Prepared for submission to JCAP

arXiv:1701.05533v1 [hep-th] 19 Jan 2017

Weyl metrics and wormholes

Gary W. Gibbons,a,b Mikhail S. Volkovb,c
a

DAMTP, University of Cambridge, Wilberforce Road, Cambridge CB3 0WA, UK

Laboratoire de Mathematiques et Physique Theorique, LMPT CNRS UMR 7350,
Universite de Tours, Parc de Grandmont, 37200 Tours, France
c
Department of General Relativity and Gravitation, Institute of Physics,
Kazan Federal University, Kremlevskaya street 18, 420008 Kazan, Russia
b

E-mail: gwg1@cam.ac.uk, volkov@lmpt.univ-tours.fr

Abstract. We study solutions obtained via applying dualities and complexifications to the vacuum Weyl metrics generated by massive rods and by point masses.
Rescaling them and extending to complex parameter values yields axially symmetric
vacuum solutions containing conical singularities along circles that can be viewed as
singular matter sources. These solutions have wormhole topology with several asymptotic regions interconnected by throats and their sources can be viewed as thin rings
of negative tension encircling the throats. For a particular value of the ring tension
the geometry becomes exactly flat although the topology remains non-trivial, so that
the rings literally produce holes in flat space. To create a single ring wormhole of one
metre radius one needs a negative energy equivalent to the mass of Jupiter. Further
duality transformations dress the rings with the scalar field, either conventional or
phantom. This gives rise to large classes of static, axially symmetric solutions, presumably including all previously known solutions for a gravity-coupled massless scalar
field, as for example the spherically symmetric Bronnikov-Ellis wormholes with phantom scalar. The multi-wormholes contain infinite struts along the symmetry axes,
apart from solutions with locally flat geometry.

Contents
1 Introduction

2 Gravitating scalar field

2.1 Spherically symmetric sector
2.1.1 Solutions with scalar field
2.1.2 Solutions with phantom
2.1.3 Wormholes
2.2 Anti-gravitating solutions with phantom field

5
6
6
7
7
8

3 Axially symmetric Weyl metrics and their symmetries

3.1 Symmetries
3.1.1 Sign flips and scaling
3.1.2 Rotations and boosts
3.1.3 Swap symmetry
3.1.4 Tachyon symmetry

8
9
9
10
10
10

4 Vacuum Weyl metrics

4.1 One-rod solution
4.2 Two-rod solution
4.3 The Chazy-Curzon metrics

11
11
12
12

5 Solutions from one rod

5.1 Prolate vacuum metrics
5.2 Oblate vacuum metrics
5.2.1 Wormhole topology
5.2.2 Ring wormholes
5.3 Solutions with scalar field

13
13
14
16
19
21

6 Solutions from two rods

6.1 Multi wormholes
6.2 Locally flat wormholes
6.3 Solutions with scalar field

23
25
28
30

7 Solutions from point masses

31

8 Conclusions

32

A Isometric embeddings of the BE wormhole

33

Introduction

Wormholes are bridges or tunnels between different universes or different parts of

the same universe. They were first introduced by Einstein and Rosen (ER) [1], who
noticed that the Schwarzschild black hole actually has two exterior regions connected
by a bridge. The ER bridge is spacelike and cannot be traversed by classical objects,
but it has been argued that it may connect quantum particles to produce quantum
entanglement and the Einstein-Pololsky-Rosen (EPR) effect [2], hence ER=EPR [3].
Wormholes were also considered as geometric models of elementary particles handles
of space trapping inside an electric flux, say, which description may indeed be valid
at the Planck scale [4]. Wormholes can also describe initial data for the Einstein
equations [5] (see [6] for a recent review) whose time evolution corresponds to the
black hole collisions of the type observed in the recent GW150914 event [7].
An interesting topic is traversable wormholes globally static bridges accessible
for ordinary classical particles or light [8] (see [9] for a review). In the simplest case
such a wormhole is described by a static, spherically symmetric line element
ds2 = Q2 (r)dt2 + dr2 + R2 (r)(d2 + sin2 d2 ),

(1.1)

where Q(r) and R(r) are symmetric under r r and R(r) attains a non-zero global
minimum at r = 0. If both Q and R/r approach unity as r then the metric
describes two asymptotically flat regions connected by a throat of radius R(0). The
Einstein equations G = T imply that the energy density = T00 and the radial
pressure p = Trr satisfy at r = 0
+ p = 2

R00
< 0,
R

p=

1
< 0.
R2

(1.2)

It follows that for a static wormhole to be a solution of the Einstein equations, the Null
Energy Condition (NEC), T v v = R v v 0 for any null v , must be violated.
Another demonstration [8] of the violation of the NEC uses the Raychaudhuri equation
[10] for a bundle of light rays described by , , : the expansion, shear and vorticity.
In the spherically symmetric case one has = = 0 [9], hence
d
1
= R v v 2 .
d
2

(1.3)

If rays pass through a wormhole throat, there is a moment of minimal cross-section

area, = 0 but d/d > 0, hence R v v < 0 and the NEC is violated.
If the spacetime is not spherically symmetric then the above arguments do not
apply, but there are more subtle geometric considerations showing that the wormhole
throat a compact two-surface of minimal area can exist if only the NEC is violated
[11, 12]. As a result, traversable wormholes are possible if only the energy density
becomes negative, for example due to vacuum polarization [8], or due to exotic matter
as for example phantom fields with a negative kinetic energy [13, 14]. Otherwise,
one can search for wormholes in the alternative theories of gravity, as for example in

the Gauss-Bonnet theory [15], in the brainworld models [16], in theories with nonminimally coupled fields [17], or in massive (bi)gravity [18].
The best known wormhole solutions were found in 1973 in the theory of gravitycoupled phantom scalar field [13, 14]. Their geometry is ultrastatic, ds2 = dt2 +dl2 ,
where the 3-metric dl2 = ik dxi dxk satisfies equations
(3)

R ik () = 2i k ,

= 0,

(1.4)

the phantom nature of being encoded in the negative sign in the right hand side of
the gravitational equations. The simplest solution of these equations is the spherically
symmetric Bronnikov-Ellis (BE) wormhole [13, 14],
 
x
2
2
2
2
2
2
2
,
(1.5)
dl = dx + (x + )(d + sin d) ,
= arctan

where is an integration constant determining the size of the wormhole throat. More
general solutions of (1.4) were obtained by Clement [19] choosing the 3-metric to be
in the axially symmetric Weyl form [20],
dl2 = e2k (d2 + dz 2 ) + 2 d2 ,
where k and depend on , z. Equations (1.4) then reduce to


1

2
2
Dk = (D) ,

+ 2 = 0,

(1.6)

(1.7)

where D = + iz . Since the -equation is linear, superposing its elementary onewormhole solutions gives solutions describing multi-wormholes [19, 21] (they can be
generalized to include also rotation and Maxwell field [22, 23]).
It is worth noting that equations (1.4) actually coincide with the vacuum Einstein
equations for the everywhere non-singular 5-metric [24]
ds25 = cos(2)[dx20 + dx24 ] + 2 sin(2)dx0 dx4 + dl2 .

(1.8)

Therefore, wormhole solutions of (1.4) can be interpreted as 5-geometries without

invoking the negative energy phantom1 .
Wormholes are usually associated with exotic matter and it is little known that
they exist also in General Relativity (GR) as classical solutions of vacuum Einstein
equations. For such solutions the metric is Ricci flat everywhere apart from geometric
sets of zero measure that can be interpreted as singular ring-shaped matter sources
whose effective energy-momentum tensor has the structure needed to provide the NEC
violation. Since these solutions are vacuum everywhere apart from singularities, they
can be constructed with the standard GR methods. The first example of this type
was discovered in 1966 by Zipoy who studied oblate vacuum metrics of the axially
1

The BE solution (1.5), when lifted to 5D, becomes a member of spherically symmetric vacuum
solutions of Chodos and Deweiler [25].

symmetric Weyl type [20] and found solutions which have a ring singularity and have
a double sheeted topology; one can get from one sheet to the other by going through
the ring [26]. The two sheets are the two asymptotically flat regions connected by
the wormhole throat the hole encircled by the ring. As Zipoy noted, this hole should
have strange properties: an object falling through it would not be seen coming out
from the other side, whereas if viewed from above the ring it could be seen dropping
through [26]. The ring singularity can be interpreted in this case as a cosmic string
loop with the negative tension [27], [28], [29], [30],
T =

(1 + 2 )c4
,
4G

(1.9)

where is a free parameter. Interestingly, if = 0 then the metric becomes flat

everywhere outside the ring, the latter thus literally creating a hole in flat space [30].
Vacuum ring wormholes may seem to be very different from the BE wormholes
supported by the phantom field. However, they are actually described by the same
solution. Specifically, the ring metric is
ds2 = e2U dt2 + e2U dl2
with dl2 given by (1.6), and the vacuum Einstein equations read in this case


U
2U
1
2

+
= 0.
Dk = + (DU ) ,

z 2

(1.10)

(1.11)

Now, comparing with (1.7) it is clear that solutions of the two sets of equations are
related by
U ,

k k.

