence between two Fermat metrics with refractive indices n(ρ) and n̄(ρ) can
be determined directly from the unparameterized geodesic equation (see Appendix C):
2 
d2 ρ
   
dρ ṅ 2 ṅ
= + + ρ + ρ2 . (3.4)
dφ2 dφ n ρ n

Here, the coefficients of the differential equation are invariant, if and only if the ratio ṅ/n remains
the same, which after integration suggests that n̄(ρ)/n(ρ) =constant. Thus, in isotropic (static)
coordinates, two Fermat metrics are projectively equivalent if and only if their refractive indices
are proportional to each other, with the same proportionality constant everywhere [1].

4 From Canonical to Isotropic coordinates

The Kottler spacetime in Schwarzschild canonical coordinates is given by

dr2
ds2 = −f (r)dt2 + + r2 dΩ2 , (4.1)
f (r)
 
Λr2
where f (r) = 1 − 2M r − 3 . For Λ > 0, it is also called the Schwarzschild-de Sitter metric
and for Λ < 0, Schwarzschild-anti-de Sitter metric. Transformation to isotropic coordinates (i.e.,
equation (3.3)) yields

dr rp r 1
= f (r), n(ρ) = p , F (ρ) = f (r). (4.2)
dρ ρ ρ f (r)

The variables in the first equation (ordinary differential) are already separated with the term
p
r f (r) polynomial of degree four under the square root. If the polynomial has no multiple roots,
then the solution of this equation reduces to an elliptic integral. The transformation between radial
coordinates ρ = ρ(r) is given by Legendre’s canonical elliptic integral of the first kind (see Appendix
A). However, since we are interested in r = r(ρ), the solution reduces to the elliptic functions of
Jacobi, obtained by the inversion of elliptical integrals of the first kind. Evaluation of the integral

5
requires knowing the roots of the polynomial (including the static region and horizons to specify
the range). Of course, r = 0 is one root of the polynomial,
 
Λ 3r 6M Λ
2
r f (r) = − r r − 3
+ = − rf¯(r). (4.3)
3 Λ Λ 3
If r1 , r2 and r3 are the roots of cubic equation f¯(r) = 0, then
6M 3
r1 r2 r3 = − , r1 + r2 + r3 = 0, r1 r2 + r2 r3 + r3 r1 = − . (4.4)
Λ Λ
Here, the corresponding values in isotropic coordinates are taken as ρ1 , ρ2 and ρ3 . Additionally, if
(see e.g., [3] for the solution of the cubic equations or Appendix B)
1


 > 0, i.e., Λ < 0 or Λ > 2
, then a real root and a complex conjugate pair.

 9M
2
  
9M 1 
1
2
− 3 = 0, i.e., Λ = , then all roots are real and at least two are equal.
Λ Λ 
 9M 2
 < 0, i.e., 0 < Λ < 1 , then all roots are real.



9M 2
(4.5)
Let’s consider each case separately (see e.g., section 5.2 of Ref.[13]).

Λ<0

In this case, from equation (4.4), the product of all roots is positive. And since the product of the
complex conjugate pair (e.g. r2 and r3 ) is positive, the real root (r1 ) is positive. Thus, there is one
horizon at r = r1 with a static region outside (for r1 < r < ∞, f (r) > 0).6

0 < Λ < 1/9M 2

All roots are real in this case. The product of all roots is negative and the sum is zero. Therefore,
one root (e.g. r1 ) is negative and the rest are positive (take r2 < r3 ). Therefore, there are two
horizons at r = r2 and r = r3 with a static region in between (f (r) > 0 for r2 < r < r3 ).7

Λ = 1/9M 2

Since the sum of all roots vanishes, only two roots (e.g. r2 and r3 ) are equal. And from the
product of the roots, the remaining root (r1 ) is negative, since Λ > 0. Thus, there is one horizon
at r = r2 = r3 , however, there is no static region.

