2.9 3.

Article

Falling into the Past: Geodesics in

a Time Travel Metric

Colin MacLaurin, Fabio Costa and Timothy C. Ralph

Special Issue
The Physics of Time Travel
Edited by
Dr. Ana Alonso-Serrano, Prof. Dr. Matt Visser, Dr. Jessica Santiago and Dr. Sebastian Schuster

https://doi.org/10.3390/universe10020095
 universe
Article
Falling into the Past: Geodesics in a Time Travel Metric
Colin MacLaurin 1 , Fabio Costa 1,2 and Timothy C. Ralph 3, *

1 School of Mathematics and Physics, University of Queensland, Brisbane, QLD 4072, Australia;
c.maclaurin@uq.edu.au (C.M.); fabio.costa@su.se (F.C.)
2 Nordita, Stockholm University and KTH Royal Institute of Technology, Hannes Alfvéns väg 12,
SE-106 91 Stockholm, Sweden
3 Centre for Quantum Computation and Communication Technology, School of Mathematics and Physics,
University of Queensland, Brisbane, QLD 4072, Australia
* Correspondence: ralph@physics.uq.edu.au

Abstract: We investigate timelike and null geodesics within the rotating “time machine” spacetime
proposed by Ralph, T.C.; et al. Phys. Rev. D 2020, 102, 124013. This is a rotating analogue of
Alcubierre’s warp drive spacetime. We obtain geodesics that begin and end in the surrounding flat
space region, yet achieve time travel relative to static observers there. This is a global property, as the
geodesics remain locally future-pointing, as well as timelike or null.

Keywords: general relativity; closed timeline curves; time machines

1. Introduction
Exotic spacetimes including wormholes, warp drives, and time machines have been
used to probe frontier physics concepts, despite being of physically dubious construc-
tion themselves. An early example is Gödel’s [1] rotating universe which has a negative
cosmological constant but motivated developments in relativistic causality [2]. More re-
cently, the Kip Thorne group has studied billiard ball collisions in the context of wormhole
spacetimes with closed timelike curves (CTCs) to see if a consistent mechanics may be
Citation: MacLaurin, C.; Costa, F.; determined from initial conditions [3,4]. Hawking argued that spacetimes of this type
Ralph, T.C. Falling into the Past: are unstable [5] as have others for different cases, including Krasnikov [6], but a more
Geodesics in a Time Travel Metric. general argument has not emerged. Other authors have studied quantum particles in the
Universe 2024, 10, 95. https:// presence of CTCs. Deutsch [7] described consistent evolutions using a quantum computa-
doi.org/10.3390/universe10020095 tion circuit model, which was clarified by Politzer [8] using a less abstract approach. An
Academic Editors: Ana Alonso-Serrano,
alternative approach to obtain consistent evolutions was described by Lloyd et al. [9] based
Matt Visser, Jessica Santiago and on teleportation.
Sebastian Schuster Closed timelike curves are spacetime trajectories that start and finish at the same space-
time event. Spacetimes that contain them may be referred to as time machine spacetimes,
Received: 15 December 2023 as they allow particles to travel into their own past. As well as in wormhole spacetimes,
Revised: 2 February 2024
CTCs have have been found in rotating spacetimes, including rotating cylinders (studied
Accepted: 10 February 2024
by van Stockum), and the Kerr spacetime [10] that describes rotating black holes (albeit
Published: 16 February 2024
within an inner region that some consider unphysical). Tippett and Tsang [11] suggested a
classification of time machine spacetimes into three classes. The first class is those that arise
naturally in geometries with strong angular momentum. The Gödel and Kerr spacetimes
Copyright: © 2024 by the authors.
previously mentioned are examples of this class. The second class is spacetimes designed
Licensee MDPI, Basel, Switzerland. specifically for time travel. These include metrics due to Ori and co-workers [12–14], the
This article is an open access article wormhole spacetimes discussed earlier, the spacetime of Mallary et al. [15], and Tippett
distributed under the terms and and Tsang’s spacetime [11]. The third class is spacetimes that are designed to produce
conditions of the Creative Commons superluminal travel but allow time travel as a natural consequence. The spacetime we
Attribution (CC BY) license (https:// investigate here is of this third type.
creativecommons.org/licenses/by/ Alcubierre [16] proposed a “warp drive” spacetime with a localized bubble region which
4.0/). travels faster than light as determined by a distant observer. Other warp drive spacetimes that

Universe 2024, 10, 95. https://doi.org/10.3390/universe10020095 https://www.mdpi.com/journal/universe

Universe 2024, 10, 95 2 of 14

have been proposed include those by Krasnikov [17], Olum [18], Natário [19] and Bobrick
and Martire [20]. Everett [21] and Krasnikov [17] considered boosting between two different
warp drive spacetimes and found that CTCs could be formed. Ralph and Chang [22] defined
a “time machine” spacetime by incorporating an Alcubierre-like warp bubble, made to
follow a circular loop, on a rotating platform. This allowed backwards-in-time travel for
certain parameters and hence the explicit formation of CTCs by a relatively simple metric.
The physical realizability of such a metric is uncertain due to the exotic matter required to
produce it, which appears to be a generic feature of warp drive spacetimes [23]. However,
its study may lead to useful advances in understanding.
In this paper, we study geodesics in the Ralph and Chang spacetime. Our main result
is that, for a particular choice of parameters, we find timelike and null geodesics which
transit the curved spacetime region and exit into flat space at an earlier time than when
they enter, according to an observer in flat space. Timelike geodesics with this property
have previously been identified by Fermi and Pizzocchero [24] for a different spacetime
(belonging to class two), albeit one with a similar structure. Whilst the Ralph and Chang
spacetime involves an Alcubierre “bubble” which travels around a circular path, here we
will consider parameters for which the front of the bubble merges with the rear, forming
a toric structure similar to [24]. Nevertheless, the metrics and their specific properties
are distinct.
Our paper is arranged in the following way. In the next section, we describe the time
machine metric, its derivation and how we parameterize it. In Section 3, we describe some
properties of the time machine metric before outlining our method for finding certain
classes of geodesics in Section 4. Section 5 describes our main results before we discuss and
conclude in Section 6.