(1.12)

Therefore, ring wormholes and phantom wormholes are actually described by the same
equations and to obtain ones from the others it is enough to interchange the phantom
with the Newtonian potential U and to flip the sign of k. As a result, to study
wormholes one can use the ordinary vacuum GR without invoking phantom fields or
extra dimensions. The price to pay is the metric singularity supported by the ring
which provides the necessary NEC violation.
In this paper we study solutions obtained via applying dualities and complexifications to the vacuum Weyl metrics. As a starting point we use the elementary metrics
generated by rods and by point masses. Rescaling them and extending to the complex
parameter values yields the prolate and oblate vacuum solutions. Further duality transformations produce a scalar field, which can be either conventional or phantom. This
gives rise to large classes of static, axially symmetric solutions, presumably including
all previously known solutions for a gravity-coupled massless scalar field. Especially
interesting are the oblate solutions which, irrespectively of whether they support a
scalar or not, describe wormholes connecting several asymptotic regions. In the onewormhole sector they reduce to the ring wormholes in the vacuum case, and to the
Bronnikov-Ellis wormhole in the phantom case.

The main point we would like to emphasise is that wormholes exist already in
vacuum GR. This was known already to Zipoy but remains largely unknown to the
community even now. The vacuum wormholes may be viewed as primary since those
sourced by a scalar can be obtained from them by duality rotations. We shall therefore
summarize the essential facts and give a description of the vacuum wormholes, before
applying the duality rotations to produce more general solutions with scalar.
The rest of the text is organised as follows. In Section II we introduce the theory
of gravitating massless scalar and describe its simplest solutions, both in the case of
conventional scalar and for the phantom . In Section III we impose the axial symmetry and list the duality transformations which map solutions to solutions. Section
IV describes the simplest vacuum Weyl metrics generated by massive rods and by point
masses. The vacuum wormholes obtained by rescaling and complexifying the one-rod
solution are discussed in Section V, where their topology and the structure of the ring
source are considered. This Section also describes solutions with scalar obtained via
applying the dualities. Milti-wormholes are considered in Section VI, where we also
analyse carefully the regularity at the symmetry axes and describe the two ring and
the N -ring solutions with locally flat geometry. Finally, we consider in Section VII
solutions obtained from the Chazy-Curzon metrics and conclude in Section VIII. The
isometric embeddings for the BE wormhole are described in the Appendix A.
Unless otherwise stated, Planck units are used everywhere in the text. A short
version of this text can be found in [30].

Gravitating scalar field

We consider a system with a real gravitating scalar field,

L = [R 2 ()2 ] g .

(2.1)

Here the parameter takes two values, either  = +1 corresponding to the conventional
scalar field, or  = 1 corresponding to the phantom field. If  = +1 then the energy
is positive and the scalar field mediates an attractive force. If  = 1 then the scalar
kinetic energy is negative (although the total energy may still be positive) and the
corresponding force is repulsive. In what follows we shall be mainly interested in the
phantom case, however, it is instructive to consider the two cases together.
Assuming the spacetime metric to be static
ds2 = e2U dt2 + e2U ik dxi dxk

(2.2)

where U , ik and are functions of the spatial coordinates xk , the Lagrangian becomes


(3)

ik
.
(2.3)
L = R 2 (i U k U +  i k )
The field equations are
1 (3)
R ik = i U k U +  i k ,
2
U = 0 ,
= 0 .

(2.4)

In what follows we shall denote = if  = 1 and = if  = 1. Formally, the

change  = +1  = 1 can be achieved by extending the scalar field to purely
imaginary values: i.
Equations (2.4) admit the following symmetry. If  = 1 and (U, , ik ) is a solution
then the following replacement will give solutions:
U U cos + sin ,
cos U sin ,
ik ik ,

(2.5)

where is a constant parameter. Likewise, if  = 1 and (U, , ik ) is a solution then

other solutions can be obtained via
U U cosh + sinh ,
cosh + U sinh ,
ik ik .

(2.6)

Many non-trivial solutions can be generated by applying these symmetries, let us

consider simple examples.
2.1

Spherically symmetric sector

All known static, spherically symmetric solutions with scalar can be obtained from the
vacuum Schwarzschild solution written in the form (here d2 = d2 + sin2 d2 )
ds2 =

xm 2 x+m 2
dt +
dx + (x + m)2 d2
x+m
xm

by applying the rotations (2.5) or boosts (2.6). Comparing with the line element (2.2)
yields
e2U =
2.1.1

xm
,
x+m

ik dxi dxk = dx2 + (x2 m2 )d2 ,

= 0.

(2.7)

Solutions with scalar field

Let us choose  = +1, = . Applying to (2.7) the rotation (2.5) with cos = 1/s
gives
e2U

xm
=

x+m

xm
x+m

1/s

e2 = 1

xm
x+m

s2 1/s
,

while the 3-metric ik does not change. Therefore,


1/s

1/s
 2

xm
x+m
2
2
ds =
dt +
dx + (x2 m2 )d2 ,
x+m
xm



2
s 1
xm
=
ln
,
|s| 1,
2s
x+m

(2.8)

(2.9)

which are the well known solutions found by Fisher [31] and by Janis-RobinsonWinicour [32] (FJRW). If s then the metric becomes ultra-static,


1
xm
2
2
2
2
2
2
ds = dt + dx + (x m ) d ,
= ln
.
(2.10)
2
x+m
2.1.2

Solutions with phantom

Let us now set  = 1, = and apply to (2.7) the boost (2.6) with cosh = 1/s.
This gives the phantom version of the FJRW solutions (2.9),
1/s

1/s


 2
x+m
xm
2
2
dt +
dx + (x2 m2 )d2 ,
ds =
x+m
xm



xm
1 s2
=
ln
|s| 1.
(2.11)
2s
x+m
Although the metric looks identical to that in (2.9), the geometry is different because
the range of s is now different. One may think that these solutions could describe
regular black holes, since setting 1/s to integer values removes the branching singularity
at the horizon x = m. However, the area of the sphere is proportional to
(x m)11/s ,

(2.12)

hence the horizon has an infinite area, therefore these solutions cannot describe regular
black holes.
2.1.3

Wormholes

Let us extend the parameters m, s in (2.9), (2.11) to imaginary values,

m i,

s is.

(2.13)

One has then

x i
xm

= e2i ,
x+m
x + i

 
x
= arctan
,

(2.14)

hence the metric in (2.9), (2.11) remains real, the scalar field in (2.9) becomes purely
imaginary, i, while the phantom field in (2.11) remains real, . As a result,
upon the complexification (2.13) both the FJRW solutions (2.9) and their phantom
counterparts (2.11) reduce to the same solutions for the phantom field,
ds2 = e2/s dt2 + e2/s [dx2 + (x2 + 2 )d2 ] ,

s2 + 1
=
.
s

(2.15)

These solutions describe globally regular and asymptotically flat wormholes (see the
Appendix A where the isometric imbedding of the geometry (2.15) into a higherdimensional Minkowski space is considered). They owe their existence to the repulsive

nature of the phantom field the gravitational attraction being compensated by the
scalar field repulsion gives rise to globally regular equilibrium configurations. In the
s limit one obtains the ultra-static wormhole of Bronnikov and Ellis [13, 14],
ds2 = dt2 + dx2 + (x2 + 2 ) d2 ,

= .

(2.16)

This can also be obtained directly from the ultra-static solution (2.10) via m i.
2.2

Anti-gravitating solutions with phantom field

Other solutions for the phantom field can be constructed in view of the following
observation [33]. Taking the infinite boost limit in (2.11),
s 0, m 0,

M =

m
= const,
s

(2.17)

gives
ds2 = e2U + e2U (dx2 + x2 d2 ),

U=

M
,
x

= U.

(2.18)

One notices that the 3-metric ik becomes flat and = U is a harmonic function.
This allows one to generalize the solutions because if U = then Eqs.(2.4) show
(3)

that R ik = 0 and hence one can choose ik = ik . The equation for U then becomes
U = 0 where is the ordinary Laplace operator. Therefore, any harmonic function
U = gives a solution, for example the multi-centre solution,
2

ds = e

2U

2U

+e

(dx) ,

U=

N
X
n=1

Mn
,
|x xn |

= U,

(2.19)

where Mn and xn are arbitrary. The existence of such solutions is the speciality of the
phantom field whose repulsion can exactly compensate the gravitational attraction.

Axially symmetric Weyl metrics and their symmetries

Let us return to equations (2.4) and assume the system to be axially symmetric. Then
the spacetime metric can be put to the Weyl form,




ds2 = e2U dt2 + e2U e2k d2 + dz 2 + 2 d2 ,
(3.1)

where U, k and depend on , z. Then equations (2.4) reduce to2

2U
2U
1 U
+
+
= 0,
2

z 2
1 2
2
+ 2 = 0,
+
2

z
"
2 
2
 2
 2 #
k
U
U

+

,

z


U U

k
= 2
+
.
z
z
z

(3.2)

Here the last two equations are compatible with each other in view of the first two
equations; their solution can be obtained from the line integral

Z 
k
k
k=
d +
dz .
(3.3)

z
This expresses, in principle, k in terms of U, , but in practice the computation of the
integral may be not easy. Since z k vanishes at = 0 (if only U, are non-singular
there), it follows that k = k0 at the symmetry axis. If k0 = 0 then the geometry at
the axis will be regular, but for k0 6= 0 there is a conical singularity on that portion of
the axis. Physically this corresponds to a cosmic strut or a cosmic string with deficit
angle
= 2(1 ek0 )
(3.4)
giving rise to a force or tension

.
8
Such conical singularity are typical for Weyl solutions.
T =

3.1

(3.5)

Symmetries

Equations (3.2) are invariant under the following operations.