Λ > 1/9M 2

In this case, the real root (e.g. r1 ) is negative, since the product of the roots is negative and the
product of the complex conjugate pair (r2 and r3 ) is positive. Since f (r) 6= 0 for r > 0, there are
no horizons. Additionally, f (r) < 0 for r > 0, therefore, no static region exists.
Thus, only for the first two cases, static region exists. The line element for second case (with
two horizons) is calculated first.
6
Since f¯(r1 ) = 0, (r1 − 2M ) = Λr13 /3. Thus, 0 < r1 < 2M since Λ < 0.
7
Here, r3 > r2 > 2M > 0, since Λ > 0 and f¯(r2 ) = f¯(r3 ) = 0. Furthermore, from (4.4),
6M
2 r33 > r2 r3 (r2 + r3 ) = > 2 r23 .
Λ
Thus, Λr33 /3 − M = r3 − 3M > 0 and Λr23 /3 − M = r2 − 3M < 0. In other words, r1 < 0 < 2M < r2 < 3M < r3 .

6
5 For 0 < Λ < 1/9M2
The systemic approach to find the solution is to first find the roots of the cubic equation, followed by
setting up the integral with appropriate limits. This integral, through appropriate transformations,
can be brought to the standard form of elliptic integrals and functions.

Roots

Since the cosmological constant is positive, the integrand can be written as,
p p Λ
r2 f (r) = h r(r3 − r)(r − r2 )(r − r1 ) where h2 = , r1 < 0 < r2 < r < r3 . (5.1)
3
The roots of the cubic equation f¯(r) = 0, after simplification, reduce to (see Appendix B)
2 2  π 2  π
r1 = − √ cos σ, r2 = √ cos σ + , r3 = √ cos σ − , (5.2)
3h 3h 3 3h 3
where8 √ p
cos (3σ) = 27M h and sin (3σ) = 1 − 27M 2 h2 . (5.3)

Integral

Finally, from (4.2), r = r(ρ) can be found by evaluating,

dr0 dr0
Z r Z r Z ρ 0
dρ
q = p = 0
. (5.4)
0 0 0 0
r2 r0 2 f (r0 ) r2 h r (r3 − r )(r − r2 )(r − r1 ) ρ2 ρ

Under the transformation (see for example, equation (256.00) of the Handbook [4])9
r2 r3
r0 = , (5.6)
r3 − (r3 − r2 ) sin2 φ
the integral (5.4) reduces to the elliptic integral of the first kind (see equation (A.2)):
  ZΦ
ρ 2 dφ 2
ln = p p = p F(Φ, k), (5.7)
ρ2 h r3 (r2 − r1 ) 2
1 − k 2 sin φ h r3 (r2 − r1 )
0

where s
−r1 (r3 − r2 ) r3 (r − r2 )
k2 = ≤ 1, Φ(r) = sin−1 . (5.8)
r3 (r2 − r1 ) r(r3 − r2 )
In terms of Jacobi’s inverse elliptic function (see equation (A.5)),
p  
h r3 (r2 − r1 ) ρ
ln = sn−1 (sin Φ, k) . (5.9)
2 ρ2
8
For reference, cos (3σ) = cos3 σ − 3 sin2 σ cos σ and sin (3σ) = 3 cos2 σ sin σ − sin3 σ.
9
One can apply the transformation (5.6) since
r3 (r0 − r2 ) r3  r2 
0
= 1 − 0 = sin2 φ,
r (r3 − r2 ) (r3 − r2 ) r
is an increasing function with values 0 and 1 at the boundaries r0 = r2 and r0 = r3 respectively. In general, a direct
approach to evaluating this type of integral is to take
a1 + a2 sn2 (ψ 0 , k) dr0
r0 = such that = a5 dψ 0 (5.5)
a3 + a4 sn2 (ψ 0 , k)
p
h r0 (r3 − r0 )(r0 − r2 )(r0 − r1 )
where ai ’s are constants and 0 ≤ ψ 0 ≤ K(k) = F(π/2, k) [4].