2. The Time Machine Spacetime

Conceptually, the Ralph and Chang time machine spacetime is a combination of
Alcubierre spacetime, with a rotating disc in Minkowski spacetime.

2.1. Alcubierre Spacetime

Alcubierre [16] starts with coordinates (t, x, y, z) on a manifold, and a path xs (t) along
the x-axis, where the “s” stands for spaceship. The goal was to choose a metric such that
this path is timelike and geodesic. Alcubierre defined a region around the worldline, since
called the “bubble”, parametrized by a function f with value 1 on the worldline, and
approximately 0 far from it. This f characterizes the amount of warp, and is discussed
further in Section 2.4. vs (t) := dxs /dt is the coordinate speed of the spaceship. Using the
lapse–shift formalism (see [25], §21.4 for background), it suffices to use a shift 3-vector
⃗bi = (− f vs , 0, 0) and lapse α = 1. Then, the Eulerian observers, meaning those orthogonal
to the hypersurfaces t = const but which we name “warp” observers, have 4-velocity
µ
uwarp = (1, f vs , 0, 0). They are geodesic since their 4-acceleration is a function of the lapse
gradient [26], §4.3.3. For f = 1, they coincide with the spaceship worldline.
The coordinate speed vs has no limits. However, we stress there is never faster-than-
light motion locally, that is, a massive test particle’s 4-velocity is never null nor spacelike. It
has at most “‘effective’ superluminal travel” [27]. Furthermore, all valid 4-velocities are
future-pointing, never past-pointing.
Alcubierre [16] briefly mentioned the case of an arbitrary lapse and shift. Here, the Eu-
µ
lerian observers are uwarp = α−1 (1, −b1 , −b2 , −b3 ), with co-velocity warp uµ = (−α, 0, 0, 0).
We generalize this construction in Appendix C to allow observers with vorticity, which are
not orthogonal to any hypersurface. In our approach, Alcubierre spacetime is obtained
by starting with Minkowski spacetime, including the metric. Take the static observers
u := ∂t as a starting point, which is the coordinate basis vector (1, 0, 0, 0) in some chosen
inertial frame. Now, define a “warp shift” 4-vector field b := − f vs ∂ x , and “warp lapse”
scalar α = 1. These combine to add terms to the Minkowski metric, yielding the Alcubierre
metric (Appendix C). Formally, our model is a single manifold with two Lorentzian metrics.
Universe 2024, 10, 95 3 of 14

But physically, we interpret this as two separate spacetimes, or two separate regions (“time”
periods) within a single spacetime.

2.2. The Rotating Disc

Use cylindrical coordinates (t, r, ϕ, z) on Minkowski spacetime. A disc rotating at
constant (coordinate) rate dϕ/dt =: Ω ∈ R has a 4-velocity field:
µ
p
udisc = 1, 0, Ω, 0 1 − Ω2 r 2 . (1)

This is valid within the region r < 1/|Ω| for subluminal motion. Intuitively, this
field describes material elements making up the disc; however, our focus is on abstract
reference frames not materials science. The motion is Born-rigid, meaning its expansion
tensor vanishes: Θ = 0. At any given point, the Lorentz factor relative to a “lab” observer
∂t is γ = −⟨udisc , ∂t ⟩ = (1 − Ω2 r2 )−1/2 , where the angle brackets refer to the metric scalar
product. The following vector field points in the tangent direction:
p
ξ µ = (Ωr, 0, 1/r, 0) 1 − Ω2 r 2 . (2)

It has unit length ⟨ξ, ξ ⟩ = 1 and is purely spatial according to the disc observers,
meaning ⟨udisc , ξ ⟩ = 0. Intuitively, these model little rulers, but technically they are merely
vectors in an orthonormal pair.

2.3. Ralph and Chang Spacetime

The time machine spacetime family we consider was introduced and analyzed in Ralph
and Chang [22]. Here, we give a novel derivation starting with Minkowski spacetime with
the disc 4-velocity field, also called the “platform”, u := udisc . Next, add an Alcubierre-like
warp shift in the tangential direction for each observer b := − f vs ξ. Warp lapse α := 1.
vs ∈ R is taken as a constant, and our specific choices of bubble functions f are given in the
following section. The resulting metric is (Appendix C):

( f vs + Ωr )2 − 1 2 f vs r (1 + Ω2 r2 + Ωr f vs )
ds2 = dt − 2 dt dϕ
1 − Ω2 r 2 1 − Ω2 r 2
r2 ((1 + Ωr f vs )2 − Ω2 r2 ) 2
+ dr2 + dϕ + dz2 . (3)
1 − Ω2 r 2
This is Equation (5) in Ralph and Chang [22]. They call these the “lab” coordinates,
because where f = 0, the metric reduces to ds2 = −dt2 + dr2 + r2 dϕ2 + dz2 , which is locally
Minkowski spacetime, and ∂t reduces to the static “lab” observers. The value r = 1/|Ω| is
not only a coordinate singularity but a curvature singularity for the spacetime parameters
tested. This is the radius at which the spinning platform’s tangential velocity is the speed of
light. Hence, we enforce f = 0 in the region around r = 1/|Ω| throughout this paper, and
as a result, Equation (3) reduces to the Minkowski metric there. Placing the point r = 1/|Ω|
outside the region of curvature such that f = 0 in its vicinity is a sufficient condition for
removing the singularity. An interesting question for future investigation would be to find
necessary conditions such that the singularity is removed.
Note that t is not necessarily timelike. (This is determined by the hypersurfaces
t = const, which have normal covector dt, satisfying ⟨dt, dt⟩ = gtt , which may have any
sign.) Similarly, the coordinate basis vector ∂t is not necessarily timelike, and, likewise, ϕ
and ∂ϕ are not necessarily spacelike. On a partly related note, the standard lapse and shift
quantities relative to the t = const hypersurfaces differ from our “warp lapse” and “warp
shift”, even when t is timelike.
Universe 2024, 10, 95 4 of 14