3.1.1

Sign flips and scaling

If (U, k, ) is a solution of Eqs.(3.2) then the following replacements will give solutions
U U,
U U,

kk ,
kk ,

;
.

(3.6)

These symmetries are not very interesting: the first one simply changes sign of the
scalar field, while the second one, when applied to solutions (2.9),(2.11),(2.15), changes
sign of the parameters m and . A more interesting symmetry is
U U,

k 2 k,

(3.7)

One can check that this reduction is indeed consistent, which would not be the case if the scalar
field had a potential. The Weyl formulation is also consistent for an electrostatic vector field, so that
it applies, for example, within the electrostatic sector of dilaton gravity.

with constant . As we shall see, this symmetry acts in a non-trivial way completely
changing properties of solutions.
These three symmetries do not mix up gravitational and scalar variables so that,
in particular, vacuum solutions with = 0 remain vacuum. There are also symmetries
which intermix the variables, they are different for the ordinary scalar field ( = +1)
and for the phantom field ( = 1).
3.1.2

Rotations and boosts

These symmetries generate solutions with scalar field. The rotations apply only in the
conventional scalar case ( = +1 and = ) acting on a solution (U, k, ) as
U U cos + sin ,

k k,

cos U sin ,

(3.8)

where is a constant parameter. The boosts apply only in the phantom case ( = 1
and = ) acting on (U, k, ) as
U U cosh + sinh ,
3.1.3

k k,

cosh + U sinh .

(3.9)

Swap symmetry

This applies only in the phantom field case,

U ,

k k,

U,

(3.10)

so that U and swap while k flips sign. The existence of this symmetry implies that
there are twice as many solutions with phantom field as those with the ordinary scalar.
In particular, this symmetry relates the ring wormholes and the phantom wormholes.
3.1.4

Tachyon symmetry

A complex change of the temporal and azimuthal variables

t i,

it

(3.11)

puts the spacetime metric (3.1) to the form




ds2 = 2 e2U dt2 + e2U e2k d2 + dz 2 + e2U d2 ,

(3.12)

which is again the Weyl metric obtained from the original one via
U ln U,

k k 2U + ln ,

(3.13)

One can check that this is a symmetry of equations (3.2) which does not change the
scalar. This symmetry has been relatively little studied. Applied to the Schwarzschild
metric it produces the Peres-Schulman-Gott (PSG) tachyon solution with isometry
group SO(2)SO(2, 1) [34], [35], [36]. This is appropriate for a neutral tachyon moving
in a spacetime with one spatial direction compactified; it describes the aftermath of
vacuum decay [37].

10

Vacuum Weyl metrics

Let us describe the simplest solutions of Eqs.(3.2) with = 0.

4.1

One-rod solution

Let us see how the Schwarzschild solution

ds2 = e2U dt2 + e2U dl2
with
xm
,
dl2 ik dxi dxk = dx2 + (x2 m2 )d2
(4.1)
x+m
looks when transformed to the Weyl form. The transformation is achieved by setting

(4.2)
z = x cos ,
= x2 m2 sin ,
e2U =

hence
d2 + dz 2 =


x2 m2 cos2  2
dx + (x2 m2 )d2 ,
2
2
x m

(4.3)

which gives


ds2 = e2U dt2 + e2U e2k (d2 + dz 2 ) + 2 d2 ,

(4.4)

with
x2 m2
xm
,
e2k = 2
.
(4.5)
x+m
x m2 cos2
There remains to express U, k in terms of , z. Inverting (4.2) gives x, in terms of
z, ,
e2U =

x = R,

x m cos = R ,

(4.6)

where
p
1
R = (R+ + R ),
R = 2 + (z m)2 .
2
Injecting this to (4.5) gives U, k expressed in terms of Weyl coordinates,

 2


Rm
R m2
1
1
U = ln
,
k = ln
,
2
R+m
2
R+ R

(4.7)

(4.8)

and one can directly check that these fulfill equations (3.2). This gives the Schwarzschild
metric in the Weyl form. One notices that U can be represented as the Newtonian
potential of a thin rod of linear mass density 1/2 and of mass m, hence3
Z
d
1 m
p
U =
.
(4.9)
2 m 2 + (z )2
For this solution one has k( = 0, z) = 0 if |z| > m, hence there are no struts and the
geometry is regular on parts of the axis not occupied by the rod.
3

Directly calculating the integral in (4.9) gives U =

11

1
2

R+ zm
ln R
which is equivalent to U in (4.8).
z+m

4.2

Two-rod solution

More general solutions can be obtained by taking several rods and superposing their
Newtonian potentials. This gives U , while k is obtained by integrating the last two
equations in (3.2). In the simplest two-rod case one has [38]
U = U1 + U2 ,

k = k1 + k2 + k12 ,

(4.10)

where (with a = 1, 2)




1
Ra ma
1
(Ra )2 (ma )2
Ua = ln
,
ka = ln
,
2
Ra + ma
2
Ra+ Ra


1
(R1+ R2 + z1+ z2 + 2 )(R1 R2+ + z1 z2+ + 2 )
k12 = ln
,
2
(R1+ R2+ + z1+ z2+ + 2 )(R1 R2 + z1 z2 + 2 )

(4.11)

with
za = z za ma ,

Ra =

1
Ra = (Ra+ + Ra ). (4.12)
2

2 + (za )2 ,

For this solution k vanishes on parts of the = 0 axis non-occupied by the rods, except
for the interval between the rods where it is constant and negative. Therefore, there is
a negative tension strut between the two rods giving rise to the repulsive force. This
strut props up the two black holes and does not let them fall on each other.
The generalization to N rods is straightforward [38],
U=

N
X

Ua ,

k=

N
X

ka +

a=1

a=1

kab ,

(4.13)

a<b

where Ua , ka are given by the same expressions as in (4.11), while kab is obtained from
k12 in (4.11) by replacing R1 Ra and R2 Rb .
4.3

The Chazy-Curzon metrics

Instead of the Newtonian potential of rods one can consider that of point masses located
at the = 0 axis. For just one mass m located at z = 0 one obtains the solution
U =

m
,
R

k=

m2 2
,
2R4

(4.14)

p
where R = 2 + z 2 .
p
For two masses m located z = m one has, with R = 2 + (z m)2 ,
m+ m

,
R+
R


m2+ 2
m2 2
m+ m 2 + z 2 m2
k=

+
1 .
2(R+ )4 2(R )4
2m2
R+ R

U =

12

(4.15)

These solutions were constructed by Chazy [39] and Curzon [40] and later independently by Silberstein [41]. The one-mass solution is regular at the axis, while for the
two masses one has


m+ m z 2 m2
k( = 0, z) =
1 .
(4.16)
2m2
|z 2 m2 |
This vanishes outside the interval between the two particles where z 2 m2 > 0 but
not inside this interval where z 2 m2 < 0 and hence [1, 42]
k( = 0, z) =

m+ m
.
m2

(4.17)

The original paper of Curzon [40]4 also gives the solution for N point masses.

Solutions from one rod

We described above the elementary vacuum solutions generated by rods and by point
masses. One could now apply rotations (3.8) and boosts (3.9) to produce new solutions
with scalar field. However, before doing this we notice that yet much larger families
of vacuum metrics can be obtained via acting first with the scale symmetry (3.7) and
then performing complexification. This gives the prolate and oblate vacuum metrics.
5.1

Prolate vacuum metrics

Applying the scaling (3.7) to the one-rod Schwarzschild solution (4.8) gives the twoparameter family of solutions labeled by and m,


 2

Rm
2
R m2

,
k=
ln
,
(5.1)
U = ln
2
R+m
2
R+ R
with
1
R = (R+ + R ),
2

R =

p
2 + (z m)2 .

(5.2)

Using (4.2), (4.3), (4.5) to pass back to the x, coordinates gives





xm
xm
2
2
ds =
dt +
dl2 ,
(5.3)
x+m
x+m
 2
12
 2

x m2 cos2
2
2
2
2
dx
+
(x

m
)d
+ (x2 m2 ) sin2 d2 .
dl =
2
2
x m
These are the prolate Zipoy-Voorhees (ZV) solutions [26, 45]. Unless for 2 = 1 they
are not spherically symmetric and exhibit a curvature singularity at R = x = m. These
4

The reader is warned that reference to Curzons paper [40] in the literature is often given incorrectly. Interestingly, this paper of 1924 also gives what is now known as the Majumdar-Papapetrou
[43, 44] multi-centre solution of Einstein -Maxwell theory.

13

metrics have been relatively well studied, as they can be used to describe deformations
of the Schwarzschild black hole (see for example [46, 47]). For all these solutions the
relation between the x, and , z coordinates is given by (4.2), hence
z2
2
+
= 1.
x2 x2 m2

(5.4)

Therefore, lines of constant x are prolate ellipses with the major semi-axis oriented
along the z-direction. In the x m limit these ellipses shrink to the m z m
segment of the z-axis. Both (, z) and (x, ) coordinate cover the R = x > m region
of the manifold.
5.2

Oblate vacuum metrics

Let us continue the solution (5.1) to imaginary parameter values,

m i ,

i.

