You Can't Get Through Szekeres Wormholes or Regularity, Topology and Causality in Quasi-Spherical Szekeres Models
You Can't Get Through Szekeres Wormholes or Regularity, Topology and Causality in Quasi-Spherical Szekeres Models
You Can't Get Through Szekeres Wormholes or Regularity, Topology and Causality in Quasi-Spherical Szekeres Models
r
X
i
v
:
g
r
-
q
c
/
0
2
0
6
0
5
2
v
4
2
7
A
u
g
2
0
0
2
You Cant Get Through Szekeres Wormholes or
Regularity, Topology and Causality in
Quasi-Spherical Szekeres Models
Charles Hellaby
Department of Mathematics and Applied Mathematics,
University of Cape Town, Rondebosch 7701, South Africa
[email protected]
and
Andrzej Krasinski
N. Copernicus Astronomical Center, Polish Academy of Sciences,
Bartycka 18, 00 716 Warszawa, Poland
[email protected]
Abstract
The spherically symmetric dust model of Lematre-Tolman can describe wormholes,
but the causal communication between the two asymptotic regions through the neck is
even less than in the vacuum (Schwarzschild-Kruskal-Szekeres) case. We investigate the
anisotropic generalisation of the wormhole topology in the Szekeres model. The function
E(r, p, q) describes the deviation from spherical symmetry if
r
E = 0, but this requires
the mass to be increasing with radius,
r
M > 0, i.e. non-zero density. We investigate
the geometrical relations between the mass dipole and the locii of apparent horizon and
of shell-crossings. We present the various conditions that ensure physically reasonable
quasi-spherical models, including a regular origin, regular maxima and minima in the
spatial sections, and the absence of shell-crossings. We show that physically reasonable
values of
r
E = 0 cannot compensate for the eects of
r
M > 0 in any direction, so
that communication through the neck is still worse than in the vacuum.
We also show that a handle topology cannot be created by identifying hypersufaces in
the two asymptotic regions on either side of a wormhole, unless a surface layer is allowed
at the junction. This impossibility includes the Schwarzschild-Kruskal-Szekeres case.
PACS:
04.20.Gz, spacetime topology & causal structure
04.40.Nr, spacetimes with uids or elds
04.70.Bw, classical black holes
This research was supported by a grant from the South African National Research Foundation,
R
E
E
)
2
( + f)
dr
2
+ R
2
(dp
2
+ dq
2
)
E
2
, (1)
where
/r, = 1, 0 and f = f(r) is an arbitrary function of r.
The function E is given by
E(r, p, q) = A(p
2
+ q
2
) + 2B
1
p + 2B
2
q + C, (2)
where functions A = A(r), B
1
= B
1
(r), B
2
= B
2
(r), and C = C(r) satisfy the relation
4(AC B
2
1
B
2
2
) = , = 0, 1, (3)
but are otherwise arbitrary.
The function R = R(t, r) satises the Friedmann equation for dust
R
2
=
2M
R
+ f, (4)
where /t and M = M(r) is another arbitrary function of coordinate radius, r. It
follows that the acceleration of R is always negative
R =
M
R
2
. (5)
Here M(r) plays the role of an eective gravitational mass for particles at comoving radius r.
For = +1, it is simply the total gravitational mass within the sphere of radius r. We assume
M 0 and R 0. In (4) f(r) represents twice the energy per unit mass of the particles in
the shells of matter at constant r, but in the metric (1) it also determines the geometry of the
spatial sections t =constant (c.f. [10]). The evolution of R depends on the value of f; it can
be:
hyperbolic, f > 0:
R =
M
f
(cosh 1), (6)
(sinh ) =
f
3/2
(t a)
M
, (7)
1
The results presented in Ref. [24] contain a few misleading typos that were corrected in Ref. [7]. The
notation used here does not follow the traditional one
3
parabolic, f = 0:
R = M
2
2
, (8)
3
6
=
(t a)
M
, (9)
i.e. R =
_
9M(t a)
2
2
_
1/3
, (10)
elliptic, f < 0:
R =
M
(f)
(1 cos ), (11)
( sin ) =
(f)
3/2
(t a)
M
, (12)
where a = a(r) is the last arbitrary function, giving the local time of the big bang or crunch
R = 0 and = 1 permits time reversal. More correctly, the three types of evolution hold
for f/M
2/3
>, =, < 0, since f = 0 at a spherical type origin for all 3 evolution types. The
behaviour of R(t, r) is identical to that in the LT model, and is unaected by (p, q) variations.
A more meaningful way to write E is
E(r, p, q) =
S
2
_
_
p P
S
_
2
+
_
q Q
S
_
2
+
_
, (13)
where S = S(r), P = P(r), and Q = Q(r) are arbitrary functions, and
A =
1
2S
, B
1
=
P
2S
, B
2
=
Q
2S
, C =
P
2
+ Q
2
+ S
2
2S
. (14)
The metric component
(dp
2
+ dq
2
)
E
2
(15)
is actually the unit sphere, plane, pseudo-sphere in Riemann projection:
= +1
(p P)
S
= cot
_
2
_
cos() ,
(q Q)
S
= cot
_
2
_
sin(), (16)
= 0
(p P)
S
=
_
2
_
cos() ,
(q Q)
S
=
_
2
_
sin(), (17)
= 1
(p P)
S
= coth
_
2
_
cos() ,
(q Q)
S
= coth
_
2
_
sin(). (18)
It seems reasonable to expect S > 0, but it is not obviously impossible for S to reach or pass
through zero.
The factor determines whether the p-q 2-surfaces are spherical ( = +1), pseudo-spherical
( = 1), or planar ( = 0). In other words, it determines how the constant r 2-surfaces foliate
4
1
1
\theta
p P
S
1
p P
S
Figure 1: The Riemann projection from (, ) to (p, q) coordinates for spheres & two-
sheeted hyperboloids. The diagrams show only the = 0, section, i.e. the q = Q section.
the 3-d spatial sections of constant t. The function E determines how the coordinates (p, q)
map onto the unit 2-sphere (plane, pseudo-sphere) at each value of r. At each r these 2-
surfaces are multiplied by the areal radius R = R(t, r) that evolves with time. Thus the
r-p-q 3-surfaces are constructed out of a sequence of 2-dimensional spheres (pseudo-spheres,
planes) that are not concentric, since the metric component g
rr
depends on p and q as well as
r and t.
The (p, q)-coordinates in the cases = +1 and = 0 have the range (, +). In
the case = 1, the parametrization (13) does not cover the subcases A = 0 and C = 0
(these subcases cannot occur with 0 because of (3)). Coming back to (2), we see that,
for = 1 and A = 0, E is zero when
(p + B
1
/A)
2
+ (q + B
2
/A)
2
= 1/(4A)
2
,
E is positive for p and q outside this circle, and is negative for p and q inside it. Fig. 1 suggests
that with = 1, we should rather take (E) as the metric function so that p and q have nite
rather than semi-innite ranges. However, both the E > 0 and E < 0 regions are Szekeres
spacetimes because they are mapped one onto another by
(p, q) = (p
, q
)/
_
p
2
+ q
2
_
, (19)
the roles of A and C being interchanged after the transformation.
If A = 0 = C, then a nonzero C is restored by a translation in the (p, q) plane. If
A = 0, then the metric of the (p, q)-surface is brought back to the standard Szekeres form
with A = 0 = C by a Haantjes transformation (a conformal symmetry transformation of a at
space, see [16] for a description) in the (p, q) surface, which also restores the appropriate form
of g
rr
.
The surface area of a (t = const, r =const) surface is nite only in the = +1 case,
where it equals 4R
2
. In the other two cases, it is innite.
The 6 arbitrary functions f, M, a, P, Q and S represent 5 physical freedoms to control
the inhomogeneity, plus a coordinate freedom to rescale r.
5
The density and Kretschmann scalar are functions of all four coordinates
2
8 = G
tt
=
2(M
3ME
/E)
R
2
(R
RE
/E)
, (20)
K = R
= (8)
2
_
4
3
8
3
+ 3
2
_
, (21)
where
8 =
6M
R
3
(22)
is the mean density within radius r. For all and we have K 0, but assumptions of
positive mass and density require 0 and 0. Clearly there are density and curvature
singularities at R = 0 the bang and/or crunch and at R
= RE
/E, M
= 3ME
/E
shell crossings. Additionally, but not K passes through zero where E
/E exceeds M
/3M.
