Preprint typeset in JHEP style - HYPER VERSION LMU-ASC 54/07

kEssence, superluminal propagation, causality and

arXiv:0708.0561v1 [hep-th] 3 Aug 2007

emergent geometry

Eugeny Babichev
INFN - Laboratori Nazionali del Gran Sasso, S.S. 17bis, 67010 Assergi (LAquila) - Italy,
Institute for Nuclear Research of RAS, 60th October Anniversary Prospect 7a, 117312 Moscow,
Russia
E-mail: [email protected]

Viatcheslav Mukhanov
Arnold-Sommerfeld-Center for Theoretical Physics, Department fur Physik,
Ludwig-Maximilians-Universitat Munchen, Theresienstr. 37, D-80333, Munich, Germany,
E-mail: [email protected]

Alexander Vikman
Arnold-Sommerfeld-Center for Theoretical Physics, Department fur Physik,
Ludwig-Maximilians-Universitat Munchen, Theresienstr. 37, D-80333, Munich, Germany
E-mail: [email protected]

Abstract: The k-essence theories admit in general the superluminal propagation of the perturba-
tions on classical backgrounds. We show that in spite of the superluminal propagation the causal
paradoxes do not arise in these theories and in this respect they are not less safe than General
Relativity.
Contents
1. Introduction 1

2. Equations of motion and emergent geometry 3

3. Causality on nontrivial backgrounds 6

4. Which initial data are allowed for the well posed Cauchy problem? 10

5. How to pose the initial conditions in a fast moving spacecraft? 11

6. Chronology protection 16

7. Is the gravitational metric universal? 18

8. Discussion 20

A. Characteristics and superluminal propagation 21

B. Action for perturbations 23

C. Action for the cosmological perturbations 25

D. Effective Hydrodynamics 27

E. Green functions for a moving spacecraft 28

1. Introduction
Over the past years, spontaneous breaking of the Lorentz invariance and questions related to this
issue, such as superluminal propagation of perturbations in nontrivial backgrounds, attracted a
renewed interest among physicists. One of the basic questions here is whether the theories allowing
the superluminal velocities possess internal inconsistencies and, in particular, inevitably lead to
the causality violation namely to the appearance of the closed causal curves (CCCs). Concerning
this issue there exist two contradicting each other points of view. Some authors (see, for instance,
[1, 2, 3, 4, 5, 6, 7, 8]) argue that the subluminal propagation condition should a priori be imposed to
make the theory physically acceptable. For example, in [1] on the P. 60 the authors introduce the
Postulate of Local Causality which excludes the superluminal velocities from the very beginning.
The requirement of subluminality is sometimes used to impose rather strong restrictions on the form
of the admissible Lagrangians for the vector and higher spin fields [5] and gravity modifications [7].
The effective field theories (EFT) allowing the superluminal propagation were considered in [8],
where it was argued that in such theories global causality and analyticity of the S-matrix may
be easily violated. The main conclusion of [8] is not favorable for the theories with superluminal
propagation. In particular the authors claim that the UV-completion of such theories must be very
nontrivial if it exists at all (for a different attitude see [9, 10]).