(5.5)

In this case instead of (4.2) one will have

z = x cos ,

p
x2 + 2 sin ,

(5.6)

hence
2
z2
+
= 1.
x 2 x2 + 2

(5.7)

Therefore, lines of constant x in the (, z) plane are oblate ellipses with the major
semi-axis oriented along the -direction. In the x 0 limit these ellipses shrink to the
segment of the -axis
I = { [0, ], z = 0}

(5.8)

(assuming that > 0). The functions R in (5.2) become complex valued,
p
p
R = 2 + (z m)2
2 + (z i)2 X iY,

(5.9)

with
X = X ,

Y = sign(z) Y,

(5.10)

where
1
X =
2

A2 + B 2 + A ,

1
Y=
2

A2 + B 2 A ,

(5.11)

with A = 2 + z 2 2 and B = 2z. Since R is replaced by a real quantity,

1
R = (R+ + R ) X,
2

14

(5.12)

and one has

1
ln
2

Rm
R+m

1
ln
2

X i
X + i


= i arctan


,

it follows that the solution remains real-valued,

 
 2

X
2
X +Y2
U = arctan
,
k=
ln
.

2
X 2 + 2

(5.13)

(5.14)

These are the oblate Zipoy-Voorhees solutions [26, 45]. These solutions are relatively
little known, therefore we shall describe some of their properties.
The two sign choices in (5.10) correspond to the two square root branches. For
each branch the real and imaginary parts of the square root, X and Y , are discontinuous
and/or not smooth at the branch cut along the segment (5.8). However, gluing the two
branch sheets together gives a Riemann surface on which all functions are smooth and
continuous. The functions X, Y defined by (5.10) are already smooth and continuous
for > , while for < one should analytically continue them through the cut,
which is achieved by replacing
X sign(z)X = sign(z)X ,

Y sign(z)Y = Y.

(5.15)

These functions are continuous for < (although discontinuous for > ). In
particular, at the symmetry axis one has after the analytic continuation either X = z,
Y = or X = z, Y = .

Figure 1. U (, z) (left) and k(, z) (right) for the solution (5.14) with = = 1. The two
branches of U (blue and yellow online) correspond to X chosen according (5.15).

As a result, one obtains the plots of the solution (5.14) shown in Fig.1. The
function U (, z) is double-valued while k(, z) is single-valued and vanishes at the
symmetry axis. The two different values of U for given , z correspond to two different
spacetime points.

15

5.2.1

Wormhole topology

As first noticed by Zipoy [26], solutions (5.14) describe wormholes. This can be seen
by making the complexification (5.5) directly in (5.3), which gives
 
x
2
2U 2
2U 2
ds = e dt + e dl ,
U = arctan
,
(5.16)

 2
1+2
 2

x + 2 cos2
2
2
2
2
dl =
dx
+
(x
+

)d
+ (x2 + 2 ) sin2 d2 .
2
2
x +
This is the standard form of the oblate Zipoy-Voorhees solutions [26, 45]. When written
in this form, it is seen that these metrics describe wormholes with two asymptotically
flat for x regions connected by a throat at x = 0. This is obvious already by
noticing that close to the symmetry axis where cos2 1 the metric (5.16) reduces to
the wormhole metric (2.15) (up to the replacement 1/s).
The coordinates (x, ) are global and all their functions are single valued, while
the Weyl coordinates are not global and one needs two Weyl charts [19]
D : { 0, < z < }

(5.17)

to cover the manifold. Let us discuss the relation between the two coordinate systems.
The (, z) coordinates are related to (x, ) via (5.6),
p
(5.18)
z = x cos , = x2 + 2 sin ,
therefore,
X iY =

p
p
2 + (z i)2 = (x i cos )2 = x i cos ,

(5.19)

hence
x = X(, z),

cos = Y (, z),

(5.20)

which gives inverse transformation from (, z) to (x, ). Since for given (, z) there are
two different branches of (X, Y ) defined by (5.10), the inverse transformation is double
valued the same pair of values (, z) can correspond either to (x, ) or to (x, ).
Therefore, the D+ chart covers the positive (x > 0) wormhole side while the D
chart covers the negative (x < 0) side (see Fig.2) [48].
The symmetry axis in Weyl coordinates, = 0, z (, +), has two images
in the spacetime: either = 0, z = x or = , z = x. Therefore, the spacetime
actually has two symmetry axes, one at = 0 and one at = , corresponding to
the edges of the (x, ) strip in Fig.2. This explains, in particular, the two values of
U (, z) in Fig.1. For example, at = 0 one has for the (yellow online in Fig.1) branch
of U (, z) with X = x = z and Y =
 
 
z
x
U (0, z) = arctan
= arctan
.
(5.21)

16

Figure 2. Wormhole topology: The x, coordinates cover the whole of the manifold, while
each of the two Weyl charts D+ and D covers only either the x > 0 or x < 0 part of the
manifold. The upper edge of the cut in the D+ patch is identified with the lower edge of the
D cut and vice-versa. A winding around the ring in the x, coordinates corresponds to
two windings in Weyl coordinates.

This gives the value at the = 0 axis. Similarly, for the second (blue online) branch
of U in Fig.1 one has X = x = z and Y = at = 0, hence
 
 
z
x
U (0, z) = arctan
= arctan
,
(5.22)

which gives the value at the = axis. These two values of U (0, z) coincide when
expressed in terms of the x-coordinate.
It is also instructive to pass to the Weyl coordinates directly in (5.16). The
starting point is the line element
dl2 = dx2 + (x2 + 2 )(d2 + sin2 d2 )

(5.23)

where x (, ). This can be transformed to the isotropic form by setting

x=

2 2
2

=x+

p
x2 + 2 [0, ),

(5.24)

which gives
1
dl =
4
2


2
2
1 + 2 [d2 + 2 d2 + 2 sin2 d2 ].

17

(5.25)

Introducing the complex variable

= iei = sin + i cos

(5.26)

one obtains
1
dl =
4
2


2
2
1+
[|d|2 + (<())2 d2 ].
||2

(5.27)

The -variable sweeps the upper half-plane <() 0 such that the x < 0 and x > 0
regions map, respectively, to the || < and || > regions. Passing to the new
complex variable w() via the Joukowski transformation [22],


1
2
w=
+
+ iz,
(5.28)
2

one finds that = <(w) and z = =(w) are related to x, by (5.18), whereas the metric
becomes

2  2
 d 
2
1
2
  |dw|2 + 2 d2 .
1+
(5.29)
dl =
 dw 
2
4
||
As a result, one has

2  2
 d 
x2 + 2
1
2
  |dw|2 =
dx + (x + )d =
1+
(dz 2 + d2 ),
4
||2  dw 
x2 + 2 cos2
(x2 + 2 ) sin2 d2 = 2 d2 ,
(5.30)
2

using which and assuming (5.20), the solution (5.16) takes up the Weyl form with U, k
given by (5.14).
The Joukowski transformation w (5.28) maps the regions || < and || >
of the -plane to the whole w-plane with the removed segment or real axis,
w [, ].

(5.31)

The inverse Joukowski transformation

=w

w 2 2

(5.32)

has branching points at w = connected by the branch cut (5.31). Any closed
contour intersecting the cut passes from one Riemann sheet to the other and the sign
of the square root in (5.32) changes. As a result, the full Riemann surface on which
the Joukowski function is single valued consistes of two sheets glued together through
the cut (5.31).
Restricting now to the upper half-plane =() 0, the positive (|| > )
and negative (|| < ) wormhole sides map, respectively, to the D+ and D Weyl
regions (5.17), each region having the branch cut (5.8) removed (see Fig.3). The entire

18

Figure 3. The Joukowski function (5.28) maps the interior (x < 0) of the semi-cercle || = ,
=() 0 to a half-plane <(w) 0 with a cut removed, which corresponds to the D Weyl
region. The exterior (x > 0) of the semi-circle maps similarly to the D+ Weyl region.

coordinate atlas of the spacetime manifold is obtained by identifying the upper edge of
the cut on the D+ chart with the lower edge of the cut on the D chart and vice-versa.
The negative part of the z-axis in the D patch then merges with the positive part of
the z-axis in the D+ patch (see Fig.2), which gives the = 0 axis; similarly for the
= axis.
It is worth emphasising again that the wormhole features of the oblate vacuum solutions had been recognised already in 1966 by Zipoy [26], but even now these solutions
remain very little known.
5.2.2

Ring wormholes

The branching point of the Weyl coordinates at + iz = gives rise to the conical
singularity of the metric (5.16) at x = 0, = /2. This singularity is distributed along
a ring of radius . Specifically, with y = cos , the spatial part of the metric (5.16)
becomes for small x, y
2

dl =

x2 + y 2
2

1+2

(dx2 + dy 2 ) + 2 d2 ,

(5.33)

and introducing instead of x, y the polar coordinates r, via

x + iy
=

r
q

1/q

ei

(5.34)

with q = 2 + 2 yields
dl2 = dr2 + r2 q 2 d2 + 2 d2 .

19

(5.35)

Since q ranges from zero to 2q > 2 it follows that the metric contais a conical
singularity at r = 0 with a negative angle deficit
= 2 2q = ( 2 + 1)2.