The matter ow u
t
, with projection tensor h
= g
+ u
=
(
R
3
RE
/E + 2R
R/R)
(R
RE
/E)
, (23)
a
= u
= 0, (24)
= g
(
(
u
)
+ u
(
a
)
)
3
h
=
(
R
R/R)
3(R
RE
/E)
diag(0, 2, 1, 1), (25)
=
[
u
]
+ u
[
a
]
= 0, (26)
E
= C
=
M(R
RM
/3M)
R
3
(R
RE
/E)
diag(0, 2, 1, 1), (27)
H
=
1
2
= 0. (28)
Note that the relation between the active gravitational mass M and the sum-of-rest-
masses M is the same in the = +1 Szekeres model as in the LT model:
M
= M
/
_
1 +f. (29)
The sum of the rest masses contained inside the sphere of coordinate radius r at the time t is
dened by:
M=
_
_
|g
3
|d
3
x, (30)
where g
3
is the determinant of the metric of the t = const hypersurface, and the integral is
taken with respect to the variables p and q from to +, and with respect to r from r
0
at the origin to the current value r. We have
2
There are only two independent curvature invariants in the Szekeres metric, for which a good choice
would be: R = 8 and C
= 48M
2
{(R
RM
/3M)
2
}/{R
6
(R
RE
/E)
2
} = (4/3)(8)(8
8)
2
a pure Ricci invariant and a pure Weyl invariant. Though less tidy, the 8 and K used above
will also suce. We thank a referee for pointing this out.
6
_
|g
3
| =
E
1 +f
(R/E)
2
(R/E)
. (31)
Consequently
M=
1
4
_
+
dq
_
+
dp
_
r
r
0
dx
_
E
1 +f
_
M/E
3
_
_
(t, p, q, x). (32)
The term containing E
1 +f
(r)
M
1 +f
(r
0
)
_
1
4
_
+
dq
_
+
dpE
2
+
1
8
_
+
dq
_
+
dp
_
r
r
0
dxE
2
_
3Mf
2(1 +f)
3/2
M
1 +f
_
. (33)
We note that
_
+
dq
_
+
dpE
2
= 4 (34)
(this is the surface area of a unit sphere), and so
M=
3
2
_
M
1 +f
(r)
M
1 +f
(r
0
)
_
+
1
2
_
r
r
0
_
3Mf
2(1 +f)
3/2
M
1 +f
_
dx. (35)
From here, we obtain the same relation that holds in the LT model, (29).
Note that this result holds only in the = +1 Szekeres model (the quasi-spherical one).
With = 0 or = 1, the total surface area of the (p, q)-surface is innite, and so M cannot
be dened.
2.1 Special Cases and Limits
The Lematre-Tolman (LT) model is the spherically symmetric special case = +1, E
= 0.
The vacuum case is (M
3ME
= 0 = S
= P
= Q
= E
(36)
and any region over which this holds is the Schwarzschild metric in LT coordinates [10], with
mass M. (See [12] for the full transformation in the general case.)
In the null limit, f , in which the dust particles move at light speed [11, 5], the
metric becomes a pure radiation Robinson-Trautman metric of Petrov type D, as given in Exact
Solutions [15], equation (24.60) with (24.62)
3
The Kinnersley rocket [14] is the = +1 case of
this null limit, which is actually more general than the axially symmetric form given in [11].
3
Ref [5] corrected ref [11]s mistaken claim that the null limit of Szekeres was a new metric.
7
The KS-type Szkeres metric was shown in [11] to be a special case of the above LT-type
metric, under a suitable limit.
2.2 Basic Physical Restrictions
1. For a metric of Lorentzian signature (+ ++), we require
+ f 0 (37)
with equality only occuring where (R
RE
/E)
2
/( +f) > 0. Clearly, pseudo-spherical
foliations, = 1, require f 1, and so are only possible for hyperbolic spatial sections,
f > 0. Similarly, planar foliations, = 0, are only possible for parabolic or hyperbolic
spatial sections, f 0, whereas spherical foliations are possible for all f 1.
2. We obviously choose the areal radius R to be positive,
R 0 (38)
(R = 0 is either an origin, or the bang or crunch. In no case is a continuation to negative
R possible.)
3. The mass M(r) must be positive, so that any vacuum exterior has positive Schwarzschild
mass,
M 0. (39)
4. We require the metric to be non degenerate & non singular, except at the bang or crunch.
Since (dp
2
+ dq
2
)/E
2
maps to the unit sphere, plane or pseudo-sphere, |S(r)| = 0 is
needed for a sensible mapping, and so S > 0 is a reasonable choice. In the cases = 0
or 1, E necessarily goes to zero at certain (p, q) values where the mapping is badly
behaved. For a well behaved r coordinate, we do need to specify
>
(R
RE
/E)
2
( + f)
> 0, (40)
i.e. ( + f) > 0 except where (R
RE
/E)
2
= 0. (41)
In Lematre-Tolman models [19, 26] (E
)
2
can occur in closed models where the areal radius on a spatial section is at a maximum,
or in wormhole models where the areal radius is minimum, R
(t, r
m
) = 0, t. These
can only occur at constant r and must hold for all (p, q) values. We will consider maxima
and minima again later.
5. The density must be positive, and the Kretschmann scalar must be nite, which adds
either M
3ME
/E 0 and R
RE
/E 0 (42)
or M
3ME
/E 0 and R
RE
/E 0. (43)
If (R
RE
change sign?
E
=
1
2
S
_
(p P)
2
+ (q Q)
2
_
/S
2
+
_
1
S
[(p P)P
+ (q Q)Q
]. (44)
The discriminant of this with respect to (p P) is
p
=
1
S
2
_
S
2
S
2
(q Q)
2
2
S
S
(q Q)Q
+ P
2
+ S
2
_
. (45)
The discriminant of
p
with respect to (q Q) is
q
= 4
S
2
S
6
(P
2
+ Q
2
+ S
2
). (46)
Since, with = +1, this is never negative, the equation E
q
= 0. They are:
1. S
= 0. Then E
= P
= Q
= 0. Then E
_
Q
_
P
2
+ Q
2
+ S
2
_
. (47)
9
p
q
p1 p2
q1
q2
E > 0
E > 0
E > 0
E > 0
E < 0
E > 0
E > 0
E > 0
E > 0
Delta p < 0
Delta p > 0
Delta p < 0
Figure 2: When S
> 0, E
> 0 outside.
Note that the (p, q) plane is a Riemann projection of a sphere, and on the sphere inside
and outside are topologically equivalent.
For every q such that q
1
< q < q
2
there will be two values of p (and one value of p when q = q
1
or q = q
2
) such that E
S
S
_
q Q
S
+
Q
_
2
+
P
2
+ Q
2
S
2
+ . (48)
The regions where E
. If S
> 0, then E
> 0
for p < p
1
and for p > p
2
, if S
< 0, then E
> 0 for p
1
< p < p
2
. E
= 0 for p = p
1
and
p = p
2
, but note that p
1
and p
2
are members of a continuous family labelled by q. All the
values of p and q from (47) (48) lie on the circle
_
p
_
P P
S
S
__
2
+
_
q
_
QQ
S
S
__
2
= S
2
_
P
2
+ Q
2
S
2
+
_
. (49)
The center of this circle is in the point
(p, q) =
_
P P
S
S
S
S
_
, (50)
and the radius of this circle is
L
E
=0
= S
P
2
+ Q
2
S
2
+ . (51)
The situation on the (p, q)-plane when S
=
S
cos + sin (P
cos + Q
sin )
1 cos
, (53)
E
=
S
cos + sin (P
cos + Q
sin )
(1 cos )
+2
_
S
S
__
S
cos + sin (P
cos + Q
sin )
(1 cos )
_
+
(S
)
2
+ (P
)
2
+ (Q
)
2
S
. (54)
The locus E
= 0 is
S
cos + P
sin cos + Q
z +P
x +Q
cos + Q
sin
. (56)
The plane has unit normal (P
, Q
, S
)/
_
(P
)
2
+ (Q
)
2
+ (S
)
2
.
Now it is easy to understand the meaning of the special case S
= 0 passes through the pole of Riemann projection. In this case, the image of the circle
E
= 0 on the (p, q) plane is a straight line passing through (p, q) = (P, Q), as indeed follows
from (44). The sign of E
is dierent on each side of the straight line. Compare also with Figs.
4 and 5.