1
 An open minded opinion concerning the superluminal propagation is expressed in [11], where
one argues that the proper change of the chronological ordering of spacetime in non-linear field
theory with superluminal propagation allows us to avoid the causal paradoxes.
Recently, in the literature were discussed several cases in which faster-than-light propagation
arises in a rather natural way. In particular we would like to mention the noncommutative solitons
[12], Einstein aether waves [13], superluminal photons in the Drummond-Hathrell effect [14,
15] and in the Scharnhorst effect [16, 17] 1 . These last two phenomena are due to the vacuum
polarization i.e. higher-order QED corrections. It was argued that this superluminal propagation
leads to the causal paradoxes in the gedanken experiment involving either two black holes [18] or
two pairs of Casimir plates [19] moving with the high relative velocities. To avoid the appearance of
the closed causal curved in such experiments the authors of [19] invoked the Chronology Protection
Conjecture [20] and showed that the photons in the Scharnhorst effect causally propagate in effective
metric different from the Minkowski one.
Note that the superluminal propagation cannot be the sole reason for the appearance of the
closed causal curves. There are numerous examples of spacetimes in General Relativity, where the
Postulate of Local Causality is satisfied and, nevertheless, the closed causal curves are present (see
[21, 22, 23, 24, 25]). Therefore an interesting question arises whether the superluminal propagation
leads to additional problems related with causality compared to the situation in General Relativity.
In these paper we will consider the k-essence fields [26, 27, 28, 29] and show that contrary
to the claim of [2, 30] the causality is not violated in generic k-essence models with superluminal
propagation (similar attitude was advocated in [31, 32, 33]). In this sense, in spite of the presence
of superluminal signals on nontrivial backgrounds, the k-essence theories are not less safe and
legitimate than General Relativity.
The paper is organized as follows. In Section 2 we discuss the equation of motion for k-essence
and derive generally covariant action for perturbations for an arbitrary k-essence background.
General aspects of causality and propagation of perturbations on a nontrivial background,
determining the new aether , are discussed in Section 3. In particular, we prove that no causal
paradoxes arise in the cases studied in our previous works [27, 28, 26] and [29].
Section 4 is devoted to the Cauchy problem for k-essence equation of motion. We investigate
under which restrictions on the initial conditions the Cauchy problem is well posed.
In Section 5 we study the Cauchy problem for small perturbations in the new aether rest
frame and in the fast moving spacecraft.
Section 6 is devoted to the Chronology Protection Conjecture, which is used to avoid the CCCs
in gedanken experiments considered in [8].
In Section 7 we discuss the universal role of the gravitational metric. Namely, we show that for
the physically justified k-essence theories the boundary of the smooth field configuration localized
in Minkowski vacuum, can propagate only with the speed not exceeding the speed of light. In
agreement with this result we derive that exact solitary waves in purely kinetic k-essence propagate
in vacuum with the speed of light.
Our main conclusions are summarized in Section 8.
All derivations of more technical nature the reader can find in Appendices. In Appendix A we
derive characteristics of the equation of motion and discuss local causality. Appendix B is devoted to
the derivation of the generally covariant action for perturbations. In Appendix C we show how the
action derived in Appendix B is related to the action for cosmological perturbations from [28, 34].

1 In this paper under superluminal we always mean faster than light in usual QED vacuum in unbounded
empty space. To avoid confusion one could say that photons propagate faster than gravitons in the Scharnhorst
effect. When this paper was in the final stage of preparation the superluminal wave-front velocity in these effects
was putted under question [10].

2
In Appendix D we consider the connection between k-essence and hydrodynamics. The derivation
of Green functions is given in Appendix E.

2. Equations of motion and emergent geometry

Let us consider the k-essence scalar field with the action:

Z
S = d4 x gL (X, ) , (2.1)

where
1
X= g ,
2
is the canonical kinetic term and by we always denote the covariant derivative associated with
metric g . We would like to stress that this action is explicitly generally covariant and Lorentz
invariant. The variation of action (2.1) with respect to g gives us the following energy-momentum
tensor for the scalar field:
2 S
T = L,X g L, (2.2)
g g
where (...),X is the partial derivative with respect to X. The Null Energy Condition (NEC)
T n n 0 (where n is a null vector: g n n = 0) is satisfied provided L,X 0. Because
violation of this condition would imply the unbounded from below Hamiltonian and hence signifies
the inherent instability of the system [35] we consider only the theories with L,X 0.
The equation of motion for the scalar field is obtained by variation of action (2.1) with respect
to ,
S
= G + 2XL,X L, = 0, (2.3)

where the effective metric is given by

G (, ) L,X g + L,XX . (2.4)

This second order differential equation is hyperbolic (that is, G has the Lorentzian signature)
and hence describes the time evolution of the system provided [6, 36, 33]
L,XX
1 + 2X > 0. (2.5)
L,X