The metric of Equation (3) may be summarized as the following, for brevity:
 
gtt 0 gtϕ 0
 0 1 0 0
gµν =
 gtϕ
. (4)
0 gϕϕ 0
0 0 0 1

2.4. The Bubble Function f

The time machine spacetime was originally conceived as a warp bubble revolving
about the origin [22]. If we write Ω̃ for the bubble’s (coordinate) rotation rate dϕ/dt, a
constant independent of the platform rate Ω, then we might expect f to have the form:

f = f rot (ϕ − Ω̃t) f r (r ) f z (z), (5)

where f z is symmetric. However, in this paper, we drop the angular dependence for
simplicity, leaving f rotationally symmetric. This can be thought of as the bubble reaching
around the entire cylinder. Also, we do not give our f an explicit time dependence. Hence,
it is a function of r alone, and in principle also of z. This describes a torus rather than a
bubble.1 Here, we use the bump functions in Figure 1, which vanish beyond a certain fixed
r. The Gaussian-inspired function has the definition:
  
exp 1 − 1
1−(r −2) 2 1<r<3
f (r ) = (6)
0 otherwise,

whereas the plateau function is:


1

1 1

1<r<2
1+exp r −1 + r −2





1 2≤r≤3
f (r ) = 1
(7)

3<r<4
− r−1 3 − r−1 4



 1+exp

0 otherwise.

These are smooth but not analytic. We define the time machine region of spacetime by
f ̸= 0, which coincides with the curved region.

f(r) f(r)
1
1

1 2 3
r 1 2 3 4
r

(a) (b)

Figure 1. Bump functions used for the bubble function f . The dotted lines indicate distinctive values
of r. The choice of integer values is for simplicity, and also hints that f need not be finely tuned
to achieve time travel. (a) A Gaussian curve squeezed into a finite domain. (b) A plateau function
with ramps.

We can think of the z-dependence of f as being a broad plateau function centered on

z = 0. However, we will start all our trajectories in the z = 0 plane, where they will remain,
so the exact details of the z dependence will not play a role in our calculations. Similarly,
the t dependence of f can be thought of as a broad plateau function centered on (but much
broader than) the time interval of interest for the trajectories. In other words, we might
envisage that at some time in the past, the value of f was slowly increased from 0 → 1. It
then remained constant for an extended period, which includes the time of interest in this
paper. At sometime in the future, it may be slowly decreased from 1 → 0, returning the
Universe 2024, 10, 95 5 of 14

relevant spacetime to Minkowski. Hence, for the purposes of our calculations, f is taken to
be constant in t and z.

3. Time Orientation and Warp Velocity

In Alcubierre spacetime, the Eulerian observers have co-velocity warp uµ = (−1, 0, 0, 0).
We observe that these are the same components as static observers in Minkowski space-
time, with whom our construction started. It turns out this is not a coincidence as
shown in Appendix C. To obtain the analogous field in the time machine spacetime,
first lower the index on the disc 4-velocity (Equation (1)) using the Minkowski metric, so
√
disc u = (−1, 0, Ωr2 , 0) 1 − Ω2 r2 in cylindrical coordinates, which is also warp uµ . Now,
µ
raise the index using the time machine metric:

µ 1  f vs 
uwarp := √ 1 + Ωr f vs , 0, Ω + ,0 . (8)
1 − Ω2 r 2 r

We denote this as the warp velocity. Equivalently, uwarp = udisc − b. It is unit timelike
in the new metric: ⟨uwarp , uwarp ⟩ = −1. Where f = 1, it reduces to the bubble velocity
defined in Ralph and Chang [22], though we have swapped the order of r and ϕ. Sharing
the same co-velocity vector with the disc leads to interesting properties. For example, uwarp
has the same Killing energy and angular momentum (defined in the following section) as
the rotating disc frames do in Minkowski spacetime. One downside is that Equation (8) is
only valid for r < 1/|Ω|.
Another possibility is a local Lorentz boost of the warp observers by speed β := Ωr
“backwards”, which intuitively undoes the speed of the platform. Now, the vector ξ points
in the tangential direction; while this is formally just the disc ruler vector from Equation (2),
it also satisfies ⟨ξ, ξ ⟩ = 1 and ⟨uwarp , ξ ⟩ = 0 in the new metric. The local boost in the
negative ξ-direction is γ(uwarp − βξ ), or:

µ
 Ωr f vs f vs 
uorient := 1 + , 0, , 0 , (9)
1 − Ω2 r 2 r (1 − Ω2 r 2 )

where ⟨uorient , uorient ⟩ = −1. This is an ideal time orientation, meaning a continuous timelike
vector field. The expression remains valid for r > 1/|Ω|, so we extend its range beyond
the initial construction. In the flat space region f → 0, it reduces to the lab observers ∂t ;
hence, it agrees with the natural requirement that t increases towards the future in that
region. It also agrees with the time orientation specified by the warp velocity on their
region of overlap because they are related by a subluminal boost. A key point is that the
time machine metric of Equation (3) is valid for r < 1/|Ω| and r > 1/|Ω|, expanding the
domain of applicability from what was considered in Ralph and Chang [22]. The metric
retains a Lorentzian signature −+++ so it is a valid spacetime, despite not necessarily being
constructed from a timelike platform. This is used in Section 5.2.
µ
One may define the platform velocity as uplat := (1, 0, Ω, 0) in the time machine
spacetime (Appendix C). This may be timelike, null, or spacelike, and is not normalized
as written. Similarly, for a revolving bubble, we may define the bubble velocity
µ
ububble := (1, 0, Ω̃, 0), which is also un-normalized.