7
Solution

From equation (5.9), r = r(ρ) can be found in terms of the Jacobian elliptic function:
 
r2 r3 hp ρ −r1 (r3 − r2 )
r= 2
where ψ(ρ) = r3 (r2 − r1 ) ln , k2 = . (5.10)
r3 − (r3 − r2 ) sn (ψ, k) 2 ρ2 r3 (r2 − r1 )
Consequently, the conformal factor is,

sn2 (ψ, k) dn2 (ψ, k) cn2 (ψ, k)

F (ρ) = r3 h2 (r2 − r1 )(r3 − r2 )2 · . (5.11)
[r3 − (r3 − r2 ) sn2 (ψ, k)]2
And the index of refraction is,
r
r2 r3 1
n(ρ) = · . (5.12)
h(r3 − r2 ) r2 − r1 ρ sn(ψ, k) dn(ψ, k) cn(ψ, k)

From equations (5.7) and (5.8), the complete elliptic integral reduces to,
 
 π  hp ρ3
K(k) = F Φ = , k = r3 (r2 − r1 ) ln . (5.13)
2 2 ρ2
Thus, the ratio between location of horizons is given by,
" #
ρ3 2
= exp p K(k) . (5.14)
ρ2 h r3 (r2 − r1 )

5.1 Schwarzschild limit

From equation (5.3),
!
√
r
1 dσ
3σ = tan−1 −1 , lim = − 3M. (5.15)
27M 2 h2 h→0 dh

Furthermore, as Λ → 0, h → 0 and from equations (5.3), (5.2) and (5.10),

r
ρ
σ → π/6, hr1 → −1, r2 → 2M, hr3 → 1, ψ → ln , k → 1. (5.16)
ρ2
Here, for r2 , both the numerator and denominator go to zero as h → 0, and the limit is evaluated by
applying L’Hospital’s rule and (5.15). For the unit modulus (i.e., k = 1), Jacobian elliptic functions
turn to hyperbolic functions.

sn(ψ, 1) = tanh ψ, cn(ψ, 1) = dn(ψ, 1) = sechψ. (5.17)

Substituting all these conditions into equations (5.10), (5.11) and (5.12) yield,

ρH 2
 r   
2 ρ M ρ
rs (ρ) = 2M cosh ln = 1+ , (5.18)
ρH 2 ρH ρ
1 − ρH /ρ 2
 r   
2 ρ
Fs (ρ) = tanh ln = , (5.19)
ρH 1 + ρH /ρ
h q i
3 ρ
2M cosh ln ρH M

ρH 3
 
ρH −1

ns (ρ) = · h  q i = 1+ 1− , (5.20)
ρ sinh ln ρ 2ρH ρ ρ
ρH

8
where the subscript s indicates the Schwarzschild and H horizon. Since this spacetime is asymp-
totically flat, ns (ρ) → 1 as ρ → ∞, yielding ρH = M/2. Thus,

ρH 3 ρH −1
   
ns (ρ) = 1 + 1− . (5.21)
ρ ρ

It is easy to see that the refractive indices (5.12) and (5.20) are not proportional; therefore, in
isotropic static coordinates, the Fermat metrics of Kottler and Schwarzschild are not projectively
equivalent.

5.2 Refractive Index

The Kottler spacetime, unlike the Schwarzschild, is not asymptotically flat. Additionally, this
spacetime in isotropic static form cannot be matched at the boundaries ρ = ρ2 and ρ = ρ3 due
to the presence of horizons (from equations (5.11) and (5.12), gρρ = F (ρ)n2 (ρ) is finite at the
horizons, but g tt = −1/F (ρ) is not). Therefore, one of the horizons, the integration constant,
cannot be determined in terms of M and Λ.
In order to plot the refractive index, we consider it as a function of z = ln(ρ/ρ2 ). Therefore,
from equation (5.12)
r
r2 r3 1 1
η(z) = · · , (5.22)
h(r3 − r2 ) r2 − r1 ρ2 ez sn(Ψ, k) dn(Ψ, k) cn(Ψ, k)
p
where 2Ψ(z) = z h r3 (r2 − r1 ). Without loss of generality, if we further consider
r
r2 r3
ρ2 = , (5.23)
4h(r3 − r2 ) r2 − r1

then from (5.16), ρ2 → M/2 as Λ → 0 and

4
η(z) = . (5.24)
ez sn(Ψ, k) dn(Ψ, k) cn(Ψ, k)