(5.36)

This singularity corresponds to an infinitely thin ring of radius and of negative tension
(energy per unit length) [30]
(1 + 2 )c4
T =
4G

(5.37)

(the correct physical dimensions are restored here for the moment). The ring encircles
the wormhole throat and plays the role of the negative energy source for the solution.
Let us consider the limit where 0. The functions U, k in (5.14) then vanish
and the metric expressed in Weyl coordinates becomes manifestly flat,
ds2 = dt2 + d2 + dz 2 + 2 d2 .

(5.38)

However, the topology is still non-trivial because the (, z) coordinates do not cover
the whole of the manifold but only either the D+ or D patches glued to each other
through the cuts. A contour around the ring core in the D+ patch does not close after
a revolution of 2 but passes to the D patch and only after a second revolution of 2
returns back to the initial position (see Fig.2). Therefore, the winding angle around
the ring core ranges from zero to 4, which indicates that the ring is still there and
has the tension
T =

c4
.
4G

(5.39)

At the same time, the geodesics which do not hit the ring are simply straight lines as
in flat space. Those which pass outside the ring always stay in the same coordinate
chart (line B in Fig.2), while those threading the ring pass to the other chart and
become invisible from the previous chart thus traversing the wormhole (line A in Fig.2).
Therefore, the ring genuinely cuts a hole in flat space through which on can observe
another universe and get there. This reminds one of Alice observing the room behind
the looking glass and next jumping there. An object falling through the ring can be
seen from behind, while viewed from the side it is not seen coming from the other side
(see Fig.4).
It seems plausible that the wormhole edges could be smoothened without changing
the global structure if the singular ring is replaced by a regular hoop-shaped energy
distribution of finite thickness and with the same tension. Inside the hoop the energy
density is finite hence the geometry must be regular, while outside the energy is zero
and the geometry should be more or less the same as for the original ring. This suggests
that wormholes could be created by negative energies concentrated in toroidal volumes,
for example by vacuum fluctuations. Therefore, traversable wormholes could actually
be more than just a mathematical curiosity.

20

Figure 4. Particles entering the ring are not seen coming out from the other side

However, the energy needed to create a ring wormhole is very high. The absolute
value of the negative tension T in (5.39) coincides with the highest possible value
for a positive tension (force), according to the maximum tension principle in General
Relativity conjectured in [49, 50]. This conjecture is supported, for example, by the
fact that the angle deficit of a cosmic string cannot exceed 2. Numerically one has
T = 3.0257 1043 Newtons 3 1039 Tonnes, hence to create a macroscopic ring
wormhole of radius R = 1 metre, say, one needs a negative energy equivalent to the
mass of Jupiter, 2RT /c2 103 M [30], [51].
At the same time, one can also imagine microscopically small rings by quantum
fluctuations appearing spontaneously from the vacuum and then disappearing again.
Particles crossing the ring during its existence would no longer be accessible from our
universe after the ring disappears. If true, this would be a potential mechanism for the
loss of quantum coherence.
5.3

Solutions with scalar field

We can now produce new solutions with scalar field by applying the rotations (3.8) and
boosts (3.9) to the prolate and oblate vacuum metrics written in the form (5.3) and
(5.16). Applying the rotation (3.8) produces solutions with the conventional scalar in
the prolate case,

/s

/s



xm
s2 1
xm
xm
2
2
2
dt +
dl , =
ln
,
ds =
x+m
x+m
s
2
x+m
12
 2
 2

x m2 cos2
2
2
2
2
dl =
dx
+
(x

m
)d
+ (x2 m2 ) sin2 d2 , (5.40)
2
2
x m
and in the oblate case,

s2 1
,
(5.41)
ds = e
dt + e
dl ,
=
s
 2
1+2
 2

x + 2 cos2
2
2
2
2
dl =
dx
+
(x
+

)d
+ (x2 + 2 ) sin2 d2 ,
2
2
x +
2

2(/s)

2(/s)

21

where = arctan(x/) and s 1. Applying boosts (3.9) gives solutions with the
phantom field boosted prolate solutions,

/s

/s



x

m
xm
x

m
1 s2
2
2
2
dt +
dl , =
ln
,
ds =
x+m
x+m
s
2
x+m
 2
12

 2
x m2 cos2
2
dl =
dx + (x2 m2 )d2 + (x2 m2 ) sin2 d2 , (5.42)
2
2
x m
and boosted oblate solutions,

1 s2
ds = e
dt + e
dl ,
=
,
(5.43)
s
1+2
 2
 2

x + 2 cos2
2
2
2
2
dx
+
(x
+

)d
+ (x2 + 2 ) sin2 d2 ,
dl =
x2 + 2
2

2(/s)

2(/s)

where is the same as before and s 1. Finally, new solutions with the phantom
scalar can be obtained by applying the swap symmetry (U, k, ) (, k, U ). This
gives the prolate solutions,



/s

/s

x

m
xm
x

m
1 + s2
2
2
2
ds =
ln
dt +
dl , =
,
x+m
x+m
s
2
x+m
 2
1+2
 2

x m2 cos2
2
dx + (x2 m2 )d2 + (x2 m2 ) sin2 d2 , (5.44)
dl =
2
2
x m
and the oblate solutions,

1 + s2
ds = e
dt + e
dl ,
=
,
(5.45)
s
 2
12
 2

x + 2 cos2
2
2
2
2
dl =
dx
+
(x
+

)d
+ (x2 + 2 ) sin2 d2 ,
x2 + 2
2

2(/s)

2(/s)

where now s (, +). The latter solutions reduce to the BE wormholes for = 1.
We have therefore obtained six families of new solutions, each carrying three free
parameters either m, , s or , , s. The oblate solutions can be obtained from the
prolate ones by analytic continuation m i and i. We also notice that
(5.44),(5.45) can be obtained either from (5.40),(5.41) via m i, s is or from
(5.42),(5.43) via i, s is. As a result, further analytic continuations do not
give new solutions.
However, any of the above solutions can in addition be acted upon by the tachyon
symmetry (3.13) interchanging t and . This doubles the size of the family of new
solutions. Most of these solutions are only axially symmetric and have never been
described before. Here are some of their properties.
There are twice as many solutions with the phantom field as those with the
conventional scalar owing to the swap symmetry (U, k, ) (, k, U ).

22

 The rotated prolate solutions (5.40) for the conventional scalar generalize the
FJRW family (2.9) and reduce to it for 2 = 1.
The boosted prolate solutions (5.42) for the phantom field reduce for 2 = 1 to
the phantom version (2.11) of the FJRW solutions.
The (boosted and swapped) oblate solutions (5.45) for the phantom field reduce
for 2 = 1 to the BE spherically symmetric wormholes (2.15). For 2 6= 0 they
describe axially symmetric deformations of the BE wormholes.
The oblate solutions (5.41),(5.43) describe ring wormholes supporting either conventional or phantom scalars. When 0, the scalar vanishes and the solutions
reduce to the ring wormhole (5.38) with locally flat geometry, even though the
metric does not look flat when expressed in the x, coordinates:
ds2 = dt2 +


x2 + 2 cos2  2
2
2
2
+ (x2 + 2 ) sin2 d2 . (5.46)
dx
+
(x
+

)d
x2 + 2

The role of the prolate solutions (5.44) is less clear they never reduce to the
Schwarzschild metric although the geometry becomes flat as 0.
Solutions (5.40), (5.41), (5.44), (5.45) become ultrastatic (U = 0) for s but
the scalar field remains non-trivial in this limit. The oblate solutions (5.45) interpolate in this limit between the spherically symmetric ultrastatic BE wormhole
for 2 = 1 and the ring wormhole (5.46) for = 0.
Yet one more interesting limit is obtained by taking in solutions with phantom
(5.42)(5.45) 0, 0, s 0 with fixed /s and /s. The three dimensional
part of the metric dl2 then reduces to that in (5.46) and hence becomes flat, but
since U = 6= 0, the 4-geometry remains non-trivial. This is a particular case
of the anti-gravitating solutions described in Section (2.2).

Solutions from two rods

Let us now apply the scaling symmetry (3.7) to the two-rod solution (4.10). First of
all we notice that this symmetry can be extended since the mass density of each rod
can be rescaled independently. As a result, there can actually be two scale parameters
1 and 2 and rescaling the solution (4.10) gives a larger family of vacuum metrics,
U = 1 U1 + 2 U2 ,

k = 21 k1 + 22 k2 + 1 2 k12 ,

(6.1)

where (a = 1, 2)




Ra ma
1
(Ra )2 (ma )2
1
,
ka = ln
,
Ua = ln
2
Ra + ma
2
Ra+ Ra


1
(R1+ R2 + z1+ z2 + 2 )(R1 R2+ + z1 z2+ + 2 )
k12 = ln
,
2
(R1+ R2+ + z1+ z2+ + 2 )(R1 R2 + z1 z2 + 2 )

23

(6.2)

with
za = z za ma ,

Ra =

1
Ra = (Ra+ + Ra ).
2

2 + (za )2 ,

(6.3)

These prolate solutions depend on 6 parameters a , ma , za .

The oblate solutions can be obtained by extending to complex parameter values,
a ia ,

ma ia .

The amplitudes Ra in (6.2) then become

p
Ra 2 + (za )2 Xa iYa ,

(6.4)

Ra Xa ,

(6.5)

where
za = z za ia .