4. Method for Finding the Geodesics

4.1. Killing Vector Fields
It is natural to classify geodesics and other curves using Killing vector fields to what-
ever extent this is possible. For the general f in Equation (5), the bubble velocity (1, 0, Ω̃, 0)
is a Killing vector field. However, we drop the angular dependence, so the metric compo-
nents are independent of t and ϕ. Hence, ∂t is a Killing vector, and gives rise to an “energy”
which is conserved along the geodesics:

e := −⟨u, ∂t ⟩. (10)
Universe 2024, 10, 95 6 of 14

Here, u is the tangent to a curve parametrized by proper time or proper length if

non-null. Since ∂t is not necessarily timelike, negative Killing energy is possible, but this is
merely kinematics in contrast to the intrinsic negative energy of exotic matter. In addition,
the Killing vector field ∂ϕ defines a natural angular momentum ℓ which is conserved
along the geodesics:
ℓ := ⟨u, ∂ϕ ⟩. (11)
Any trajectory with z = 0 everywhere is classified by e and ℓ at each point, along with
the sign of the r component. Its tangent is expressed most simply with a lowered index:
s !
e2 gϕϕ + 2eℓ gtϕ + ℓ2 gtt
uµ = − e, ± ϵ+ , ℓ, 0 , (12)
−g

where ϵ := ⟨u, u⟩ is −1/0/1 for a timelike/null/spacelike vector respectively, and

g := det( gµν ) evaluates to −r2 for the time machine spacetime. However, this is only
a “candidate” velocity, as it must satisfy several tests to be considered physical as follows.

4.2. Inequalities
We consider three tests of the tangent to a curve, ranging from essential down to desirable.
1. Radicand test. Equation (12) involves a square root, so the radicand must be non-
negative. Otherwise, one’s choice of e, ℓ and ϵ does not define a valid trajectory at
that particular spacetime point. With the time machine metric components, this is:

ℓ2 2
2 f vs (ℓ − eΩr ) − 2r ( e − Ω ℓ)
ϵ + e2 − + f v s (ℓ − eΩr ) ≥ 0. (13)
r2 r 2 (1 − Ω2 r 2 )

2. Future-pointing test. Timelike and null tangents are required to be future-pointing if

interpreted as physical particles. This is the local property: ⟨u, uorient ⟩ < 0, which
utilizes Equation (9), giving:

ℓ − eΩr2
f vs − e < 0. (14)
r (1 − Ω2 r 2 )

3. Backwards-in-t test. A desirable property of timelike and null tangents is to have ut < 0
for at least part of the curve. We seek paths with time travel, in the sense of returning
to a lab observer worldline at an earlier lab time. For lab observers, t is the proper
time, and t must increase in the flat region. Note that we make no insinuation of
global simultaneity but have merely used the continuity of t on the manifold. Now,
ut < 0 is:
Ωr (e − Ωℓ) + (eΩr2 − ℓ)(1 + Ωr f vs )/r
e + f vs < 0. (15)
1 − Ω2 r 2
This trio of simultaneous inequalities is solvable as a large number of separate cases
if we treat f as being independent of r as a first step. One may also show some simpler
statements, for example, the restriction r < 1/|Ω| implies Ωvs < 0, so these parameters
have opposite signs. In the flat region f = 0, timelike and null test particles must have
positive energy, an increasing t coordinate, and ϵ + e2 − ℓ2 /r2 ≥ 0.

4.3. Circular Paths

Circular trajectories are a simpler special case and aid in the discovery of other trajec-
tories. With r = const and z = 0, the candidate tangent is:
q
ℓ gtϕ − ± − g(ℓ2 − ϵgϕϕ )
!
uµ = , 0, ℓ, 0 . (16)
gϕϕ
Universe 2024, 10, 95 7 of 14

The ‘±’ indicates there are two possible solutions for the circular trajectories. It can be
shown to correspond to the sign of ut , at least when the latter is nonzero. Hence, it must
be −1 to pass the backwards-in-t test. Note that e and ℓ are not independent because of
the constraint of circular motion; rather, e = −ut may be read off from Equation (12).2 The
radicand test is ℓ2 ≥ ϵgϕϕ , and the future-pointing test is ⟨u, uorient ⟩ < 0.
Again, these simultaneous inequalities are solvable, assuming that, at first, f is
independent. For instance, in the case ϵ = −1 and r < 1/|Ω|, we must have either
Ωr f vs < −(1 + |Ω|r ); or else −(r + 1/Ω) < r f vs < (r − 1/Ω) together with:
s
Ω2 r2 − (1 + Ωr f vs )2
Ω ℓ ≥ | Ω |r . (17)
1 − Ω2 r 2

Whilst for many cases now, requiring ℓ and e to be constant along a trajectory is
necessary and sufficient to find the geodesics (such as the free-fall cases considered in the
next section), this is not sufficient when circular motion is imposed. Given that ℓ and e are
constant along a circular path, we further require that the particle feels no acceleration. The
vector field a := ∇u u, called the 4-acceleration for ϵ = −1, is directed radially:
q
′ /g ) ± ℓ
ℓ2 (1 − rgϕϕ ′ − g g′ /g ) + g
ℓ2 + gϕϕ ( gtϕ 1 ′
 ϕϕ tϕ ϕϕ ϕϕ ϕϕ − 2 rgϕϕ 
aµ = 0, , 0, 0 . (18)
rgϕϕ

A dash means derivative with respect to r. We used the determinant of the metric
g = −r2 . It is computationally simpler to obtain the co-acceleration using a♭ = u⌟du♭
(Appendix B). Now, a = 0 for an affinely parametrized geodesic. This does not typically
have an exact solution for r. But conversely, if the spacetime parameters and a specific
value of r are substituted, then this rearranges to a quadratic in ℓ2 —that is, a quartic with
terms in ℓ4 , ℓ2 , and 1. See Figure 2 for example solutions.
The abundance of circular geodesics suggests a strategy for time travel. Set up a test
particle with ℓ and r chosen from the figure, and the corresponding e for circular motion,
perhaps via a collision. It will accumulate negative t indefinitely, up to the time machine’s
construction, or aside from any orbital instabilities. Then, arrange a collision to bump the
particle out. From a lab frame, this may mean orchestrating the exit collision before the test
particle is sent in. A different strategy is to pick a trajectory which is not circular but has
similar parameters to a those of a circular geodesics. Most of our examples will be like this,
starting from the flat region, making several roughly circular loops in the curved region
and then exiting.