It is easy to see that σ, from equation (5.3), depends upon a single parameter y = M 2 h2 = M 2 Λ/3
where 0 < y < 1/27. Therefore, hri (where i = 1, 2 and 3), modulus k and argument Ψ all depend
upon single parameter y. The refractive indices for different values of y are plotted in figure 1.

6 For Λ < 0
This section follows a procedure similar to that in the previous section, except for the compactness
of notations, the modulus k in Jacobian elliptic functions is dropped as is usually done.

Roots

Since the cosmological constant is negative, the integrand can be written as,
p p Λ
r2 f (r) = H r(r − r1 )(r − r2 )(r − r3 ) where H 2 = − , 0 < r1 < r < ∞. (6.1)
3
The roots of the cubic equation f¯(r) = 0, after simplification, reduces to (see Appendix B)
√
r1 3
r1 = S1 + S2 , Re(r2 ) = Re(r3 ) = − , Im(r2 ) = −Im(r3 ) = (S1 − S2 ), (6.2)
2 2

9
Figure 1: The refractive index (5.12) can be written as a function of ln (ρ/ρ2 ) with the parameter
y = M 2 h2 = M 2 Λ/3 by utilizing the relationship (5.23). Here, 0 < y < 1/27. Although for the
relationship (5.23) the refractive index depends only on z and y, the horizon ρ2 depends upon both
M and h. Here, the vertical dashed line represents second horizon and horizontal, refractive index
equal to one.

where Re and Im represent the real and imaginary parts of a complex number respectively and
r !1/3 r !1/3
1 1 1 1
S1 = MH + + M 2H 2 , S2 = MH − + M 2H 2 . (6.3)
H 27 H 27

Integral

Thus, r = r(ρ) can be found by evaluating,

dr0 dr0
Z r Z r Z ρ 0
dρ
q = p = 0
. (6.4)
0 0 0 0
r1 r0 2 f (r0 ) r1 H r (r − r1 )(r − r2 )(r − r3 ) ρ1 ρ

10
This integration can be solved in a similar fashion as in the previous section, except in equation (5.5)
sn(ψ 0 , k) turns to cn(ψ 0 , k) since r2 and r3 form complex conjugate pair (see for example, equation
(260.00) of the Handbook [3]). However, the resultant modulus k is greater than one. One can
apply reciprocal modulus transformation, such that the modulus converts to k1 = 1/k. However,
the Jacobian amplitude becomes imaginary (or the cosine of the amplitude becomes greater than
one). One way to circumvent these problems is to apply the transformation x = 1/r which leads to
a polynomial of degree three under square root (this will lead to compactifiation of space however):
dx0
  Z x1
ρ 1 1
ln =√ p where xi = , 0 < x < x1 (6.5)
ρ1 2M x 0 0 0
(x1 − x )(x2 − x )(x3 − x ) r i

This integral can be evaluated, in general, by substitution (see for example, equation (243.00) of
the Handbook [4] for this case),
a1 + a2 cn(ψ 0 , k) dx0
x0 = such that = a5 dψ 0 . (6.6)
a3 + a4 cn(ψ 0 , k)
p
0 0 0
(x1 − x )(x2 − x )(x3 − x )
where ai ’s are constants, ψ 0 = ψ 0 (ρ0 ) and 0 ≤ ψ 0 ≤ 2K(k).