(6.6)

Computing the square root gives

Xa = sX (a)Xa ,

Ya = sY (a)Ya ,

(6.7)

where sX (a) and sY (a) take values +1 or 1 and will be specified below, while
qp
qp
1
1
2
2
Xa =
(Aa ) + (Ba ) + Aa , Ya =
(Aa )2 + (Ba )2 Aa (6.8)
2
2
with
Aa = 2 + (z za )2 2a ,

Ba = 2a (z za ).

(6.9)

This gives the oblate solutions,

k = 12 k1 + 22 k2 + 1 2 k12 ,

U = 1 U1 + 2 U2 ,

(6.10)

where

Ua = arctan

Xa
a


,

1
ka = ln
2

Xa2 + Ya2
Xa2 + 2a


,

(6.11)

and
k12


!
 (X1 + iY1 )(X2 + iY2 ) + z1+ z2+ + 2 2
1
 .
= ln 
2
(X1 + iY1 )(X2 iY2 ) + z1+ z2 + 2 

24

(6.12)

6.1

Multi wormholes

We are mainly interested in the oblate solutions (6.10) describing a pair of wormholes
(they can be generalized to the multi wormhole case as explained around Eq.(4.13)),
hence we shall describe some of their properties.
A closer inspection reveals that a proper choice of signs sX (a), sY (a) in (6.7) is
essential, since otherwise the field equations are fulfilled only in some regions of space.
Specifically, since Ua and ka in (6.11) have the same structure as for the one wormhole
solution, they satisfy
2 Ua
1 Ua 2 Ua
+
+
= 0,
2

z 2
"
2 
2 #
Ua
Ua
ka
=

z
ka
Ua Ua
= 2
,
z
z
and these equation hold everywhere. The k12 amplitude should satisfy


U1 U2 U1 U2
k12
= 2

z z


k12
U1 U2 U2 U1
= 2
+
,
z
z
z

(6.13)

(6.14)

but these equations do not always hold. For example, choosing in (6.7) sX (a) =
sY (a) = 1 these equations are satisfied only if z > z1 and z > z2 and only if both 1
and 2 are positive. The reason is that k12 depends on phases of za ,


a
,
(6.15)
a (z) = arctan
z za
which jump when z crosses values za . As a result, k12 jumps too, and the equations
are satisfied only from one side of the jump. The proper sign choice for which the
equations are satisfied everywhere is
Xa = s(a) Xa ,

Ya = s(a) sign(a ) sign(z za ) Ya ,

(6.16)

which implies that the Ya amplitudes experience jumps through the cuts
Ia = { [0, |a |], z = za }.

(6.17)

There are four options of choice of the coefficients s(a) in (6.16):

s(1) = s(2) = 1

or

s(1) = s(2) = 1.

(6.18)

These four options correspond to the four different branches of the same solution that
should be considered all together, since for each individual branch functions Xa , Ya are

25

either discontinuous or not smooth at the cuts. However, considering them together
for all four different values of s(a) gives a smooth Riemann surface consisting of four
sheets continuously joining each other through the cuts. Specifically, amplitudes (6.16)
are already smooth for > |a |, while for < |a | one should continue them through
the cuts via (compare with (5.15)) Xa sign(z za )Xa and Ya sign(z za )Ya which
gives
Xa = s(a) sign(z za )Xa ,

Ya = s(a) sign( a )Ya .

(6.19)

These functions are smooth and continuous for < |a |. Using them in (6.10) gives
plots of the solution shown in Fig.5 and Fig.6. One can see that U (, z) is four-valued,
hence one needs four different Weyl charts to cover the solution:
s(1)

Ds(2) = D
: { 0, < z < }.

(6.20)

On each chart the signs of (Xa , Ya ) are chosen according to the values of s(a) in (6.16).
Each chart has two branch cuts (6.17) and the complete coordinate atlas covering
the manifold is obtained by gluing together the four charts to analytically continue
+

the functions through the cuts. One glues the I1 cuts on the D+
and D+
charts
by identifying the upper edge of one cut with the lower edge of the other and vice
+

versa, similarly for the I1 cuts on the D

and D
charts. One also glues I2 cuts on
+
+

the D+
and D
charts and on the D+
and D
charts. The resulting atlas covers a
manifold M with four asymptotic regions connected through the wormhole throats.
The continuation through cuts implies that for < |a | (for example at the symmetry
axis = 0) the field amplitudes are given by (6.19).
As a result, expanding the function in (6.19) at small yields


2
4

(6.21)
Xa + iYa = s(a) za + + O( )
2za
where za is defined in (6.6) and s(a) is given by (6.18). Using this to compute U in
(6.10) gives, if s(1) = s(2) = 1,





z z1
z z2
U ( = 0, z) = 1 arctan
+ 2 arctan
,
(6.22)
1
2
and, if s(1) = s(2) = 1,





z z2
z z1
2 arctan
,
U ( = 0, z) = 1 arctan
1
2

(6.23)

which are the values of U on the four symmetry axes. This can be compared with
Eqs.(5.21),(5.22) in the single ring case, in which case there are only two symmetry
axes.
Let us now similarly compute the value of k at the symmetry axes. The amplitudes
ka in (6.12) are insensitive to signs of Xa , Ya hence for given (, z) their values are the
same on all Weyl charts. The k12 amplitude in (6.12) depends on products (X1 +

26

Figure 5. U (, z) defined by (6.10) for 1 = 1, 2 = 1.5, 1 = 1.2, 2 = 0.5, z1 = z2 = 1,

and for [0, 2], z [4, 4]. The four different branch sheets (different colours online)
correspond to values of U in four different spacetime regions.

iY1 )(X2 + iY2 ) and (X1 + iY1 )(X2 iY2 ) for which there are two sign options. If one
chooses in (6.21) s(1) = s(2) then one will have for small
(X1 + iY1 )(X2 + iY2 ) = z1+ z2+ + O(2 ),

(X1 + iY1 )(X2 iY2 ) = z1+ z2 + O(2 ),

inserting which to (6.11),(6.12) gives

k1 (0, z) = k2 (0, z) = k12 (0, z) = 0.

(6.24)

The other possibility is to chose in (6.21) s(1) = s(2), then one has at small



2
2
+
+
(X1 + iY1 )(X2 + iY2 ) = z1 + +
z2 + + + O(4 ),
2z1
2z2



2

2
+

(X1 + iY1 )(X2 iY2 ) = z1 + +

z2 + + O(4 ).
(6.25)
2z1
2z2
Inserting this to (6.11),(6.12) gives
k1 (0, z) = k2 (0, z) = 0,
but k12 assumes a constant non-zero value


(z1 z2 )2 + (1 2 )2
.
k12 (0, z) = ln
(z1 z2 )2 + (1 + 2 )2

(6.26)

(6.27)

The conclusion is that k(, z) vanishes at the two symmetry axes with s(1) = s(2)
where U is given by (6.22), hence these two axes are regular. However, k assumes the

27

Figure 6. k(, z) defined by (6.10) for the same parameter values as in Fig.5 and for [0, 2],
z [4, 4]. It shows only two sheets made of (blue and yellow online) branches computed
from Xa , Ya chosen either as in (6.16) (left panel) or as in (6.19) (right panel). The resulting
sheets are the same in both cases.

constant non-zero value (6.27) everywhere at the other two axes, where s(1) = s(2)
and U is given by (6.22). Hence these two axes are singular and contain infinite struts
(see Fig.6). Therefore, two of the four symmetry axes of the solution are regular while
the other two are everywhere singular.
The two-wormhole system was first discussed in Ref.[19], where the solution for
U (, z) was obtained (see also Ref.[48]). It took next more than 30 years to obtain
also the explicit solution for k(, z) and not just the integral representation (3.3) [21].
The formula (6.27) was also obtained in Ref.[21], however, it was concluded there that
k(0, z) is discontinuous and assumes the value (6.27) either inside the interval [z1 , z2 ]
while vanishing outside, or the other way round. As one can see in Fig.6 (left panel)
precisely this type of behaviour is shown separately by the two (blue and yellow online)
solution branches. However, k is clearly continuous on the combined surface made
of both branches. Moreover, choosing regular near the axis amplitudes Xa , Ya (6.19)
renders k(0, z) constant for each separate branch, as one can see in the right panel of
Fig.6. In any case, since U is everywhere finite and smooth, the field equations imply
that the derivative z k should vanish at the axis hence k(0, z) cannot jump.
6.2

Locally flat wormholes

Summarising the above discussion, the oblate vacuum solutions obtained from the
two-rod metrics describe a pair of ring wormholes. They are covered by four Weyl
charts and have four symmetry axes, two of which are regular while the two others
contain a conical singularity since the function k assumes a constant non-zero value
there. Therefore, these solutions are singular, and it seems the same is true also for
the multi-ring solutions.