ℓ
3

1.5 2.0 2.5 3.0

r

-1

-2

-3

Figure 2. Timelike circular geodesics for Ω = −1/5, vs = 7, and the Gaussian-like f . The brown
curves give the angular momentum required at each r. There are four candidate geodesics just
above r = 1, and none near r = 5/2 nor just below r = 3. The dashed parts of the curves are not
time-traveling: they are either past-pointing or have ut > 0. The aqua curve is the warp velocity,
included for comparison. The flat region(s) are not shown; there, the only constant–r geodesics have
ℓ = 0, the lab observers.
Universe 2024, 10, 95 8 of 14

5. Free Fall to the Past

A very interesting situation is if particles can free fall from flat space into the time
machine metric region and scatter back into flat space in such a way that they emerge in
the past. We find such trajectories are possible given particular conditions for the metric
and the initial trajectories. These trajectories do not form CTCs as such; indeed, closed
loops are not what we desire for time travel. However, it is important to demonstrate in an
operational way that there are events on the outward part of the trajectory that are indeed
in the past of events on the inward part of the trajectory. We introduce a correction term
that achieves this.
We start from the geodesic velocity field defined by Equation (12) for parameters
where the physicality conditions, Equations (13) and (14), are satisfied. Then, for specific
initial conditions (i.e., a particular incoming geodesic in flat space), the relevant geodesic in
the region of curvature is calculated numerically from the velocity field. Timelike geodesics
are found when ϵ = −1 and null geodesics when ϵ = 0.

5.1. Timelike Geodesic Examples

In this section, we present timelike time-traveling geodesics in the spacetime with
parameters Ω = −1/5, vs = 7, and the Gaussian-like f of Equation (6). The case of circular
geodesics for the same parameters was already considered in Figure 2; however, these are
stuck inside the time machine. Figure 3 exhibits a geodesic that achieves time travel, while
starting and ending in the flat region.
In Figure 3b, the exit (orange) curve typically has a lesser t-value than the entry (green)
curve, within the region shown. The t–r plot has striking steepness at r ≈ 2.64, which
corresponds to the outer loop in Figure 3a. Here, the trajectory parameters are close to
those of a circular geodesic at that point, including the e; this appears as a point near the
lower-right in Figure 2.
One way to unambiguously achieve time travel is if the particle returns to a given
lab location at an earlier lab time. Now, a lab worldline appears vertical in the spacetime
diagram; however, most apparent intersections actually have angular separation in ϕ. In the
flat region, it is natural to use t as a simultaneity convention, as this is a proper time for the
lab observers and is Poincaré–Einstein synchronized, meaning t = const is orthogonal to ∂t .
This justifies calling it an earlier lab time; however, this is largely just a convention. Physical
or operational evidence would be the ability to send a causal signal from an outgoing event
to an ingoing event. Suppose a lab observer is at some location along the ingoing curve.
Time travel is achieved if there exists a timelike or null curve from the end (orange curve,
later particle proper time) back to the start, arriving before the particle begins.
Suppose our causal signal is a null signal, sent along a circular arc at the time machine
boundary, guided by an optical fiber, say. This curve segment takes lab time ∆tarc = r |∆ϕ|,
where the angular separation is at most π. Because we have not incorporated all possibil-
ities, any conclusions that a given test particle trajectory achieves time travel should be
understood as a sufficient condition but not necessary. Another approach is to compute the
finite Minkowski interval between two points; this gives the length of a signal geodesic if
the entire line passes through flat space. Another possibility is to send a geodesic signal
though the curved region, but this calculation is more involved.
For the example trajectory, at r = 3, the t-difference is ≈−6.5. Assuming our geometric
units are meters, then 1 m of time is 1 m/c ≈ 3.3 nanoseconds in SI units. The angular
separation of ≈25◦ (see Figure 3a) wastes ≈ 4.4 ns, but in total the signal arrives with
∆t ≈ −17 ns, which is before the particle leaves! For any readers creating their own plots,
other time-traveling timelike geodesics with the Gaussian f include Ω = −1/5, vs = 4.85,
e = 1.9, and ℓ = −2.08; or Ω = −1/4, vs = 5.5, e = 1.5, and ℓ = −3.22. These are adjusted
so the ϕ-values roughly coincide at the boundary r = 3.
Universe 2024, 10, 95 9 of 14

t
4

0
1 2 3 4 5 6
r

-2

-4

(a) (b)

Figure 3. Example timelike geodesic with e = 2, ℓ = −2.28, and z = 0. The “space” diagram shows
the test particle entering from the right, moving inwards along the green curve. The dotted lines
mark significant values of f , as in Figure 1a, with the outermost one being the boundary of the time
machine region. Soon after entering, the particle is swept around. Its exit path where r is increasing
is colored orange. The “spacetime” diagram shows the same trajectory. In the flat space region, its
t-coordinate must increase. But inside the time machine, its t-value mostly decreases, in this case.
Yet, the trajectory remains future-pointing, i.e., the arrows on the trajectories indicate the direction of
increasing proper time for the particle. It exits at an earlier t than it entered, but t must increase again
thereafter. Still, a well-positioned lab observer would see the particle exit before it entered, so time
travel is achieved. (a) “Space” diagram, the r–ϕ plane. (b) “Spacetime” diagram.