Solution

All constants (ai ’s) from equation (4.4) can be evaluated in terms of r1 , M and H. After the
simplification, √
r1 M (1 + cnψ)
r=√ √ √ √ , (6.7)
M − 3M − r1 + ( M + 3M − r1 ) cnψ
where  1/4  
4M (3M − r1 ) ρ 1 6M − r1
ψ(ρ) = ln , k2 = +p ≤ 1. (6.8)
r12 ρ1 2 64M (3M − r1 )
Now, as r → ∞, from the equation (6.7)
√ √
3M − r1 − M
√ = φ∞ =⇒ ψ∞ = cn−1 (φ∞ , k) = F cos−1 (φ∞ ), k ,
 
cnψ∞ = √ (6.9)
3M − r1 + M
yielding the range,
" 1/4 #
r12

ρ∞
ρ ∈ (ρ1 , ρ∞ ) where = exp ψ∞ . (6.10)
ρ1 4M (3M − r1 )

Thus, the transformation x = 1/r has compactified the space. Moreover, the solution is finite
(i.e., the Jacobian functions involved in the solution are non-zero), since the complete integral K(k)
occurs for r < 0 and r < r1 i.e., outside the domain. Finally, the conformal factor is given by,
s 3
dn2 ψ (1 − cnψ)

3M
F (ρ) = 8 M r1 −1 · h√ √ √ √ i2 . (6.11)
r1
(1 + cnψ) M − 3M − r1 + ( M + 3M − r1 ) cnψ

As expected, the conformal factor F (ρ∞ ) → ∞, the requirement for conformal compactification
(see for example, Chapter 2.4.1 of Ref.[17] and references within). The index of refraction,
1/4
M r16 (1 + cnψ)2

n(ρ) = · , (6.12)
64 (3M − r1 )3 ρ snψ dnψ
shows that the projective equivalence no longer exist under transformation to isotropic coordinates.

11
6.1 Schwarzschild limit
As Λ → 0, H → 0 and from equations (6.2), (6.3) and (6.8)
 
ρ
r1 → 2M, k → 1, ψ → ln , (6.13)
ρ1

and since Jacobian elliptic functions turn to hyperbolic functions (see equation (5.17)), the Schwarzschild
solution is recovered. Moreover, asymptotic flatness yields ρ1 → ρH = M/2 and since K(k = 1) =
F(π/2, 1) → ∞, from equation (6.10), ρ∞ → ∞.

6.2 Refractive Index

Figure 2: The refractive indices are plotted as a function of ρ/ρ1 using the relationship (6.16)
The vertical dashed line represents the location where ρ = ρ∞ and horizontal, the refractive index
at ρ∞ . The asymptotic value of refractive index can be found, since the spacetime has only one
horizon and the space has been compactified. Another horizontal line, dot dashed, indicates unit
index of refraction. The parameter y = M 2 H 2 = −M 2 Λ/3, when less than one, gives rise to index
of refraction less than one. However, if the relationship (6.16) is not applied, then the index of
refraction may depend upon both M and H (rather than only on y) such that it is greater than
one.

12
 Once again, to plot the refractive index, we consider it as a function of z = ln(ρ/ρ1 ). Therefore,
from equation (6.12)
1/4
M r16 1 (1 + cnΨ)2

η(z) = · · , (6.14)
64 (3M − r1 )3 ρ1 ez snΨ dnΨ
where  1/4
4M (3M − r1 )
Ψ(z) = z . (6.15)
r12
Without loss of generality, if we consider
1/4
M r16

1
ρ1 = . (6.16)
2 64 (3M − r1 )3

then ρ1 → M/2 as Λ → 0, and

2 (1 + cnΨ)2
η(z) = . (6.17)
ez snΨ dnΨ
From equations (6.3) and (6.2), it is easy to see that Hr1 and in turn amplitude Ψ(z) and modulus
k depend upon a single parameter y = M 2 H 2 = −M 2 Λ/3. The refractive indices for different
values of y are plotted in figure 2.