When this condition holds everywhere the effective metric G determines the characteristics (cone
of influence) for k-essence, see e.g. [36, 33, 37, 38]. For nontrivial configurations of k-essence field
6= 0 and the metric G is generally not conformally equivalent to g ; hence in this case the
characteristics do not coincide with those ones for canonical scalar field the Lagrangian of which
depends linearly on the kinetic term X. In turn, the characteristics determine the local causal
structure of the space time in every point of the manifold. Hence, the local causal structure for the
k-essence field is generically different from those one defined by metric g (see Appendix A for
details). For the coupled system of equations for the gravitational field and k-essence the Cauchy
problem is well posed only if the initial conditions are posed on the hypersurface which is spacelike
with respect to both metrics: g and G (see P. 251 of Ref. [39] and Refs. [40, 36, 41] for details).
We postpone the detailed discussion of this issue until Section 4 and now we turn to the behavior
of small perturbations on a given background. With this purpose it is convenient to introduce the
function  1
L,XX
c2s 1 + 2X , (2.6)
L,X

3
which for the case X > 0 plays the role of speed of sound for small perturbations [28] prop-
agating in the preferred reference frame, where the background is at rest. It is well known that
in the case under consideration there exists an equivalent hydrodynamic description of the system
(see Appendix D) and the hyperbolicity condition (2.5) is equivalent to the requirement of the
hydrodynamic stability c2s > 0.
The Lerays theorem (see P. 251 of Ref. [39] and Ref. [40] ) states that the perturbations on
given background 0 (x) propagate causally in metric G (0 , 0 ). In Appendix B we show that
neglecting the metric perturbations g , induced by , one can rewrite the equation of motion for
the scalar field perturbations in the following form
1  
GG + Meff2
= 0, (2.7)
G
here we denote
cs
q
G
2 G , G detG1
where G1
G

= , (2.8)
L,X

and !
2 cs G
Meff 2 2XL,X L, + 0 . (2.9)
L,X

Note that the metric G is conformally equivalent to G and hence describes the same causal
structure as it must be. The equation for the perturbations has exactly the same form as equation
for the massive Klein-Gordon field in the curved spacetime. Therefore the metric G describes the
emergent or analogue spacetime where the perturbations live. In particular this means that
the action for perturbations
1
Z
d4 x G G Meff2 2
 
S = , (2.10)
2
and the equation of motion (2.7) are generally covariant in the geometry G . Introducing the
covariant derivatives D associated with metric G (D G = 0), equation (2.7) becomes

G D D + Meff
2
= 0. (2.11)

Using the inverse to G matrix

   
L,X L,XX
G1
= 2
g cs 0 0 , (2.12)
cs L,X
one can define the emergent interval

dS 2 G1
dx dx , (2.13)

which determines the influence cone for small perturbations of k-essence on a given background2.
This influence cone is larger than those one determined by the metric g , provided L,XX /L,X <
0 [6, 36, 33, 38], and the superluminal propagation of small perturbations becomes possible (see
Appendix A). At first glance it looks like the theory under consideration has emergent bimetric
structure. However, this theory is inherently different from the bimetric theories of gravity [42]
because the emergent metric refers only to the perturbations of k-essence and is due to the non-
linearity of the theory, while in the bimetric gravity theories both metrics have fundamental origin
and are on the same footing.
2 Note that in order to avoid confusion we will be raising and lowering the indices of tensors by gravitational metric

g (g ) throughout the paper.