To summarize the situation depicted in Figure 3a, the region of curvature is a cylinder
of outer radius 3 m and inner radius 1 m. The particle moves in the z = 0 plane. If the
particle was to send a signal as it exits the region of curvature (orange curve, just outside
the outer dotted radius), it would arrive at the entry point of the particle (green curve, just
outside the outer dotted radius) approximately 17 nanoseconds before it enters.

5.2. Null Geodesic Examples

In this section, we consider null geodesics in the spacetime with parameters Ω = 1.2,
vs = 1, and the plateau f from Equation (7). Hence, in this section, the region of curvature
is a cylinder of outer radius 4 m and inner radius 1 m. Note that this means our time
machine spacetime ( f ̸= 0) is in the region r > 1/|Ω| for this section. Figure 4 shows an
example trajectory chosen for its significant time travel. At the time machine boundary
r = 4, the difference between the t coordinates for the ingoing and outgoing path segments
is −46 m to the nearest integer, or −152 ns. The angular difference here is 74◦ , which would
take 17 ns for a causal signal to circumnavigate. The total time travel in our reckoning is
still a respectable ∆t = −135 ns.

20

10

0
1 2 3 4 5
r

-10

-20

(a) (b)

Figure 4. Example null geodesic with ℓ/e = 0.8337 and z = 0. As before, the dotted lines mark
significant values of f , in this case from the plateau bump function as in Figure 1b. The test particle
completes many loops at the plateau itself where f = 1, accruing a lot of negative t. Note the vertical
axis scale on the spacetime diagram. (a) “Space” diagram, the r–ϕ plane. (b) “Spacetime” diagram.
Universe 2024, 10, 95 10 of 14

We focus on null geodesics, which start from the flat region, and hence have positive
energy: e > 0. Conveniently, their paths are classified by impact parameter ℓ/e, up to
rotational symmetry and the sign of ur , and assuming constant z = 0. A useful property is
the “turning point”, where ur changes sign. This occurs at ur = 0, with solution:

ℓ − gtϕ ± r 1 + Ωr f vs ∓ Ωr
= =r . (19)
e gtt Ωr + f vs ∓ 1

Figure 5 plots the turning points for the present spacetime parameters. In Minkowski
spacetime or Newtonian physics, these would simply form a wedge ℓ/e = ±r, and
correspond to the point of closest approach of a straight line to the origin. For the time
machine with plateau f , this wedge graph does appear for r ≤ 1 and r ≥ 4. Also, we see
from inspection that none of the outside null geodesics can reach the origin. The closest is
for ℓ/e just above 5/6, which reach to just beyond r = 5/6. There are two critical values of
ℓ/e at 5/6 and ≈3.85, where the minimum r attained by geodesics changes abruptly. This
discontinuity is apparent in the diagram. Near the former value will turn out to be a rich
parameter space for time travel. The latter corresponds to a turning point of Equation (19),
which occurs at r ≈ 3.82, with corresponding ℓ/e ≈ 3.85 as mentioned, which in turn has
another intercept at r ≈ 2.70.

ℓ/e

1 2 3 4 5
r

-2

-4

Figure 5. Turning points for null geodesics, where the blue plot satisfies ur = 0 (Equation (19)). A
sample trajectory is shown, which approaches inwards from large r. Since its impact parameter is
constant, its worldline traces a horizontal line in the diagram. The first point of intersection with the
blue curve corresponds to reaching its minimum r before traversing away, in general. The unbroken
curves are the rightmost parts; only these are relevant to geodesics approaching from large r.

Figure 6 shows the amount of time travel for null geodesics. Similarly to before, we
define this as the difference in the t coordinate between the entry and exit points on the
boundary r = 4, plus a penalty of r |∆ϕ| for any angular offset, being the lab time (t) for a
causal signal to return to the initial angle via a null circular arc. By definition, ∆t = 0 for
ℓ/e = ±4 because these trajectories just graze the boundary.
Figure 7 exhibits the time travel to the past region by zooming in on Figure 6. From
analytic and numerical investigation, it appears the number of loops increases without
limit as ℓ/e → 5/6 from above, as does the time travel. In the figure, some points approach
1 µs of time travel. As before, having ∆t < 0, as we have defined ∆t, is a sufficient but not
necessary condition for time travel because of our restricted choice of signal mechanism.
Universe 2024, 10, 95 11 of 14

ℓ/e

-10 -5 5 10 15 20
Δt

-2

-4

Figure 6. Total time travel for null paths in terms of impact parameter, with the same spacetime
parameters Ω = 1.2, vs = 1, and plateau f . Some specific trajectories are shown as insets. Most of
the parameter space has ∆t > 0, meaning time travel is not achieved; however, see Figure 7 for the
exceptions. The extended horizontal lines correspond to the critical values of ℓ/e discussed previously.

ℓ/e
0.87

0.86

0.85

0.84

-100 -10 -1 1 10
Δt

Figure 7. Time traveling null geodesics, a zoom in on Figure 6. In this region, the null curves make
multiple loops, and most accrue substantial time travel—note the logarithmic scale. This reveals
a sawtooth fractal pattern. The “teeth” on the left side are the local minima of ∆t, that is, the local
maxima of time travel. These correspond to paths which exit at the same ϕ as they entered, at least
approximately. The teeth on the right are paths where the exit ϕ value is roughly opposite, which
wastes signal travel time. The horizontal line is ℓ/e = 5/6 or 0.83̄.