7 Conclusion
For the Kottler spacetime, the static region exists between the two horizons when 0 < Λ < 1/9M 2 ,
and beyond a single horizon when Λ < 0. For these static regions, we have calculated the metric
in isotropic coordinates. For Λ < 0, the refractive index can be found asymptotically, as the space
is conformally compactified and the static region extends to spatial infinity.
In static coordinates, the spatially projected (affinely parameterized) null geodesics of a Lorentzian
spacetime are the (non-affinely parameterized) geodesics of the associated Riemannian Fermat met-
ric. As mentioned in the introduction, the parameterization changes due to the conformal transfor-
mations. The property that the unparameterized geodesics of two Riemannian metrics coincide is
called the projective equivalence. And a necessary condition for this to occur is that both metrics
share the identical projective curvature tensor. In isotropic static coordinates, we have shown that
the necessary and sufficient condition for the projective equivalence between two Fermat metrics
is that the ratio of refractive indices is constant. In other words, the projective equivalence in
isotropic coordinates is analogous to the invariance of Snell’s law in ordinary geometric optics.
The Nariai spacetime is a spherically symmetric static solution of the Einstein field equations
with cosmological constant and empty matter distribution, i.e., Rµν = Λgµν . The canonical form
for this spacetime does not exist; however, the line element in isotropic coordinates is given by [12],
 
2 1 2 2 1 2 2 2


ds = − (A1 cos(ln ρ) + A2 sin(ln ρ)) dt + 2 dρ + ρ dΩ . (7.1)
Λ ρ

As can be seen, an attempt to transform to the canonical form due to the presence of ρ−2 results in r
being constant. Since the null geodesics retain affine parameterization if the metric is multiplied by
a constant conformal factor [18], the affinely parameterized null geodesics of the Nariai metric (7.1)
are independent of Λ. It is interesting to note that the refractive indices of the Nariai, Schwarzschild
and Kottler spacetimes follow trigonometric, hyperbolic (equation (5.20)) and elliptic (Jacobian)
functions respectively with the natural log of the radial coordinate as the argument. It is known

13
that the Fermat metrics of the Kottler and Schwarzschild spacetimes in canonical coordinates are
projectively equivalent. We have shown that this projective equivalence is lost when transformed
to isotropic static coordinates which leads to spatially projected null geodesics that are different.
In other words, for the Kottler spacetime in canonical coordinates and Nariai in isotropic, both of
which satisfy the field equations Rµν = Λgµν , the spatial trajectories of affinely parameterized null
geodesics are independent of Λ.

Acknowledgement
I thank my supervisor, Dr. Mohammad Akbar, for his excellent course Theory of Black Holes and
for introducing me to the fascinating area of projective equivalence and gravitational lensing with
Noah Bright from Pittsburgh during his REU at UTD in summer 2020. I also thank him for his
guidance and sharing his insights from his work with Behshid Kasmaie on the subject. I am grateful
to Dr. Gary Gibbons for his encouraging and insightful comments. I thank Noah for stimulating
exchanges and for being my first mentee. Special thanks to Charlie Brewer for a careful reading
of the manuscript. I also thank the anonymous referees for their useful suggestions. I have been
partially supported by Julia Williams Van Ness Merit Scholarship and Margie Renfrow Student
Funds.

A Elliptic integrals and functions

This appendix lists some of the fundamental formulae related to elliptic integrals and functions
(see [4] for more details). The integral
Z h p i
I = R x, p(x) dx, (A.1)

is known as elliptic integral if p(x) is a polynomial of degree four or three with no multiple roots.
p
Here, R is a rational function of x and p(x). It was shown by Legendre that any elliptic integral
can be expressed linearly in terms of elementary functions and three fundamental integrals known
as the canonical elliptic integrals of first, second and third kind. The elliptic integral of the first
kind
dψ 0 dy 0
Z ψ Z y
F(ψ, k) = p = p (A.2)
0 1 − k 2 sin2 ψ 0 0 (1 − y 0 2 )(1 − k 2 y 0 2 )
is of prime importance to this paper. Here k is called the modulus (0 < k < 1) and ψ (or y) is
called the argument (0 < ψ ≤ π/2 or 0 < y ≤ 1). Moreover, y 0 = sin ψ 0 . When ψ = π/2 (or y = 1),
it is known as complete elliptic integral,
π
dψ 0
Z
2
K(k) = p . (A.3)
0 1 − k 2 sin2 ψ 0