From (53) and (52) we nd
E
E
=
S
cos + sin (P
cos + Q
sin )
S
(57)
thus
E
E
= constant S
z + P
x + Q
y = S constant (58)
which is a plane parallel to the E
/E)
=
sin (P
sin Q
cos )
S
= 0 (59)
11
tan
e
=
Q
cos
e
=
1
P
_
(P
)
2
+ (Q
)
2
,
1
= 1, (60)
(E
/E)
=
S
sin P
cos cos Q
cos sin
S
= 0 (61)
tan
e
=
P
cos
e
+ Q
sin
e
S
=
1
_
(P
)
2
+ (Q
)
2
S
(62)
cos
e
=
2
S
_
(S
)
2
+ (P
)
2
+ (Q
)
2
,
2
= 1. (63)
The extreme value is then
_
E
E
_
extreme
=
2
_
(S
)
2
+ (P
)
2
+ (Q
)
2
S
. (64)
Since (sin
e
cos
e
, sin
e
sin
e
, cos
e
) =
2
(P
, Q
, S
)/
_
(P
)
2
+ (Q
)
2
+ (S
)
2
, eq. (55)
shows that the extreme values of E
/E has a dipole variation around each constant r sphere, changing sign when
we go over to the anitipodal point: (, ) ( , + ). Writing
_
R
R
E
E
_
=
_
R
+ R
S
cos + sin (P
cos + Q
sin )
S
_
(67)
we see that RE
of neighbouring constant
r shells, due to their not being concentric. In partuicular RS
/S is the forward ( = 0)
displacement, and RP
/S & RQ
/E is maximum.
It will be shown in section 5.2 that, where R
> 0, E
/E M
/(3M) and E
/E R
/R
are required to avoid shell crossings, and also in eq (130) that R
/R > M
/3M. These
inequalities, together with M
/E:
=
2M
R
2
R
1 3Mx/M
1 Rx/R
(68)
has a negative derivative by x:
,
x
=
R/R
3M/M
(1 Rx/R
)
2
2M
R
2
R
< 0, (69)
12
and so the density is minimum where E
/E is maximum.
The density, eq. (20), can be decomposed into a spherical part and a dipole-like part, as
noted by Szekeres [25] and de Souza [9] (see also p. 30 in Ref.[17]). Rewriting de Souzas
result into our notation of eqs. (1) and (2), we obtain:
=
s
+ , (70)
where
s
is the spherical part:
s
=
2M
(A+ C) 6M(A
+ C
)
R
2
[R
(A+ C) R(A
+ C
)]
, (71)
and is the dipole-like part:
=
A
+ C
(A+ C)E
/E
R
RE
/E
6MR
2M
R
R
2
[R
(A+ C) R(A
+ C
)]
. (72)
The dipole-like part changes sign on the surface where E
/E = (A
+ C
/E) = (E
= Q
= S
= 0 (= A
= C
/E = (A
+C
)/(A+C) = const,
is a circle parallel to the great circle E
+ C
+ C
)R
max/min
=
S(S
+ P
+ Q
) [(S
)
2
+ (P
)
2
+ (Q
)
2
](2S
+
2
_
(S
)
2
+ (P
)
2
+ (Q
)
2
)
S(S
2
_
(S
)
2
+ (P
)
2
+ (Q
)
2
)
, (73)
while around the E
= 0 circle
E
=
(S
)
2
+ (P
)
2
+ (Q
)
2
S
S
(P
cos + Q
sin ) S
(P
cos + Q
sin )
_
(P
cos + Q
sin )
2
+ (S
)
2
(P
cos + Q
sin )
. (74)
4 Regular Origins
When = +1, R = 0 occurs at an origin of spherical coordinates, e.g. R(t, 0) = 0, t,
where the 2-spheres have no size. Similarly,
R(t, 0) = 0 =
R(t, 0), etc t. There will be a
second origin, at r = r
O
say, in any closed, regular, f < 0 model. Thus, by (11) and (6) and
their combinations with (4) & (5), for each constant
lim
r0
M
f
= 0 , lim
r0
f = 0 , lim
r0
f
2
M
= 0. (75)
13
The type of time evolution at the origin must be the same as its neighbourhood, i.e., along
a constant t slice away from the bang or crunch, by (12) and (7),
0 < lim
r0
|f|
3/2
(t a)
M
< . (76)
Clearly, we need M 0, f 0 and
0 < lim
r0
|f|
3/2
M
< . (77)
Using lHopitals rule, this gives
lim
r0
3Mf
2M
f
= 1. (78)
The density and Kretschmann scalar must be well behaved. We dont consider a vacuum
region of nite size at the origin, as that is just Minkowski space, M = 0, and we dont consider
the obscure case of a single vacuum point at the origin. Because & in (21) evolve dierently,
we also need
0 < lim
r0
6M
R
3
= lim
r0
2M
R
2
R
< lim
r0
3R
M
RM
= 1 (79)
and
0 < lim
r0
2(M
3ME
/E)
R
2
(R
RE
/E)
= lim
r0
2M
R
2
R
(1 3ME
/M
E)
(1 RE
/R
E)
< , (80)
but in fact the latter is ensured by the former, and the anisotropic eect of E vanishes at the
origin. However, since E
/M
E or lim
r0
RE
/R
=
M
2
(f)
3
_
1
3Mf
2M
f
_
sin ( sin )(1 cos )
+
M
2
(f)
3
_
1
Mf
f
_
(1 cos )
3
M
2
a
(f)
3/2
M
< . (82)
Lastly, the metric must be well behaved, so E should have no unusual behaviour, such as
S = 0, that would compromise a valid mapping of (dp
2
+ dq
2
)/E
2
to the unit sphere. Also,
to ensure the rate of change of proper radius with respect to areal radius is that of an origin,
g
rr
/(R
)
2
should be nite
0 < lim
r0
(R
RE
/E)
2
(1 +f)(R
)
2
< (83)
0 < lim
r0
_
1
3ME
E
_
2
< lim
r0
ME
< (84)
and lim
r0
ME
=
1
3
, (85)
14
where the last of (79) has been used. This should hold for all (p, q), i.e. all (, ). Thus (57)
gives
lim
r0
MS
< , lim
r0
MP
< , lim
r0
MQ
< , (86)
all three limits being dierent from 1/3.
All of the above suggests that, near an origin,
M R
3
, f R
2
, S R
n
, P R
n
, Q R
n
, n 0. (87)
The condition E
/E M
/3M that will be obtained in the next section implies that near an
origin
n 1 (88)
5 Shell Crossings
5.1 Occurrence and Position of Shell Crossings in a Surface of Con-
stant t and r.
A shell crossing, if it exists, is the locus of zeros of the function R
RE
RE
= P
= Q
= R
= 0. Since
P, Q and S depend only on r, this means they are constant throughout the spacetime. As
seen from (1) and (20), the Szekeres metric reduces then to the LT metric, and so this case
need not be considered.
Suppose that R
RE
= Q
= S
= R
(t, r
0
) will vanish for all t. This will either be a singularity (when M
(r
0
) = 0) or a
neck (when M
(r
0
) = 0), familiar from the studies of the LT model, see Refs. [10] and [13].
Hence, R
RE
> 0 and R
RE
> ER
RE
< 0 and
R
RE
RE
/E < 0.
Assuming R
> 0, can R
RE
/R E
=
1
2S
_
S
S
+
R
R
_
_
(p P)
2
+ (q Q)
2
_
1
2
S
_
S
S
R
R
_
+
1
S
[(p P)P
+ (q Q)Q
], (89)
the discriminants of this with respect to (p P) and (q Q) are
p
=
P
2
S
2
1
S
2
_
S
S
+
R
R
___
S
S
+
R
R
_
(q Q)
2
+ (q Q)Q
1
2
_
S
S
R
R
__
(90)
15
q
= 4
1
S
2
_
S
S
+
R
R
_
2
_
P
2
+ Q
2
+ S
2
S
2
R
2
R
2
_
. (91)
Thus ER
/R E
p
< 0 for all q). Hence, ER
/RE
has the same sign for all p and q (i.e. there are no shell
crossings) if and only if
R
2
R
2
>
P
2
+ Q
2
+ S
2
S
2
:=
2
(r). (92)
Note that when = 0, this can fail only at those points where R
= 0.
If R
2
/R
2
=
2
, then
q
= 0, and so
p
= 0 at just one value of q = q
SS
. At this value
of q, ER
/RE
= 0 at one value of p = p
SS
. In this case, the shell crossing is a single point
in the constant (t, r)-surface, i.e. a curve in a space of constant t and a 2-surface in spacetime.
If R
2
/R
2
<
2
, then the locus of ER
/RE
/S = R
/R) in the (p, q) plane. The straight line is just a projection onto
the (p, q) plane of a circle on the sphere of constant t and r, and so is not really any special
case.