6. Discussion
We have discovered and analyzed certain classes of geodesic trajectories in the Ralph
and Chang time machine spacetime, which result in time travel. The time travel is a
large-scale property of a trajectory. As we have defined it, it occurs relative to static
observers in the flat spacetime region.We emphasize again, that the trajectories are locally
future-pointing everywhere, and remain timelike or null.
Most geodesics are not time-traveling. Those which are most successful, which com-
plete many loops inside the time machine region, can be very sensitive to parameters. On
the other hand, we found some very generic choices of bubble functions f were sufficient,
without any fine-tuning. For the case of the massive particle geodesics, it would be inter-
Universe 2024, 10, 95 12 of 14

esting future work to look at the local environment (and tidal forces) of the particle as it
traverses the trajectory. Another interesting future direction would be to try to engineer a
collision between the ingoing and outgoing trajectories, thus forming an explicit geodesic
CTC similar to the type discussed by Thorne et al. [3]. This would presumably require fine
tuning of the parameters.
We also further analyzed the spacetime itself, obtaining a time orientation, and point-
ing out some Killing vector(s). We have also made a small contribution to the physical
interpretation of spacetimes involving rotation, which has been a longstanding challenge
in some cases. We also generalized Alcubierre’s construction based on hypersurface–
orthogonal observers to allow vorticity.

Author Contributions: Conceptualization; methodology; validation, C.M., F.C. and T.C.R.; writing—
original draft preparation, C.M.; writing—review and editing, F.C. and T.C.R.; supervision, T.C.R. All
authors have read and agreed to the published version of the manuscript.
Funding: This work was partially supported by the Australian Research Council Centre of Excellence
for Quantum Computation and Communication Technology (Project No. CE110001027). Nordita is
supported in part by NordForsk.
Data Availability Statement: No new data were created or analyzed in this study. Data sharing is
not applicable to this article.
Conflicts of Interest: The authors declare no conflicts of interest.

Abbreviation
The following abbreviations are used in this manuscript:

CTCs closed timelike curves

Appendix A. Relative Velocity

Consider an observer u, and a vector n at the same point in spacetime. The latter may
be timelike, null, or spacelike. Recall velocity is displacement over time. In u’s frame of
reference, define the relative velocity of n as its space part divided by its time part, where
this splitting is determined by u. The pure time part of n is −⟨n, u⟩u and pure space part:
n + ⟨n, u⟩u, an orthogonal decomposition. The relative velocity ⃗v follows from dividing
these, where we use only the magnitude of the time part:

1
⃗v = − n − u. (A1)
⟨n, u⟩

Note that ⟨u, ⃗v⟩ = 0, so the relative velocity is purely spacelike according to u (or
zero). Its length is the relative speed β, which is not limited by 1. Equation (A1) is invariant
under the rescaling of n. If u and n are orthogonal, then ⃗v is not defined, and one might
perhaps say the relative velocity is infinite, or more cautiously that n is purely spacelike
according to our observer. Jantzen et al. [28], §2.1.4 give a similar result to Equation (A1),
and also allow n to be timelike, null, or spacelike. Misner et al. [25], Equation (2.35), have
essentially this formula for the special case of timelike n.
Reversing the problem, we can rearrange it to obtain:

n ∝ u +⃗v. (A2)

This shows how to construct a vector n with supplied relative velocity, subject to
⟨u, u⟩ = −1 and ⟨u, ⃗v⟩ = 0. If the result is non-null, it may then be normalized. This
generalizes the local Lorentz boost to the null and spacelike cases.
Universe 2024, 10, 95 13 of 14

Appendix B. 4-Acceleration Formula

For a 4-velocity field u, its 4-acceleration is a := ∇u u, where ∇ is the Levi–Civita con-
nection. The following is equivalent but computationally simpler [29], §7.8 and [28], §2.2.3:

a♭ = u⌟du♭ . (A3)

The flat symbol ‘♭ ’ here denotes the covector corresponding to a vector, which in
components is simply aµ or uµ . The notation ‘⌟’ is from differential forms, and means
the vector u is substituted into the 2-form d(u♭ ). In a coordinate basis, we have simply
aν = uµ (∂µ uν − ∂ν uµ ). To show Equation (A3):

du♭ ≡ ∇u♭ − (∇u♭ ) T (A4)

♭ ♭
= −u ∧ a + 2ω, (A5)

using the kinematic decomposition. The result follows. Here, ‘T ’ is the transpose. We used
the index-ordering convention ∇µ uν .

Appendix C. Generalized Warp Field Construction

Given any spacetime with metric, we “add” a warp field to obtain a new metric.
This is based on Alcubierre’s application of the lapse–shift formalism but generalized to
a non-coordinate basis. This might be contrasted with the Kerr–Schild ansatz or Gordon
form. Suppose a 4-velocity field u is provided, along with some “warp shift” 4-vector field
b satisfying ⟨u, b⟩ = 0, and “warp lapse” α. The specific values used for the Alcubierre
and time machine spacetimes are given in Sections 2.1 and 2.3. Whereas the standard
lapse–shift construction takes 3-dimensional metrics on hypersurfaces and welds them into
a 4-dimensional metric, we take one 4-dimensional metric and define a new one:
warp
gµν := 1 − α2 + ⟨b, b⟩ uµ uν − uµ bν − bµ uν +original gµν


(A6)

Note that in Equation (A6) (and throughout this Appendix), all angle brackets and
the raising or lowering of indices refer to the original metric, original gµν . By design, the
metric components match the standard lapse–shift metric [25], §21.4, when expressed in
an adapted basis (u, e1 , . . .). (The ei satisfy ⟨u, ei ⟩ = 0, but need not be constructed.) The
extra ‘1’ appears because the spatial metric is u♭ ⊗ u♭ + ds2original . The dual co-basis is
(−u♭ , e1 , . . .). The analogue of the Eulerian observers is (u − b)/α, which is dual to αu♭
under the new metric, and normalized. We label it the warp velocity (c.f. Section 3).
What happens to the original observers u after the warp field is added, for example,
the platform? Physically, it seems questionable to claim a correspondence between vectors
in different spacetime regions. Hence, we define those observers to have relative velocity
b/α in the warp frame. Note that this is orthogonal to the warp velocity in the new
metric. Hence, add them to obtain u/α (see Equation (A2)). This vector has squared-norm
⟨b, b⟩/α2 − 1, so it is not necessarily timelike. (This expression uses the warp p metric,
despite being expressed in terms of the original metric.) The relative speed is ⟨b, b⟩/α,
which is possibly superluminal.