The elliptic functions of Jacobi are obtained through inversion of elliptic integrals of the first kind.
ψ
dψ 0
Z
−1
am (ψ, k) = p = F(ψ, k) = u, (A.4)
0 1 − k 2 sin2 ψ 0
y
dy 0
Z
−1
sn (y, k) = p = F(sin−1 y, k) = u. (A.5)
0 2 2
(1 − y )(1 − k y ) 0 2
0

14
Thus, y = sin ψ = sn(u, k) and ψ = am(u, k). Here, sn(u, k) may be read sine amplitude u and
am(u, k), amplitude u. Moreover,
dy 0
Z 1 p
−1
cn (y, k) = p = F(sin−1 1 − y 2 , k), (A.6)
y (1 − y 0 2 )(K 2 + k 2 y 0 2 )
dy 0
Z 1  
−1
= F sin−1 (1 − y 2 )/k 2 , k , K ≤ y < 1 (A.7)
p
dn (y, k) = p
y (1 − y 0 2 )(y 0 2 − K 2 )
√
where K = 1 − k 2 is called the complementary modulus. Moreover, the range is given by

− 1 ≤ snψ ≤ 1, −1 ≤ cnψ ≤ 1, K ≤ dnψ ≤ 1, (A.8)

and special values,

am K = π/2, sn K = 1, cn K = 0, dn K = K, am 0 = 0, sn 0 = 0, cn 0 = 1, dn 0 = 1.
(A.9)
The Jacobian elliptic functions turn to hyperbolic when the modulus k = 1

sn(ψ, 1) = tanh ψ, cn(ψ, 1) = sechψ, dn(ψ, 1) = sechψ, (A.10)

and to trigonometric when it vanishes,

sn(ψ, 0) = sin ψ, cn(ψ, 0) = cos ψ, dn(ψ, 0) = 1. (A.11)

Additionally, differentiation with respect to the argument is given by,

∂ ∂ ∂ ∂
(amψ) = dnψ, (snψ) = cnψ dnψ, (cnψ) = −snψ dnψ, (dnψ) = −k 2 snψ cnψ.
∂ψ ∂ψ ∂ψ ∂ψ
(A.12)

B Roots of Cubic Equation

The roots x1 , x2 and x3 of the cubic equation

x3 + a2 x2 + a1 x + a0 = 0, (B.1)

are given by [3]

a2
x1 = S1 + S2 − , (B.2)
3 √
1 a2 i 3
x2 = − (S1 + S2 ) − + (S1 − S2 ) , (B.3)
2 3 2
√
1 a2 i 3
x3 = − (S1 + S2 ) − − (S1 − S2 ) , (B.4)
2 3 2
where
h
3

1 i1
2 2 3
h
3

1 i1
2 2 3
S1 = p + q + p , S2 = p − q + p , (B.5)
and
a1  a2 2 1  a 3
2
q= − , p= (a1 a2 − 3a0 ) − . (B.6)
3 3 6 3
Moreover, if

 > 0, then a real root and a complex conjugate pair.

q 3 + p2 = 0, then all roots are real and at least two are equal. (B.7)


< 0, then all roots are real.