When
q
> 0 (R
2
/R
2
<
2
), the two limiting values of q at which
p
changes sign are
q
1,2
=
Q
_
P
2
+ Q
2
+ (S
2
S
2
R
2
/R
2
)
S
/S + R
/R
, (93)
and then for every q such that q
1
< q < q
2
, there are two values of p (only one if q = q
1
or
q = q
2
) such that ER
/R E
= 0. These are
p
1,2
=
P
__
S
S
+
R
R
_
(q Q) Q
_
2
+ P
2
+ Q
2
+ (S
2
S
2
R
2
/R
2
)
S
/S + R
/R
. (94)
The values of p and q from (93) and (94) lie on the circle with the center at
(p
SC
, q
SC
) =
_
P
P
/S + R
/R
, Q
Q
/S + R
/R
_
, (95)
and with the radius L
SC
given by
L
SC
2
=
P
2
+ Q
2
+ (S
2
S
2
R
2
/R
2
)
(S
/S + R
/R)
2
. (96)
This is in general a dierent circle than the one dened by E
/E = R
/R =
const. As noted after eq. (58), this line is a circle that lies in a plane parallel to the E
= 0
great circle. It follows immediately that the E
= 0 = B
1
= B
2
= C
there. (3)
To keep the density non-singular, 0 ((S
)
2
+(P
)
2
+(Q
)
2
)/S
2
< min((R
/R)
2
, (M
/3M)
2
),
which is a no shell crossing condition. We shall improve on the latter below.
For positive density, (20) shows that (M
3ME
/E) & (R
RE
3ME
/E) 0
and (R
RE
3ME
/E) & (R
RE
/3M = R
/R, but the latter cannot hold for all time. This case
can only hold for all (p, q) if M
= 0, E
= 0, R
, f
, a
, S
, P
,
Q
3ME
3M
E
max
=
_
(S
)
2
+ (P
)
2
+ (Q
)
2
S
r. (97)
It is obvious that this is sucient, and also that
M
0 r. (98)
We will now consider (R
RE
R
=
M
M
(1
4
) +
f
f
_
3
2
4
1
_
f
3/2
a
M
5
, (99)
where
4
=
sinh (sinh )
(cosh 1)
2
,
5
=
sinh
(cosh 1)
2
. (100)
At early times,
0, (101)
R
M
f
2
2
+ O(
4
) 0, (102)
5
4
3
+ O() +, (103)
4
2
3
+ O(
2
)
2
3
, (104)
we nd
5
dominates and
R
R
f
3/2
a
M
5
, (105)
17
so that (R
RE
< 0 r. (106)
Similarly, at late times,
, R ,
5
0 ,
4
1, (107)
we nd
5
vanishes and
R
R
1
2
f
f
, (108)
so that
_
R
R
E
E
_
> 0
f
2f
E
E
> 0. (109)
Following the above analysis of (M
3ME
/E) 0 we obtain
f
2f
>
_
(S
)
2
+ (P
)
2
+ (Q
)
2
S
r, (110)
which obviously implies
f
> 0 r. (111)
Again, since we already have M
> 0. (112)
5.2.2 Parabolic evolution, f = 0
The easiest way to obtain the conditions for this case, f = 0, f
= 0, is to put = /
f > 0
in the hyperbolic case, and take the limit f 0, 0. All terms involving f
/f cancel and
we retain exactly the same conditions, viz (106) & (111) (and of course (97)). Naturally, (110)
ceases to impose any limit.
5.2.3 Elliptic evolution, f < 0
For elliptic models, we can write:
R
R
=
M
M
(1
1
) +
f
f
_
3
2
1
1
_
(f)
3/2
a
M
2
, (113)
where
1
=
sin ( sin )
(1 cos )
2
,
2
=
sin
(1 cos )
2
. (114)
At early times,
0, (115)
R
M
(f)
2
2
+ O(
4
) 0, (116)
2
4
3
+ O() +, (117)
1
2
3
+ O(
2
)
2
3
, (118)
18
we nd
2
dominates and
R
R
f
3/2
a
M
2
, (119)
so that R(R
/R E
< 0 r. (120)
Similarly, at late times,
2, (121)
R
M
(f)
(2 )
2
2
+ O((2 )
4
) 0, (122)
2
4/(2 )
3
+ O((2 )) , (123)
1
8/(2 )
3
+ 2/3 +O((2 )) , (124)
we nd
R
R
M
M
_
8
(2 )
3
_
f
_
12
(2 )
3
_
+
(f)
3/2
a
M
_
4
(2 )
3
_
, (125)
so that R
3/2
(R
/R E
M
3f
2f
_
+ a
> 0 r, (126)
which is the condition that the crunch time must increase with r. Since we already have
M
> 0 (127)
for all .
We now show the above also ensure R(R
/R E
R
=
M
3M
+
b
(b a)
_
2
3
1
_
+
(a
)
(b a)
_
2
3
1
+ 2
2
_
. (129)
The derivative of (2/3
1
) is (2 3 sin + cos )/(1 cos )
2
, and the third derivative of
the numerator of the latter is sin . It follows that (2/3
1
) 0 and declines monotonically
from + to 0 as goes from 2 to 0. Since (2/3
1
+ 2
2
) is the mirror image in =
of (2/3
1
), we have that
R
R
>
M
3M
, (130)
so that (97) guarantees that for each given r, the maximum of E
/R as varies.
19
Note that although (126) implies
f
2f
<
M
3M
, (131)
a condition such as (110) is not needed in this case. As an indication of the approximate
magnitude of R
/R|
min
, at the moment of maximum expansion along any given worldline,
R
R
=
M
M
f
f
, (132)
so it would be possible to have E
/E|
max
close to R
/R|
min
around the time of maximum
expansion.
6 Regular Maxima & Minima
Certain topologies necessarily have extrema in R. For example, closed spatial sections have a
maximum areal radius, and wormholes have a minimum areal radius, i.e. R
(t, r
m
) = 0, t.
Suppose ( + f) = 0 at some r = r
m
. By (41) we must have
f
(r
m
) = 0 (133)
(unless f
RE
/E) = 0 to
keep g
rr
nite, and hence (M
3ME
RE
/E)
+ f
L , 0 < L < , (134)
(M
3ME
/E)
(R
RE
/E)
N , 0 N < . (135)
As noted above, we require
M
= f
= a
= S
= P
= Q
= 0 (136)
to ensure
R
= 0. (137)
The limits (134) and (135) must hold good for all t and for all (p, q), so using (13), (99) &
(113) with R > 0, M > 0, S > 0 shows that
M
+ f
,
f
+ f
,
a
+ f
,
R
+ f
, (138)
S
+ f
,
P
+ f
,
Q
+ f
,
E
+ f
(139)
must all have nite limits, that do not have to be zero. Using lHopitals rule, each of the
above limits can be expressed in the form
L
M
= lim
f1
M
+ f
=
2M
+ f
f
=
2M
L
f
. (140)
20
Thus, for f = 1, the above condidtions for no shell crossings in elliptic regions should be
re-expressed in terms of these limits.
It is worth pointing out that E
RE
/E)
+ f
, (141)
K
pp
=
R( + f)
1/2
E
2
, (142)
K
qq
=
R( + f)
1/2
E
2
. (143)
Now at an extremum in R, the factor (R
RE
RE
) = K
ij
(+n
) K
+
ij
= K
ij
, (144)
which is only possible if
f = . (145)
If however, r
m
is only a shoulder i.e. R
(r
m
) = 0, but R
| > 0.
6.1 Summary: Conditions for No Shell Crossings or Surface Layers
The conditions found here are exactly those on M, f & a for LT models (see [13] which
generalises those of [1] for a = 0 LT models), with extra conditions involving S, P, Q also.
21
R
f M
, f
, a
, P
, Q
+1 > 0 all M
(S
)
2
+(P
)
2
+(Q
)
2
S
M
3M
+1 > 0 0 f
(S
)
2
+(P
)
2
+(Q
)
2
S
f
2f
a
M
3f
2f
_
+ a
0
a
0
but not all 3 equalities at once
+1 = 0 1 M
= 0, f
= 0, a
= 0 S
= 0, P
= 0, Q
= 0
R
M
3f
2f
_
+ a
(S
)
2
+(P
)
2
+(Q
)
2
S
M
3M
a
0
+1 = 0 1 M
= 0, f
= 0, a
= 0 S
= 0, P
= 0, Q
= 0
R
M
3f
2f
_
+ a
(S
)
2
+(P
)
2
+(Q
)
2
S
M
3M
a
0
+1 < 0 all M
(S
)
2
+(P
)
2
+(Q
)
2
S
M
3M
+1 < 0 0 f
(S
)
2
+(P
)
2
+(Q
)
2
S
f
2f
a
M
3f
2f
_
+ a
0
a
0
but not all 3 equalities at once
7 Impossibility of a Handle Topology
Since the function E has the eect of making the distance between adjacent constant r shells
depend on angle, this allows us to create a wormhole that is bent, so that the two asymptotic
world sheets on either side can be thought of as intersecting in the embedding.
This leads to the question of whether those two world sheets can be smoothly joined across
a junction surface. In fact the possibility of matching the two world sheets together can be
considered independently of whether there is a natural embedding that would allow them to
intersect at the appropriate angles.