28

There is, however, a notable exception obtained by taking in (6.10) the 1 0

and 2 0 limit, which gives U = k = 0 everywhere. The conical singularity along the
z-axes then disappears and the metric becomes manifestly flat although the topology
remains non-trivial as one still needs four Weyl charts to cover the manifold. This
implies that there are conical singularities at the points (a , za ) of the (, z) plane (we
now assume without loss of generality that a > 0 since in the opposite case one should
simply replace in all formulas below a by |a |). For example, a contour around the
+
point (1 , z1 ) in the D+
patch shown in Fig.7 does not close since when arriving at the

lower edge of the cut it continues to the D+

patch and only after a second revolution
+
returns back to the D+
patch to close. Therefore, the total angle increment is 4 hence

Figure 7. The two-ring wormhole is covered by four Weyl charts, each having two branch
+

cuts. The upper cuts on D

and D
are glued to each other such that the upper edge of the
one is identified with the lower edge of the other and vice-versa; similarly for the lower cuts

on D+
and D
. As a result, the geodesic A crosses several charts.

there is an angle deficit of 2 corresponding to a cosmic string of negative tension

T = c4 /(4G). This string stretches in the -direction, hence this is a ring of radius
1 placed at z = z1 . Similar argument shows that there is also a ring of the same
tension and of radius 2 placed at z = z2 . Therefore, the solution is sourced by a pair
of negative tension rings whose positions and radii can be arbitrary.
Since the metric is locally flat, the geodesics are simply straight lines. Those
initially parallel to the z-axis will always remain parallel to it, which helps to understand once again why there are four symmetry axes. Consider a geodesic which is close

29

enough to the axis to thread both rings, as for example the geodesic A in Fig.7 which
+
stars at the D+
chart at z = . When it threads the ring at z = z2 it passes to the
+
D chart and continues there, until it threads the second ring at z = z1 and passes to

the D
chart where it stays till reaching z = +. Therefore, this geodesic follows the
pattern
+
+

D
D
D+

(6.28)

which determines one of the four z-axes of the solution. Similarly, a geodesic which

starts on D
and follows the z-axis proceeds as
+

D+
D+
D

(6.29)

which determines the second axis. The other two axes are similarly defined by
+
+

D
D+
D+
,

+
D+
D
D
.

(6.30)

The above arguments can be straightforwardly generalized to the case of N rings

of arbitrary radii a and positions za and of the same tension T = c4 /(4G). The
metric expressed in Weyl coordinates is manifestly flat but the topology is non-trivial
and one needs 2N Weyl charts
Ds(1)s(2)...s(N ) : { 0, < z < }

(6.31)

to cover the whole of the manifold. Here s(a) = 1 and each chart has N cuts
Ia = { [0, a ], z = za } ,

a = 1, . . . N.

(6.32)

The atlas is obtained by gluing together the m-th cuts on all pairs of charts with
coinciding indices s(a 6= m). For example, the upper edge of the first cut on D+s(2)...s(N )
should be identified with the lower edge of the first cut on Ds(2)...s(N ) and vise versa for
all possible values of s(2), . . . s(N ). This gives an exact solution of Einstein equations
sourced by N singular rings whose geometry is everywhere flat except at the ring
positions.
6.3

Solutions with scalar field

Any of the discussed above vacuum Weyl metrics described by

(U, k, = 0)

(6.33)

can be promoted to solutions with 6= 0. Applying symmetries (3.8) and (3.9) gives
solutions with the conventional scalar,

1
s2 1
Us = U,
ks = k, s =
U, s 1,
(6.34)
s
s
boosted solutions with phantom field,
1
Us = U,
s

ks = k,

s =

30

1 s2
U,
s

s 1,

(6.35)

and their swapped version

1
Us = U,
s

ks = k,

s =

1 + s2
U,
s

s (, ).

(6.36)

Since the original solution can exist in prolate and oblate versions, this gives six oneparameter families of solutions with scalar. In addition, twice as many solutions can
be obtained by acting on (6.34)(6.36) with the tachyon symmetry (3.13).
For example, taking the two-wormhole vacuum metric (6.10), applying the swap
symmetry (6.36), and then taking the limit s gives the two-wormhole counterpart
of the ultrastatic solution of Bronnikov and Ellis. This solution is no longer spherically
symmetric and it contains infinite struts along two symmetry axes.
More generally, all solutions with 6= 0 obtained from the two-rod metrics are
singular since already their prolate and oblate vacuum versions have struts. The only
solutions without struts are the vacuum rings with U = k = 0, but symmetries (6.34)
(6.36) act trivially in this case and give = 0.

Solutions from point masses

Let us finally discuss solutions obtained from the Chazy-Curzon metric (4.14),(4.15).
This metric does not admit the prolate and oblate generalizations since the scale
symmetry (U, k) (U, 2 k) acts trivially only redefining the parameter m in (4.14).
Hence, there is only one vacuum solution in the case,
U =

m
,
R

k=

m2 2
,
2R4

(7.1)

p
with R = 2 + z 2 , inserting which to (6.34)(6.36) gives three families of solutions
with scalar field.
More possibilities exist for the two mass solution (4.15) because it contains three
free parameters m and m. Applying (U, k) (U, 2 k) again results in the trivial
rescaling of m but this time there is a non-trivial possibility to complexify the solution
via m i and hence
p
p
(7.2)
R = 2 + (z m)2 2 + (z i)2 = X iY = x i cos ,
the expression of X, Y in terms of , z being given by (5.10). Since
X iY =

X 2 + Y 2 ei Rei

with

tan() =

Y
X

(7.3)

there are two possibilities to keep U, k in (4.15) real, either to set m = M/2i which
gives


M
M 2 2
M 2 2 + z 2 + 2
U=
sin(), k =
cos(4) 2
1
(7.4)
R
4R4
8
R2

31

or to set m = M/2 which gives

M
cos(),
U=
R

M 2 2
M2
k=
cos(4)

4R4
82


2 + z 2 + 2
1 .
R2

(7.5)

These solutions satisfy the elementary flatness condition at the symmetry axis, since
k(0, z) = 0. However, they are all singular at the point (, z) = (, 0) where R vanishes.
Taking as the pair (U, k) either the original prolate two-point solution (4.15) or one
of the oblate solutions (7.4),(7.5) and injecting to (6.34)(6.36) gives nine threeparameter families of solutions with scalar field. The dimension of the solution space
can be doubled by further applying the tachyon symmetry (3.13).

Conclusions

To recapitulate, we applied above the duality rotations to the vacuum Weyl metrics to
produce new solutions with scalar field. We were mainly interested in the wormhole
solutions with several asymptotic regions. Such solutions are best known in systems
with exotic matter, but it is little known that they exist also in vacuum GR, being
sourced by thin negative tension rings. The one-ring solutions are completely regular
away from the ring and do not show pathologies like closed timelike curves. In this
respect the ring wormholes are different from other known solutions showing singular
rings/strings, like the Kerr black hole containing inside the horizon a ring singularity
which can also be viewed as a wormhole [52], or the NUT wormholes containing inside
a singular Misner string [53].
The vacuum ring wormholes could be viewed as primary since solutions with
scalar field are obtainable from them via duality rotations. This gives rings dressed
with scalar, conventional or phantom. In particular, the spherically symmetric BE
wormhole can be obtained by dressing the ring with the phantom field. As we have
seen, one can construct large families of such dressed up solutions. However, most
of them, excepting the one-ring oblate solutions, contain singular struts.
The multi-ring solutions become regular only in the limit where the modulus of
the ring tension attains the (same for all rings) minimal non-zero value and the metric
becomes exactly flat away from the rings. The topology then remains non-trivial and
shows several interconnected asymptotic regions, while the strut singularity along the
symmetry axes disappears. Such solutions cannot be dressed with scalar because the
dualities act trivially on them. This suggests once again that vacuum ring wormholes
could be viewed as fundamental.
In some parts of our discussion we summarised and generalised little known facts
scattered in the literature, whereas other parts are original, as for example the description of the scalar-dressed solutions obtained via duality transformations and of
the ring and multi-ring solutions with locally flat geometry. The main message we
were trying to convey is that traversable wormholes could be less exotic objects than
is usually thought, since they exist already in vacuum GR and not necessarily only in
systems with exotic matter.

32

Acknowledgements
G.W.G. thanks the LMPT for hospitality and acknowledges the support of Le Studium
Institute for Advanced Studies of the Loire Valley. M.S.V. was partly supported by
the Russian Government Program of Competitive Growth of the Kazan Federal University.

Isometric embeddings of the BE wormhole

p
x2 + 2 (sin cos , sin sin , cos ) and also Z,
 

Z
,
(A.1)
dZ = p
dx

x = sinh
2
2

x +

Introducing (X1 , X2 , X3 ) =

the ultrastatic BE solution (2.16) can be represented as the geometry induced on a

hypersurface in five dimensional Minkowski space,
ds2 = dt2 + dx2 + (x2 + 2 )d2 = dt2 + dX12 + dX22 + dX32 + dZ 2 ,

(A.2)

where
X12

X22

X32

 
Z
.
= cosh

(A.3)

It follows that the spatial 2-geometry of the equatorial plane X3 = 0 is a catenoid

Figure 8. Embedding of the equatorial section of the BE wormhole to the 3-dimensional

Euclidean space spanned by X1 , X2 , Z defined by (A.4),(A.5).

(see Fig.8) a surface of revolution in three-dimensional Euclidean space with the line
element
dX12 + dX22 + dZ 2 = dr2 + r2 d2 + dZ 2

(A.4)

whose meridional curve is the catenary,

 
Z
r = cosh
.