15
C Light Rays in Isotropic Coordinates
The geodesics of the Fermat metric (3.2) can be derived from the Lagrangian,

n2 (ρ)  2 
L= ρ̇ + ρ2 θ̇2 + ρ2 sin2 θ φ̇2 , (C.1)
2
where ρ̇ = dρ/dλ and λ is affine parameter. The Euler-Lagrange equation in θ reduces to,
 
2 ṅ 1
θ̈ = φ̇ sin θ cos θ − 2 ρ̇ θ̇ + , (C.2)
n ρ

where ṅ = dn/dρ. For the initial conditions, θ = π/2 and θ̇ = 0, the equation above yields
θ̈ = 0. Thus, θ remains π/2 and the motion is confined to the equatorial plane. Since the metric
is rotationally invariant, we can consider motion in equatorial plane without loss of generality.
Therefore, the Euler-Lagrange equations in ρ and φ reduces to
   
ṅ 2 2 2 ṅ
ρ̈ + ρ̇ = ρ φ̇ + ρ φ̇2 , (C.3)
n n
 
ṅ 1
φ̈ + 2 ρ̇ φ̇ + = 0. (C.4)
n ρ
Since we are interested in unparameterized trajectory i.e., ρ = ρ(φ), from the chain rule,

dρ dρ dφ
= , (C.5)
dλ dφ dλ
and,
2
d2 ρ dρ d2 φ d2 ρ

dφ
= + . (C.6)
dλ2 dφ dλ2 dφ2 dλ
Substituting these relations into (C.3) and (C.4), upon simplification, yields
2 
d2 ρ
   
dρ ṅ 2 2 ṅ
= + +ρ+ρ , (C.7)
dφ2 dφ n ρ n

which is unparameterized spatial trajectory of null geodesics of metric (3.3) in equatorial plane.

References
[1] M. M. Akbar and B. Kasmaie, “Gravitational Lensing in Static Spherically Symmetric Space-
times: Symmetries and Equivalences,” forthcoming.

[2] M. A. Abramowicz, B. Carter and J. P. Lasota, “Optical Reference Geometry for Stationary
and Static Dynamics,” Gen. Relat. Gravit. 20, no. 11, pp. 1173–1183 (1988).

[3] M. Abramowitz and I. A. Stegun (1972) Handbook of Mathematical Functions with Formulas,
Graphs, and Mathematical Tables (United States National Bureau of Standards).

[4] P. F. Byrd and M. D. Friedman (1971) Handbook of Elliptic Integrals for Engineers and Sci-
entists (Second Edition, Revised, Springer-Verlag Berlin Heidelberg).

[5] S. Dodelson (2003) Modern Cosmology (Academic Press, Amsterdam).

16
 [6] L. P. Eisenhart (1926) Riemannian Geometry (Princeton University Press, Princeton).

[7] G. W. Gibbons, C. M. Warnick and M. C. Werner, “Light-bending in Schwarzschild-de-Sitter:

Projective Geometry of the Optical Metric,” Class. Quant. Grav. 25, 245009 (2008).

[8] J. Islam, “The Cosmological Constant and Classical Tests of General Relativity,” Phys. Lett.
A 97, 239-241 (1983).

[9] V. Lakshminarayanan, A. Ghatak, Ajoy, Thyagarajan, K. (2002) Lagrangian Optics (Springer

US).

[10] L. D. Landau and E. M. Lifschits (1975) The Classical Theory of Fields (Fourth Revised
English Edition, Elsevier).

[11] G. C. McVittie, “The mass-particle in an expanding universe,” Mon. Not. Roy. Astron. Soc.
93, 325-339 (1933).

[12] H. Nariai, “On a New Cosmological Solution of Einstein’s Field Equations of Gravitation,”
Gen. Relat. Gravit. 31, 963–971 (1999).

[13] V. Perlick, “Gravitational Lensing from a Spacetime Perspective,” Living Rev. Relativ. 7, 9
(2004)

[14] V. Perlick (2000) Ray optics, Fermat’s principle and applications to general relativity (Springer,
Heidelberg).

[15] P. Schneider, J. Ehlers and E. .E. Falco (1992) Gravitational Lenses (Springer, Berlin).

[16] J. L. Synge and A. Schild (1978) Tensor Calculus (Dover publications, New York).

[17] P. K. Townsend, “Black Holes: Lecture Notes,” arXiv:gr-qc/9707012

[18] R. M. Wald (1984) General Relativity (University of Chicago Press, Chicago).

17