Thus we investigate whether it is possible to create a Szekeres model with a handle topology
in the following way. Take a wormhole topology an = +1 model with r = 0, f(0) = 1
at the wormhole & f < 0 nearby and let it be mirror symmetric about r = 0. Choose a
comoving open surface on one side of the wormhole, and its mirror image on the other side,
and match the two sheets together along , as shown schematically in Fig. 3. Because the 2
sheets are mirror images, this is equivalent to matching to its own mirror image.
22
Figure 3: Conceptual illustration of joining a Szekeres wormhole model to itself across a
boundary surface , shown as a heavy line. The boundary may be close to the wormhole, as
shown, or out in the asymptotic regions. There is no signicance to the change from solid to
dotted circles, other than picture clarity. The handle topology is shown as an embedding of
a constant time section, with one angular coordinate suppressed, although a valid matching
across does not require the result to have a natural embedding. However, it is shown that
the matching fails because it is not preserved by the model evolution.
To implement this we choose a comoving surface:
r
= Z(p, q) (146)
and surface coordinates:
i
= (t, p, q). (147)
The two fundamental forms and the normal are calculated in appendix B.
Obviously 1st fundamental forms match by construction, and normal vectors are equal and
opposite:
n
+
= n
. (148)
In fact the 2 surfaces
+
&
. Thus K
ij
= K
+
ij
, so
the only way to match the 2nd fundamental forms, K
+
ij
= K
ij
, is to make them zero:
K
ij
= 0. (149)
The K
pt
= K
pt
& K
qt
= K
qt
equations give
R
= 0 r
= Z = constant. (152)
23
If the matching surface is at constant r, then only a closed torus topology is possible. So the
answer is: no, a handle topology is not possible.
Suppose
R = 0 possible, then it should be possible to solve
K
pp
= 0 , K
pq
= 0 , K
qq
= 0 (153)
for Z(p, q), by specifying suitable functions E(r, p, q) & R(r) on an initial time slice. In other
words, you can probably match on a constant time slice, but the matching is not preserved by
the model evolution.
8 Szekeres Wormholes?
It has been shown in [10] that LT models can describe the Schwarzschild-Kruskal-Szekeres
manifold, as well as models that have the same topology but non-zero density. It has also
been shown that the matter ows from past to future singularity, with possibly some matter
escaping to J
+
or some being captured from J
= (1) (k
t
)
2
+
_
R
R
E
E
_
2
+ f
(k
r
)
2
+
R
2
E
2
((k
p
)
2
+ (k
q
)
2
) (155)
_
R
R
E
E
_
2
+ f
_
dr
dt
_
2
= 1
R
2
E
2
_
_
_
dp
dt
_
2
+
_
dq
dt
_
2
_
_
. (156)
24
It is obvious that at each event dr/dt is maximised by choosing k
p
= 0 = k
q
. Since R is
independent of (p, q), this also gives the direction of maximum dR/dt|
null
at any event. We
will call this radial motion, and radial null paths rays. Thus, the DE
t
n
=
dt
dr
n
=
j
1 +f
_
R
RE
E
_
, j = 1 (157)
in principle solves to give
t = t
n
(r) (158)
along the ray. We dont expect this to be geodesic, but we regard it as the limit of a sequence
of accelerating timelike paths, and thus the boundary to possible motion through a wormhole.
The acceleration of this path may be calculated from a
= k
, as given in appendix C.
8.2 Apparent Horizons
The areal radius along a ray is
R
n
= R(t
n
(r), r), (159)
(R
n
)
=
Rt
n
+ R
(160)
= j
1 +f
_
R
RE
E
_
+ R
(161)
= j
_
2M
R
+ f
1 +f
_
R
RE
E
_
+ R
, = 1. (162)
These rays are momentarily stationary when
(R
n
)
= 0. (163)
Now light rays initially along constant p and q will not remain so, owing to the anisotropy of the
model. However, since these radial directions are at each point the fastest possible escape
route, we dene this locus to be the apparent horizon (AH).
(Szekeres [25] dened a trapped surface as the locus where null geodesics that are (mo-
mentarily) radial have zero divergence, k
;
= 0, where k
;
k
= 0, k
= 0, k
p
= 0 = k
q
.
He obtained
k
;
=
2
R
_
R
RE
E
__
R
1 +f
+ j
_
. (164)
Given the anisotropy of the model, one doesnt expect this to be the same locus as our AH.)
Assuming a normal spacetime point will have non-zero metric components, and taking R
increasing with r on constant t slices,
R
> 0 and
_
R
RE
E
_
> 0, (165)
we require
j = 1, (166)
25
i.e.
Either (future AH: AH
+
) j = +1 (outgoing rays)
= 1 (in a collapsing phase) (167)
Or (past AH: AH
) j = 1 (incoming rays)
= +1 (in an expanding phase). (168)
Note that we want outgoing to mean moving away from the neck at r = 0. A ray passing
through the neck would change from incoming to outgoing at r = 0, and, since R
ips sign
there, j would also have to ip there.
8.2.1 The Apparent Horizon and its Location with Respect to E
= 0.
Dene
D :=
_
1 +f
2M
R
+ f. (169)
Then
(D > 0) (R > 2M). (170)
Since M/R and (2M/R+f) are positive, we see that D 1 leads to a contradiction, and so
D < 1. (171)
However, |D| can be greater than 1 because D < 1 is not prohibited. We have
(D < 1) =
_
R <
M
1 +
1 +f
_
. (172)
This will always occur when R is close to the Big Bang/Big Crunch.
Using D, the equation of the AH is
RE
+ DR
E = 0, (173)
and in terms of p and q this equation is
_
S
S
D
R
R
_
_
(p P)
2
+ (q Q)
2
_
+ 2[(p P)P
+ (q Q)Q
] S
2
_
S
S
+ D
R
R
_
= 0.
(174)
The discriminant of this with respect to p is
p
= 4P
2
4
_
S
S
D
R
R
___
S
S
D
R
R
_
(q Q)
2
+ 2(q Q)Q
S
2
_
S
S
+ D
R
R
__
.
(175)
The discriminant of this with respect to q is
q
= 64
_
S
S
D
R
R
_
2
_
P
2
+ Q
2
+ S
2
_
S
2
S
2
D
2
R
2
R
2
__
. (176)
26
Now, if
q
< 0 everywhere, then
p
< 0 for all q, in which case there is no p obeying
(174), i.e. the apparent horizon does not intersect this particular surface of constant (t, r).
If
q
= 0, then
p
< 0 for all q except one value q = q
0
, at which
p
= 0. At this value
of p = p
0
, (174) has a solution, and so the intersection of the apparent horizon with this one
constant (t, r) surface is a single point. Note that the situation when the apparent horizon
touches the whole 3-dimensional t = const hypersurface at a certain value of t is exceptional,
this requires, from (174), that P
= Q
= S
= R
_
P
2
+ Q
2
+ S
2
_
S
2
S
2
D
2
R
2
R
2
_
S
/S DR
/R
(177)
and then a solution of (174) exists given by
p
1,2
=
P
__
S
S
D
R
R
_
(q Q) + Q
_
2
+ P
2
+ Q
2
+ S
2
_
S
2
S
2
D
2
R
2
R
2
_
S
/S DR
/R
. (178)
Except for the special case when S
/S = DR
/S DR
/R
, Q
Q
/S DR
/R
_
, (179)
and with the radius L
AH
given by
L
AH
2
=
P
2
+ Q
2
+ S
2
_
S
2
S
2
D
2 R
2
R
2
_
(S
/S DR
/R)
2
. (180)
The special case S
/S = DR
> 0 we have
(D > 0) =(E
< 0)
(D < 0) =(E
= 0 circle and the AH cannot intersect unless they coincide. Indeed, these circles
lie in parallel planes, by the same argument that was used at the end of sec. 5.1: the line on
the (t, r) = const surface dened by (173) has the property E
/E = DR
/R = const, and
so it must be a circle in a plane parallel to the E
= 0 and AH circles are disjoint, they may either be one inside the other or
each one outside the other. However, when projected back onto the sphere, these two situations
turn out to be topologically equivalent: depending on the position of the point of projection,
the same two circles may project onto the plane either as one circle inside the other or as two
separate circles, see Figs. 4 and 5.
8.2.2 Location of the AH Compared with R = 2M
Along R = 2M
(R
n
)
= R
(1 +j)
RE
E
, (182)
so R = 2M is not the AH except where E
= 0.
28
C1
C2
P1
P2
Figure 5: The same circles as in Fig. 4 projected onto a plane from a dierent pole will
project as one inside the other. The transition from the situation of Fig. 4 to that of Fig.