Notes
1 Intuitively, the bubble is a topological ball, at least under the original Minkowski metric as restricted to a hypersurface t = const.
Alcubierre’s f only approaches 0 asymptotically, and hence is not compactly supported, but one might consider instead f ≥ ϵ̃ for
some small constant.
2 For the special case gϕϕ = 0, we have instead:
 g
tt ϵgtϕ 
uµ = ℓ + , 0, ℓ, 0 .
2gtϕ 2ℓ

Note that for gϕϕ = 0, both gtt and gtϕ must be nonzero for a nondegenerate metric, and ℓ ̸= 0 for any nonzero vector.
Universe 2024, 10, 95 14 of 14

References
1. Gödel, K. An Example of a New Type of Cosmological Solutions of Einstein’s Field Equations of Gravitation. Rev. Mod. Phys.
1949, 21, 447–450. [CrossRef]
2. Ellis, G.F.R. Editor’s Note. Gen. Relativ. Gravit. 2000, 32, 1399–1408. [CrossRef]
3. Echeverria, F.; Klinkhammer, G.; Thorne, K.S. Billiard balls in wormhole spacetimes with closed timelike curves: Classical theory.
Phys. Rev. D 1991, 44, 1077–1099. [CrossRef] [PubMed]
4. Friedman, J.; Morris, M.S.; Novikov, I.D.; Echeverria, F.; Klinkhammer, G.; Thorne, K.S.; Yurtsever, U. Cauchy problem in
spacetimes with closed timelike curves. Phys. Rev. D 1990, 42, 1915–1930. [CrossRef] [PubMed]
5. Hawking, S.W. Chronology protection conjecture. Phys. Rev. D 1992, 46, 603–611. [CrossRef] [PubMed]
6. Krasnikov, S. Time machines with the compactly determined Cauchy horizon. Phys. Rev. D 2014, 90, 024067. [CrossRef]
7. Deutsch, D. Quantum mechanics near closed timelike lines. Phys. Rev. D 1991, 44, 3197–3217. [CrossRef] [PubMed]
8. Politzer, H.D. Path integrals, density matrices, and information flow with closed timelike curves. Phys. Rev. D 1994, 49, 3981–3989.
[CrossRef] [PubMed]
9. Lloyd, S.; Maccone, L.; Garcia-Patron, R.; Giovannetti, V.; Shikano, Y.; Pirandola, S.; Rozema, L.A.; Darabi, A.; Soudagar, Y.;
Shalm, L.K.; et al. Closed Timelike Curves via Postselection: Theory and Experimental Test of Consistency. Phys. Rev. Lett. 2011,
106, 040403. [CrossRef] [PubMed]
10. Visser, M. The Kerr spacetime: A brief introduction. arXiv 2007, arXiv:0706.0622.
11. Tippett, B.K.; Tsang, D. Traversable acausal retrograde domains in spacetime. Class. Quantum Gravity 2017, 34, 095006. [CrossRef]
12. Ori, A. A Class of Time-Machine Solutions with a Compact Vacuum Core. Phys. Rev. Lett. 2005, 95, 021101. [CrossRef]
13. Ori, A. Formation of closed timelike curves in a composite vacuum/dust asymptotically flat spacetime. Phys. Rev. D 2007,
76, 044002. [CrossRef]
14. Soen, Y.; Ori, A. Improved time-machine model. Phys. Rev. D 1996, 54, 4858–4861. [CrossRef]
15. Mallary, C.; Khanna, G.; Price, R.H. Closed timelike curves and ‘effective’ superluminal travel with naked line singularities. Class.
Quantum Gravity 2018, 35, 175020. [CrossRef]
16. Alcubierre, M. Letter to the editor: The warp drive: Hyper-fast travel within general relativity. Class. Quantum Gravity 1994,
11, L73–L77. [CrossRef]
17. Krasnikov, S.V. Hyperfast travel in general relativity. Phys. Rev. D 1998, 57, 4760–4766. [CrossRef]
18. Olum, K.D. Superluminal Travel Requires Negative Energies. Phys. Rev. Lett. 1998, 81, 3567–3570. [CrossRef]
19. Natário, J. Warp drive with zero expansion. Class. Quantum Gravity 2002, 19, 1157–1165. [CrossRef]
20. Bobrick, A.; Martire, G. Introducing physical warp drives. Class. Quantum Gravity 2021, 38, 105009. [CrossRef]
21. Everett, A.E. Warp drive and causality. Phys. Rev. D 1996, 53, 7365. [CrossRef] [PubMed]
22. Ralph, T.C.; Chang, C. Spinning up a time machine. Phys. Rev. D 2020, 102, 124013. [CrossRef]
23. Lobo, F.S.N.; Visser, M. Fundamental limitations on ‘warp drive’ spacetimes. Class. Quantum Gravity 2004, 21, 5871–5892.
[CrossRef]
24. Fermi, D.; Pizzocchero, L. A time machine for free fall into the past. Class. Quantum Gravity 2018, 35, 165003. [CrossRef]
25. Misner, C.; Thorne, K.; Wheeler, J. Gravitation; W.H. Freeman and Co.: New York, NY, USA, 1973.
26. Gourgoulhon, E. 3+1 Formalism in General Relativity; Springer: Berlin/Heidelberg, Germany, 2012; Volume 846. [CrossRef]
27. Alcubierre, M.; Lobo, F.S.N. Warp Drive Basics. Fundam. Theor. Phys. 2017, 189, 257. [CrossRef].
28. Jantzen, R.; Carini, P.; Bini, D. GEM: The User Manual: Understanding Spacetime Splittings and Their Relationships; Unpublished
work, 2013.
29. Hestenes, D.; Sobczyk, G. Clifford Algebra to Geometric Calculus; Springer: Berlin/Heidelberg, Germany, 1984.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.