33

(A.5)

One can also construct the embedding of the non-ultrastatic BE geometry (2.15),
ds2 = A2 (x)dt2 +

dx2
+ r2 (x)(d2 + sin2 d2 )
B 2 (x)

(A.6)

with
/s

A(x) = B(x) = e

p
x2 + 2
,
r(x) =
B(x)

 
x
= arctan
,

(A.7)

following the procedure of Refs.[54], [55]. This geometry can be embedded into the
7-dimensional Minkowski space with the metric
ds2 = dX02 + dX12 + dX22 + dX32 + dX42 + dX52 dX62

(A.8)

by the following explicit formulas for Xk = Xk (t, x, , ):

X0 = A(x) sinh(t),
X1 = A(x) cosh(t),
X2 = r(x) sin cos , X3 = r(x) sin sin , X4 = r(x) cos ,
Z
Z p
dx
X5 =
,
X6 =
A02 (x) + r02 (x) dx ,
B(x)

(A.9)

where the prime denotes the derivative with respect to x.

References
[1] A. Einstein and N. Rosen, The particle problem in the General Theory of Relativity,
Phys. Rev. 48 (1935) 7377.
[2] A. Einstein, B. Podolsky and N. Rosen, Can quantum mechanical description of
physical reality be considered complete?, Phys. Rev. 47 (1935) 777780.
[3] J. Maldacena and L. Susskind, Cool horizons for entangled black holes, Fortsch. Phys.
61 (2013) 781811, [1306.0533].
[4] C. W. Misner and J. A. Wheeler, Classical physics as geometry: gravitation,
electromagnetism, unquantized charge, and mass as properties of curved empty space,
Annals Phys. 2 (1957) 525603.
[5] C. W. Misner, Wormhole initial conditions, Phys. Rev. 118 (1960) 11101111.
[6] M. Cvetic, G. W. Gibbons and C. N. Pope, Super-geometrodynamics, JHEP 03 (2015)
029, [1411.1084].
[7] Virgo, LIGO Scientific collaboration, B. P. Abbott et al., Observation of
gravitational waves from a binary black hole merger, Phys. Rev. Lett. 116 (2016)
061102, [1602.03837].
[8] M. Morris, K. Thorne and U. Yurtsever, Wormholes, time machines, and the weak
energy condition, Phys.Rev.Lett. 61 (1988) 14461449.
[9] M. Visser, Lorentzian wormholes: From Einstein to Hawking. AIP, 1996.

34

[10] S. W. Hawking and G. F. R. Ellis, The Large Scale Structure of Space-Time.

Cambridge Monographs on Mathematical Physics. Cambridge University Press, 2011,
10.1017/CBO9780511524646.
[11] J. L. Friedman, K. Schleich and D. M. Witt, Topological censorship, Phys. Rev. Lett.
71 (1993) 14861489, [gr-qc/9305017].
[12] D. Hochberg and M. Visser, The Null energy condition in dynamic wormholes, Phys.
Rev. Lett. 81 (1998) 746749, [gr-qc/9802048].
[13] K. Bronnikov, Scalar-tensor theory and scalar charge, Acta Phys.Polon. B4 (1973)
251266.
[14] H. G. Ellis, Ether flow through a drainhole - a particle model in general relativity, J.
Math. Phys. 14 (1973) 104118.
[15] P. Kanti, B. Kleihaus and J. Kunz, Wormholes in dilatonic Einstein-Gauss-Bonnet
Theory, Phys.Rev.Lett. 107 (2011) 271101, [1108.3003].
[16] K. Bronnikov and S.-W. Kim, Possible wormholes in a brane world, Phys.Rev. D67
(2003) 064027, [gr-qc/0212112].
[17] S. V. Sushkov and R. Korolev, Scalar wormholes with nonminimal derivative coupling,
Class.Quant.Grav. 29 (2012) 085008, [1111.3415].
[18] S. V. Sushkov and M. S. Volkov, Giant wormholes in ghost-free bigravity theory, JCAP
1506 (2015) 017, [1502.03712].
[19] G. Clement, Axisymmetric regular multiwormhole solutions in five-dimensional
General Relativity, Gen. Rel. Grav. 16 (1984) 477.
[20] H. Weyl, Zur Gravitationstheorie, Ann. der Physik 54 (1917) 117145.
[21] G. Clement, Axisymmetric multiwormholes revisited, Gen. Rel. Grav. 48 (2016) 76,
[1511.06249].
[22] G. Clement, Regular multiparticle solutions of Einstein-Maxwell scalar field theories,
Class. Quant. Grav. 1 (1984) 275.
[23] G. Clement, A class of stationary axisymmetric solutions of Einstein-Maxwell scalar
field theories, Class. Quant. Grav. 1 (1984) 283.
[24] G. Clement, A class of wormhole solutions to higher dimensional General Relativity,
Gen. Rel. Grav. 16 (1984) 131.
[25] A. Chodos and S. L. Detweiler, Spherically symmetric solutions in five-dimensional
General Relativity, Gen. Rel. Grav. 14 (1982) 879.
[26] D. Zipoy, Topology of some spheroidal metrics, J. Math. Phys. 7 (1966) 11371143.
[27] K. A. Bronnikov and J. C. Fabris, Weyl space-times and wormholes in D-dimensional
Einstein and dilaton gravity, Class. Quant. Grav. 14 (1997) 831842.
[28] G. Clement, Selfgravitating cosmic rings, Phys. Lett. B449 (1999) 1216,
[gr-qc/9808082].
[29] G. Clement, Spinning ring wormholes: a classical model for elementary particles?, in
Proceedings, 4th Alexander Friedmann International Seminar on Gravitation and
Cosmology: St. Petersburg, Russia, June 17 - 25, 1998, 1998. gr-qc/9810075.

35

[30] G. W. Gibbons and M. S. Volkov, Ring wormholes via duality rotations, Phys. Lett.
B760 (2016) 324328, [1606.04879].
[31] I. Z. Fisher, Scalar mesostatic field with regard for gravitational effects, Zh. Eksp.
Teor. Fiz. 18 (1948) 636640, [gr-qc/9911008].
[32] A. I. Janis, E. T. Newman and J. Winicour, Reality of the Schwarzschild singularity,
Phys. Rev. Lett. 20 (1968) 878880.
[33] G. W. Gibbons, Phantom matter and the cosmological constant, hep-th/0302199.
[34] A. Peres, Gravitational field of a tachyon, Phys.Lett. A7 (1970) 361362.
[35] L. S. Schulman, Gravitational shock waves from tachyons, Nuovo Cim. B2 (1971)
3844.
[36] J. R. Gott, Tachyon singularity - a spacelike counterpart of the schwarzschild black
hole, Nuovo Cim. B22 (1974) 4969.
[37] G. W. Gibbons and D. A. Rasheed, Dyson pairs and zero mass black holes, Nucl.
Phys. B476 (1996) 515547, [hep-th/9604177].
[38] W. Israel and K. Khan, Collinear particles and Bondi dipoles in General Relativity,
Nuov.Cim. XXXIII (1964) 331344.
[39] J. Chazy, Sur le champ de gravitation de deux masses fixes dans la theorie de la
relativite, Bull.Soc.Math.France 52 (1924) 1738.
[40] H. Curzon, Cylindrical solutions of einsteins gravitation equations, Proc.Math.Soc.
London 25 (1924) 477480.
[41] L. Silberstein, Two-centers solution of the gravitational field equations, and the need
for a reformed theory of matter, Phys. Rev. 49 (1936) 268270.
[42] N. Schleifer, Condition of elementary flatness and the two-particle curzon solution,
Phys.Let. A 112 (1985) 204207.
[43] A. Papapetrou, A static solution of the equations of the gravitational field for an
arbitrary charge distribution, Proc.Roy.Irish.Acad. A51 (1947) 191.
[44] S. D. Majumdar, A class of exact solutions of Einsteins field equations, Phys. Rev. 72
(1947) 390398.
[45] B. H. Voorhees, Static axially symmetric gravitational fields, Phys. Rev. D2 (1970)
21192122.
[46] H. Kodama and W. Hikida, Global structure of the Zipoy-Voorhees-Weyl spacetime
and the delta=2 Tomimatsu-Sato spacetime, Class. Quant. Grav. 20 (2003) 51215140,
[gr-qc/0304064].
[47] G. Lukes-Gerakopoulos, The non-integrability of the Zipoy-Voorhees metric, Phys. Rev.
D86 (2012) 044013, [1206.0660].
[48] A. I. Egorov, P. E. Kashargin and S. V. Sushkov, Scalar multi-wormholes, Class.
Quant. Grav. 33 (2016) 175011, [1603.09552].
[49] C. Schiller, Simple derivation of minimum length, minimum dipole moment and lack of
space-time continuity, Int. J. Theor. Phys. 45 (2006) 221235.

36

[50] G. W. Gibbons, The Maximum tension principle in general relativity, Found. Phys. 32
(2002) 18911901, [hep-th/0210109].
[51] S. Krasnikov, The Quantum inequalities do not forbid space-time shortcuts, Phys. Rev.
D67 (2003) 104013, [gr-qc/0207057].
[52] B. Carter, Global structure of the Kerr family of gravitational fields, Phys. Rev. 174
(1968) 15591571.
[53] G. Clement, D. Galtsov and M. Guenouche, NUT wormholes, Phys. Rev. D93 (2016)
024048, [1509.07854].
[54] C. Fronsdal, Completion and embedding of the Schwarzschild solution, Phys. Rev. 116
(1959) 778781.
[55] M. Ferraris and M. Francaviglia, An algebraic isometric embedding of Kruskal
spacetime, Gen.Rel.Grav. 10 (1979) 283296.

37