5 is continuous and occurs when the sphere is rotated, but the pole and the plane are not
moved. Then one of the circles (C1 when a clockwise rotation is applied to Fig. 4) will
pass through the pole at one value =
0
of the rotation angle. Its image on the plane is
acquiring a larger and larger radius with increasing , until it becomes a straight line when
=
0
. When increases further, the straight line bends in the opposite direction so that
it surrounds the second circle P2.
Eq (162) with (163) and j = 1 can be written
R
AH
=
2M(1 RE
/R
E)
2
1 +f[1 (1 RE
/R
E)
2
]
(183)
=
2M(1 V )
2
1 +f(2V V
2
)
, V = RE
/R
E. (184)
The eect of E(r, p, q) is to create a dipole in the geometry and density around each (t, r)
shell, with E
extreme
=
_
(S
)
2
+ (P
)
2
+ (Q
)
2
S
(185)
at the poles.
It is clear then that radial displacements between two nearby surfaces of constant r are
shortest where E
/E is maximum, and light rays move outwards fastest (max dr/dt, min dt/dr,
i.e. most rapid transfer between constant r shells at the same (p, q) value). It has also been
shown that the density is minimum here. The longest radial displacements, slowest light ray
motion and maximum density occur at the opposite pole.
We will call the direction where E
/E is
minimum, the slow pole.
Now the conditions for no shell crossings require
E
extreme
<
M
3M
(186)
29
and for an elliptic region we have
M
3M
<
R
R
, (187)
so
V
2
< 1 , (1 V )
2
> 0 , 3 < (2V V
2
) < 1. (188)
In those places on the AH, where V = 0 = E
= 0, R 0 and R
> 0?
As noted in [10] the introduction of matter into a wormhole slows the progress of light rays
through it. Can this eect be compensated for by a suitable choice of E
= 0 = E
n
=
1
1 +f
_
_
M(1 cos )
(f)
_
1 +f)
_
_
_
_
sin
(1 cos )
2
_
(f)
3/2
a
_
1
3 sin( sin )
2(1 cos )
2
_
f
f
+
_
1
sin ( sin )
(1 cos )
2
_
M
M
E
E
_
. (192)
Consider a region in which R
> 0 and M
/E is always negative. In
particular, because of the no shell crossing condition (97), |E
/E| M
/3M, the E
/E term
gives at most a partial cancellation of the M
/E
cannot compensate for the eect of non-zero M
_
(f)
(1 +f)
sin
(1 cos )
_
_
_
(f)
3/2
a
M
_
sin
(1 cos )
2
_
f
_
1
3 sin( sin )
2(1 cos )
2
_
+
M
M
_
1
sin ( sin )
(1 cos )
2
__
_
_
_
(f)
(1 +f)
sin
(1 cos )
_
_
E
E
. (194)
The solution is the parametric locus =
AH
(r). It is evident that if E
= 0, varying M
has
no eect at all on the AH locus for a given M, as the solution is
_
(f)
(1 +f)
sin
(1 cos )
= 1 cos = 1 + 2f R = 2M. (195)
Similarly E
R=M
=
R
_
2
+ f
, (196)
31
where > 0. This is null or outgoing timelike wherever
(R
ME
/E)
1 +f
R
_
2
+ f
. (197)
For M
= 0, which forces E
/E < M
/3. (198)
For any given M
> 0, R
> 0 and f > 1, this is always greater than the numerator on the
rhs, so, to satisfy the equality, the denominator on the lhs must be greater than that on the
rhs, so once again
> 2. (199)
Thus R = M surfaces can only be tangent to outgoing null rays for R = M > 2M, and for
R < 2M they are spacelike, incoming null, or incoming timelike.
This allows us to conclude that, along the entire length of the future R = 2M surface,
outgoing rays pass inside it, or run along it where M
1 +f
_
R
RE
E
__
Rdt +
R
dr
dr
2(1 +f)
+
_
R
dt + R
dr
(
Rdt+R
dr)E
E
RE
dr
E
+
R(E
)
2
dr
E
2
_
_
R
RE
E
_
_
+
R
dt + R
dr, (201)
32
dt
dr
AH
=
_
j
1 +f
_
R
RE
E
_
j
Rf
2(1 +f)
3/2
_
R
RE
E
_
+j
1 +f
_
R
E
RE
E
+
R(E
)
2
E
2
_
+ R
__
_
j
1 +f
_
R
RE
E
_
+ j
1 +f
_
R
RE
E
_
+
R
_
. (202)
At the neck of the wormhole, r = r
n
, the regularity conditions of sec. 6 give us the
following limits, where L
f
etc are being dened in each case:
f 1, (203)
f
0, f
/
_
1 +f L
f
= 2f
/L
f
0 (L
f
)
2
= 2f
, (204)
E
0, E
/
_
1 +f L
E
= 2E
/L
f
0, (205)
R
0, R
/
_
1 +f L
R
= 2R
/L
f
> 0, (206)
0,
R
/
_
1 +f L
R
= 2
R
/L
f
= (/t)L
R
. (207)
Thus the terms
R
1 +f ,
R
RE
/E
1 +f , R
/E
1 +f , R(E
)
2
/E
2
1 +f and
AH,N
= jR
__
2
R
L
f
_
R
RE
E
_
+
1 +f
_
R
RE
E
__
, (208)
dt
dr
n,N
= j
2(R
RE
/E)
L
f
. (209)
Since the AH only intersects the neck when
R = 0, the behaviour of
R/
1 +f and
R
2
/
will be veried).
At the moment of maximum expansion in the neck we have
R = 2M ,
R = 0 ,
_
1 +f = 0, (210)
= , cos = 1 , sin = 0, (211)
and so, using
R
=
M
(f)
__
M
M
f
f
_
(1 cos )
_
M
M
3f
2f
_
sin ( sin )
(1 cos )
_
a
_
f
sin
(1 cos )
, (212)
R
=
1
(1 cos )
2
M
(f)
_
( sin )
_
M
M
3f
2f
_
+
a
(f)
3/2
M
_
2
+
sin
(1 cos )
_
( sin )
M
(f)
_
2f
f
_
M
M
9f
8f
_
_
M
M
3f
2f
__
_
f
_
a
+ a
f
__
+ (1 cos )
M
(f)
_
M
M
f
f
2f
f
_
M
M
f
f
__
(213)
33
gives, by virtue of (204) and (140):
L
R
= 2(L
M
+ ML
f
) = 4(M
+ Mf
)/L
f
, R
= 2(M
+ Mf
). (214)
We nd the limit of
R/
1 +f
MEN
=
jL
R
L
R
RL
E
/E
=
j(M
+ Mf
)
M
+ Mf
ME
/E
, (215)
so it is clear that
R
2
E
=
1
R
_
M
R
MR
R
2
+
f
2
_
R
R
1 +f
=
_
L
M
R
ML
R
R
2
+
L
f
2
_
, (216)
and because of (214) and R = 2M all terms in the bracket cancel, verifying that
R
R
1 +f =
0.
These together with (5) give us
dt
dr
AH,MEN
= j
4Mf
(M
+ Mf
)
L
f
(M
+ Mf
ME
/E)
, (217)
dt
dr
n,MEN
= j
4(M
+ Mf
ME
/E)
L
f
. (218)
For a light ray to pass through the neck at the moment of maximum expansion without
falling inside the AH, we need dt/dr|
AH,MEN
> dt/dr|
n,MEN
, in other words
dt/dr|
AH,MEN
dt/dr|
n,MEN
=
Mf
(M
+ Mf
)
(M
+ Mf
ME
/E)
2
MEN
> 1, (219)
or
(M
+ Mf
)
_
Mf
(M
+ Mf
)
M
<
E
E
<
(M
+ Mf
) +
_
Mf
(M
+ Mf
)
M
.
(220)
Since M(r) is positive, and both M(r) and f(r) are at a minimum at the neck, i.e. M
> 0
& f
> 0, we have
_
Mf
(M
+ Mf
) < M
+ Mf
= 0, f = 1, R
> 0
E
max
3M
. (221)
To be able to satisfy this as well as (220) we would need
M
+ Mf
_
Mf
(M
+ Mf
)
M
<
M
3M
, (222)
34
but this leads to
M
(4M
+ 3Mf
) < 0, (223)
which is clearly not possible. Indeed, although is zero rather than divergent where E
/E =
M
/3M, where E
/E exceeds M
/E = M
max
=
9Mf
(M
+ Mf
)
(2M
+ 3Mf
)
2
, (224)
which rises from 0 at f
/M
/M
, i.e.
vacuum.
Therefore, even at the neck, E
to inside AH
+
.
8.3.3 Summary
In a Szekeres wormhole, every particle worldline encounters R = 2M, twice for most r values
and once where f = 1, making this a pair of 3-surfaces that span the spacetime. The
apparent horizons coincide with R = 2M at an extremum of R(t = const, r) a neck or belly
where f = 1. Where M
0), then the event horizon is the set of rays that are asymptotic to R = 2M,
but always lie outside. The future event horizon EH
+
emerges from R = 2M surface, and
vice-versa for EH
.
Thus we conclude that the causal structure of a regular Szekeres wormhole is only a quan-
titative modication of the LT wormhole (dense black hole), and the possible causal diagrams
for Szekeres models are essentially the same as those for LT models, as given in [10].
8.3.4 Numerical Examples
A few numerical examples were produced as follows.
We choose the 3 LT arbitrary functions to produce a Kruskal-like topology, with the neck
at r = 0, that is mirror symmetric about r = 0 and t = 0. The choice must therefore satisfy
f(0) = 1, f
(0) = 0, f
(0) > 0, M
(0) = 0, M
/E becomes
E
max
=
(230)
along the direction (p, q) = (0, 0), i.e. = 0. Since numerical integrations will only be done
along this path and the = one, we treat E as a function of r only. We make E
/E as large
as possible without violating the no shell crossings condition E
/E M
/3M with
E = E
0
(1 +E
1
r
2
) + E
2
, E
0
, E
1
, E
2
> 0, (231)
where the shell crossing occurs somewhere if E
2
= 0.
Because of the two reection symmetries, we can start integrating a null ray from maximum
expansion at the neck,
= , r = 0, (232)
where AH
+
& AH
/f
M
0
= 1 , M
1
= 1 , E
0
= 1 , E
1
= 1 , E
2
= 0 , r
s
= 1. (236)
We found that the rays & AHs were well split, while the rays were strongly trapped.
36
Run 2 low M
/f
M
0
= 1 , M
1
= 1 , E
0
= 1 , E
1
= 1 , E
2
= 0.1 , r
s
= 0.01. (237)
The rays were mildly split, the AHs were indistinguishable in the range plotted, and the
rays were mildly trapped.
Run 3 slightly less low M
/f
M
0
= 1 , M
1
= 2 , E
0
= 1 , E
1
= 2 , E
2
= 0.01 , r
s
= 0.01. (238)
This was very similar to the previous run, with the rays less mildly trapped.
Run 4 high M
/f
M
0
= 1 , M
1
= 3 , E
0
= 1 , E
1
= 3 , E
2
= 0.0001 , r
s
= 10. (239)
Here the rays & AHs were well split, and the rays were very strongly trapped.
These examples cover the main possibilites, and run 1 is shown in Fig. 6.
9 Conclusions
Szekeres (S) models are a generalisation of the spherically symmetric Lematre-Tolman (LT)
models. Both describe inhomogeneous dust distributions, but the former have no Killing vectors.
There are 3 arbitrary functions of coordinate radius in LT models (M, f & a), and a further 3
in S models (S, P & Q).
For quasi-spherical Szekeres (S) models, we established 3 sets of regularity conditions
the conditions for a regular origin, the conditions for no shell crossings, and the conditions for
regular maxima and minima in the spatial sections. The last two contain exactly those for the
LT models, with extra conditions on the arbitrary functions that are peculiar to S. Thus, for
every regular LT model that is non-vacuum (M
= 0, S models
must be spherically symmetric.)
Since LT models can reproduce the Schwarzschild-Kruskal-Szekeres topology of a wormhole
connecting two universes, but with non-zero density everywhere, this is also possible with
S models. In the vacuum case (M
/E = (A
+ C
/E at constant r is bounded (see (64)), it must be veried whether eq. (240) has a
solution in every sphere of constant t and r. The solution will exist when
(E
/E)
min
(A
+ C
)/(A+ C) (E
/E)
max
. (241)
Since (E
/E)
min
= (E
/E)
max
, eq. (241) is equivalent to
(A
+ C
)
2
/(A+ C)
2
(E
/E)
2
extreme
=
1
S
2
_
P
2
+ Q
2
+ S
2
_
. (242)
We have
A+ C =
1
2S
_
1 +P
2
+ Q
2
+ S
2
_
A
+ C
=
S
2S
2
_
S
2
P
2
Q
2
1
_
+
1
S
(PP
) . (243)
Substituted in (242), this leads to
4S
2
S
2
_
1 +P
2
+ Q
2
_
4SS
(PP
)
_
S
2
P
2
Q
2
1
_
4S
2
(PP
)
2
+
_
P
2
+ Q
2
_ _
1 +P
2
+ Q
2
+ S
2
_
0. (244)
The discriminant of this with respect to S
is
= 16S
2
_
1 +P
2
+ Q
2
+ S
2
_
2
_
(PQ
QP
)
2
+ P
2
+ Q
2
_
, (245)
40
and is always negative unless P
= Q
, Q
= Q
= 0 as
well. However, P
= Q
= S
= 0 implies A
= C
= 0 and E
= Z(p, q) (246)
and surface coordinates,
i
= (t, p, q) (247)
we calculate the basis vectors in the surface,
e
i
=
x
i
, (248)
the 1st fundamental form,
3
g
+
ij
=
3
g
ij
= g
i
e
j
, (249)
the normal vector,
n
, n
= 1 , n
i
= 0, (250)
and the 2nd fundamental form
K
ij
= n
_
2
x
j
+
i
x
j
_
. (251)
Using GRTensor/GRJunction [20] we nd the following for the intrinsic metric:
3
g
tt
= 1, (252)
3
g
pp
=
Z
2
p
(R
RE
/E)
2
E
2
+ R
2
( + f)
E
2
( + f)
, (253)
3
g
pq
=
Z
p
Z
q
(R
RE
/E)
2
( + f)
, (254)
3
g
qq
=
Z
2
q
(R
RE
/E)
2
E
2
+ R
2
( + f)
E
2
( + f)
, (255)
41
the surface basis vectors:
e
t
i
= (1, 0, 0), (256)
e
r
i
= (0, Z
p
, Z
q
), (257)
e
p
i
= (0, 1, 0), (258)
e
q
i
= (0, 0, 1), (259)
the surface normal:
n
r
=
(R
E RE
)R
E
, (260)
n
p
=
Z
p
(R
E RE
)R
E
, (261)
n
q
=
Z
q
(R
E RE
)R
E
, (262)
where Z
p
=
Z
p
, Z
q
=
Z
q
, (263)
=
_
R
2
( + f) + (Z
2
p
+ Z
2
q
)(R
E RE
)
2
_
1/2
, (264)
and the extrinsic curvature:
K
pt
=
[Z
p
(R
R
RR
)]
, (265)
K
qt
=
[Z
q
(R
R
RR
)]
, (266)
K
pp
=
1
2E
2
( + f)
_
2RE(R
E RE
)( + f)Z
pp
+2(R
E RE
)
2
(E
p
E
p
E)Z
3
p
+2(R
E RE
)
2
(E
q
E
q
E)Z
2
p
Z
q
+
_
2
_
3ER
RE
+ RR
E
2
(E
)
2
R
2
2E
2
(R
)
2
R
2
E
E
_
( + f)
Rf
E(R
E RE
)
_
Z
2
p
2R(2RE
p
E E
p
R
E RE
p
E
)( + f)Z
p
2RE
q
(R
E RE
)( + f)Z
q
2R
2
( + f)
2
_
, (267)
K
pq
=
1
2E
2
( + f)
_
2RE(R
E RE
)( + f)Z
pq
+2(R
E RE
)
2
(E
p
E
p
E)Z
2
p
Z
q
+2(R
E RE
)
2
(E
q
E
q
E)Z
p
Z
2
q
+
_
2
_
3ER
RE
+ RR
E
2
(E
)
2
R
2
2E
2
(R
)
2
R
2
E
E
_
( + f)
42
Rf
E(R
E RE
)
_
Z
p
Z
q
2RE(RE
q
R
E
q
)( + f)Z
p
2RE(RE
p
R
E
p
)( + f)Z
q
_
, (268)
K
qq
=
1
2E
2
( + f)
_
2RE(R
E RE
)( + f)Z
qq
+2(R
E RE
)
2
(E
p
E
p
E)Z
p
Z
2
q
+2(R
E RE
)
2
(E
q
E
q
E)Z
3
q
+
_
2
_
3ER
RE
+ RR
E
2
(E
)
2
R
2
2E
2
(R
)
2
R
2
E
E
_
( + f)
Rf
E(R
E RE
)
_
Z
2
q
2RE
p
(R
E RE
)( + f)Z
p
2R(2RE
q
E E
q
R
E RE
q
E
)( + f)Z
q
2R
2
( + f)
2
_
, (269)
where all quantities are evaluated on .
C The Acceleration of a Given Tangent Vector
Starting from
a
= k
= k
, (270)
the individual acceleration components for a given k
k
t
+
t
rr
(k
r
)
2
+
t
pp
(k
p
)
2
+
t
qq
(k
q
)
2
(271)
= k
t
t
k
t
+ k
r
r
k
t
+ k
p
p
k
t
+ k
q
q
k
t
+
_
R
RE
E
__
R
RE
E
_
1
+ f
(k
r
)
2
+
_
R
R
E
2
_
_
(k
p
)
2
+ (k
q
)
2
_
, (272)
a
r
= k
k
r
+ 2
r
tr
k
t
k
r
+
r
rr
(k
r
)
2
+ 2
r
rp
k
r
k
p
+ 2
r
rq
k
r
k
q
+
r
pp
(k
p
)
2
+
r
qq
(k
q
)
2
(273)
= k
t
t
k
r
+ k
r
r
k
r
+ k
p
p
k
r
+ k
q
q
k
r
+
2
_
R
RE
E
_
_
R
RE
E
_
k
t
k
r
+
_
_
_
R
RE
E
_
_
R
RE
E
_
f
2( + f)
_
_ (k
r
)
2
2R
_
E
p
E
E
Ep
E
2
_
_
R
RE
E
_
k
r
k
p
2R
_
E
q
E
E
Eq
E
2
_
_
R
RE
E
_
k
r
k
q
R( + f)
E
2
_
R
RE
E
_
_
(k
p
)
2
+ (k
q
)
2
_
, (274)
43
a
p
= k
k
p
+ 2
p
tp
k
t
k
p
+
p
rr
(k
r
)
2
+ 2
p
rp
k
r
k
p
+
p
pp
(k
p
)
2
+ 2
p
pq
k
p
k
q
+
p
qq
(k
q
)
2
(275)
= k
t
t
k
p
+ k
r
r
k
p
+ k
p
p
k
p
+ k
q
q
k
p
+
2
R
R
k
t
k
p
_
R
RE
E
_
(EE
p
E
E
p
)
R( + f)
(k
r
)
2
+
2
_
R
RE
E
_
R
k
r
k
p
E
p
E
(k
p
)
2
2E
q
E
k
p
k
q
+
E
p
E
(k
q
)
2
, (276)
a
q
= k
k
q
+ 2
q
tq
k
t
k
q
+
q
rr
(k
r
)
2
+ 2
q
rq
k
r
k
q
+
q
pp
(k
p
)
2
+ 2
q
pq
k
p
k
q
+
q
qq
(k
q
)
2
(277)
= k
t
t
k
q
+ k
r
r
k
q
+ k
p
p
k
q
+ k
q
q
k
q
+
2
R
R
k
t
k
q
_
R
RE
E
_
(EE
q
E
E
q
)
R( + f)
(k
r
)
2
+
2
_
R
RE
E
_
R
k
r
k
q
+
E
q
E
(k
p
)
2
2E
p
E
k
p
k
q
E
q
E
(k
q
)
2
. (278)
For radial paths k
p
= 0 = k
q
,
p
k
= 0 =
q
k
these reduce to
a
t
= k
t
t
k
t
+ k
r
r
k
t
+
_
R
RE
E
__
R
RE
E
_
1
( + f)
(k
r
)
2
, (279)
a
r
= k
t
t
k
r
+ k
r
r
k
r
+
2
_
R
RE
E
_
_
R
RE
E
_
k
t
k
r
+
_
_
_
R
RE
E
_
_
R
RE
E
_
f
2( + f)
_
_ (k
r
)
2
, (280)
a
p
=
_
R
RE
E
_
(EE
p
E
E
p
)
R( + f)
(k
r
)
2
, (281)
a
q
=
_
R
RE
E
_
(EE
q
E
E
q
)
R( + f)
(k
r
)
2
. (282)
Using the radial null condition
k
t
=
jk
r
+ f
_
R
RE
E
_
(283)
44
the acceleration becomes
a
t
=
j
+ f
__
R
RE
E
_
k
t
k
r
+
_
R
RE
E
_
k
t
t
k
r
_
+
j
+ f
___
R
RE
E
_
2( + f)
_
R
RE
E
__
(k
r
)
2
+
_
R
RE
E
_
k
r
r
k
r
_
+
_
R
RE
E
__
R
RE
E
_
1
( + f)
(k
r
)
2
, (284)
a
r
= k
r
r
k
r
+
j
+ f
__
R
RE
E
_
k
r
t
k
r
+ 2
_
R
RE
E
_
(k
r
)
2
_
+
_
_
_
R
RE
E
_
_
R
RE
E
_
f
2( + f)
_
_ (k
r
)
2
, (285)
a
p
=
_
R
RE
E
_
(EE
p
E
E
p
)
R( + f)
(k
r
)
2
, (286)
a
q
=
_
R
RE
E
_
(EE
q
E
E
q
)
R( + f)
(k
r
)
2
. (287)
While E
S
_
_
(S
)
2
+ (P
)
2
+ (Q
)
2
S
_ (288)
q
e
= Q +
Q
S
_
_
(S
)
2
+ (P
)
2
+ (Q
)
2
S
_ (289)
and it is easily veried that a
p
= 0 = a
q
in these two antipodal directions. It follows that
initially radial geodesics in these directions remain radial if p
e
& q
e
are constant with r. For
example, if p
e
= 0 = q
e
, this would require arbitrary functions satisfying
P
P
=
2S
S
S
2
P
2
Q
2
=
Q
Q
(290)
or P = 0 ,
Q
Q
=
2S
S
S
2
Q
2
, when P
= 0 (291)
or
P
P
=
2S
S
S
2
P
2
, Q = 0 , when Q
= 0 (292)
or P = 0 , Q = 0 , when P
= 0 and Q
= 0 (293)
45
D Other Features of the AH
The FLRW Case
The dust FLRW limit is M = M
0
r
3
, f = kr
2
, a = 0, R = rS(t), and the R = 2M locus is
given by
S(t
AH
) = 2M
0
r
2
(294)
In the collapse phase of a k = +1 model, the time of the future AH is
cos
AH
= 1 2r
2
(295)
t
AH
= M
0
_
+ arccos(2r
2
1) + 2r
1 r
2
_
(296)
which has slope
_
dt
dr
_
AH
=
4M
0
r
2
1 r
2
(297)
while the light rays have slopes
_
dt
dr
_
n
=
S
1 r
2
=
2M
0
r
2
1 r
2
(298)
Clearly the future AH is incoming timelike. The result extends to all k values, and the converse
holds for the past AH in the expansion phase.
Behaviour Near the Bang or Crunch
Consider eq (194) for the locus of the future AH in a collapsing elliptic region, f < 0, <
2, in terms of parameter . Near the crunch, = 2 0, we nd
0 =
_
_
1
_
(f)
(1 +f)
_
2
_
_
_
_
(f)
3/2
a
M
_
4
3
_
f
_
12
3
_
+
M
M
_
8
3
__
+
_
_
_
(f)
(1 +f)
_
2
_
_
_
E
E
. (299)
As noted previously, when E
f 0.
(Even in this case, where we know R = 2M is the AH, the fact that R
diverges at R = 0
means we must multiply through by to make the rhs zero there.) Notice too that the no shell
crossing condition (126) ensures the second bracket is generically non-negative where M
> 0
and non-positive where M
3
6
4M
3
(300)
where the crunch time b(r) is dened in (128). Therefore the future AH does not intersect the
crunch away from an origin. The result is just the time reverse for the past AH near the bang
46
in an expanding region, and a similar calculation applies for hyperbolic or extended parabolic
regions, giving the same result.
An f = 0 locus is where an interior elliptic region joins to an exterior hyperbolic or parabolic
region. The transition involves the lifetime of the worldlines diverging, so either the crunch
goes to the innite future, or the bang goes to the innite past. Since there is only one AH
in a hyperbolic or parabolic region, one of the two AH loci in the elliptic region also exits to
innity before f = 0 is reached.
A third possibility where f is only asymptotically zero (the asymptotically at case) is that
both the bang and the crunch diverge to the innite past and future, and the two AHs go with
them.
Behaviour Near an Origin
Consider eq (194) again. Near a regular origin, along a constant t or constant surface (see
section 4), M (f)
3/2
, E (f)
n/2
for some positive constants & , and f 0, so
that
0 =
_
1 +
_
f
sin
(1 cos )
_ _
_
sin
(1 cos )
2
_
+
f
2f
_
_
_
f
sin
(1 cos )
_
nf
2f
. (301)
We divide through by f
and dene X =
f
sin
(1cos )
which is positive for > , giving
0 = [1 X]
_
_
sin
(1 cos )
2
_
+
1
2f
+
X
1 X
n
2f
_
. (302)
Though a
and M
3
6
0. (304)
As always, the time reverse applies in an expanding phase, and the hyperbolic and parabolic
cases give the same result.
47