4
 The derived above form of the action and of the equation of motion for perturbations is very
useful. In particular, it simplifies the stability analysis of the background with respect to the per-
turbations of arbitrary wavelengths, while the hyperbolicity condition (2.5) guarantees this stability
only with respect to the short-wavelength perturbations.
It is important to mention that besides of the usual hyperbolicity condition (2.5) one has to
require that L,X is nowhere vanishes or becomes infinite. The points where L,X vanishes or diverges,
generally correspond to the singularities of the emergent geometry. It follows from equations (2.8)
and (2.12) that these singularities are of the true nature and cannot be avoided by the change of the
coordinate system. Therefore one can argue that before the singularities are formed the curvature
of the emergent spacetime becomes large enough for efficient quantum production of the k-essence
perturbations which will destroy the classical background and therefore L,X cannot dynamically
change its sign. Hence, if one assumes that at some moment of time the k-essence satisfies the
null energy condition, that is, L,X > 0 everywhere in the space (or + p > 0 in hydrodynamical
language; see Appendix D) then this condition can be violated only if one finds the way to pass
through the singularity in the emergent geometry with taking into account the quantum production
of the perturbations. This doubts the possibility of the smooth crossing of the equation of state
w = 1 and puts under question recently suggested models of the bouncing universe ([43]). The
statements above generalize the results obtained in [44] and re-derived later in different ways in [45]
in cosmological context.
In deriving (2.10) and (2.11) we have assumed that the k-essence is sub-dominant component in
producing the gravitational field and consequently have neglected the metric perturbations induced
by the scalar field. In particular the formalism developed is applicable for accretion of a test scalar
field onto black hole [29]. For k-essence dark energy [26] action (2.10) can be used only when
k-essence is a small fraction of the total energy density of the universe, in particular, this action is
applicable during the stage when the speed of sound of a successful k-essence has to be larger than
the speed of light [2, 32]. During k-inflation [46, 47, 27] or DBI inflation [48] the geometry g is
determined by the scalar field itself and therefore the induced scalar metric perturbations are of
the same order of magnitude as the perturbations of the scalar field. For this case the action for
cosmological perturbations was derived in [28], see also [34]. We have shown in Appendix C that
the correct action for perturbations in k-inflation has, however, the same structure of the kinetic
terms as (2.10) or, in other words, the perturbations live in the same emergent spacetime with
geometry G . One can expect therefore that this emergent geometry G has a much broader
range of applicability and determines the causal structure for perturbations also in the case of other
backgrounds, where one cannot neglect the induced metric perturbations.
If the hyperbolicity condition (2.5) is satisfied, then at any given point of spacetime the metric
G1
can always be brought to the canonical Minkowski form diag (1, 1, 1, 1) by the appro-
priate coordinate transformation. However, the quadratic forms g and G1 are not positively
defined and therefore for a general background there exist no coordinate system where they are
both simultaneously diagonal. In some cases both metrics can be nevertheless simultaneously diag-
onalized at a given point, so that, e.g. gravitational metric g is equal Minkowski metric and the
induced metric G1 2


is proportional to diag cs , 1, 1, 1 , where cs is the speed of sound (2.6).
For instance, in isotropic homogeneous universe both metrics are always diagonal in the Friedmann
coordinate frame.
We conclude this section with the following interesting observation. The effective metric (2.12)
can be expressed through the energy momentum tensor (2.2) as

G1
= g + T (2.14)

5
where
L,X L,XX L,XX
= Lcs and = cs .
cs L,X L,X
As we have pointed out the cosmological perturbations propagate in G1 even if the background
field determines the dynamics of the universe. In this case the energy momentum tensor for the
scalar filed satisfies the Einstein equations and eventually we can rewrite the effective metric in the
following form  

G1
= R g + R . (2.15)
2
This looks very similar to the metric redefinition g G1
in string theory where the quadratic
in curvature terms in the effective action are fixed only up to metric redefinition (2.15) see e.g.
[49]. The metric redefinition does not change the light cone and hence the local causality only in
the Ricci flat R = 0 spacetimes. However, neither in the matter dominated universe nor during
inflation the local causals structures determined by g and G1
are equivalent.

3. Causality on nontrivial backgrounds

t
t

future
R

signal
O
x

ponse
res
P

past

Figure 1: This figure represents the causal paradox constructed using tachyons. Someone living along the
worldline x = 0 sends a tachyon signal to the astronaut in a fast moving spacecraft, OR. In the spacecraft
frame (x , t ), the astronaut sends a tachyon signal back, RP . The signal RP propagates in the direction
of growing t as it is seen by the astronaut, however it travels back in time in the rest frame. Thus it is
possible to send a message back in the own past.

In this section we discuss the causality issue for superluminal propagation of perturbations
on some nontrivial backgrounds, in particular, in Minkowski spacetime with the scalar field, in
Friedmann universe and for black hole surrounded by the accreting scalar field.

6
 t
t

light future

kessence future

O x

kessence past

light past

Figure 2: The causality paradox is avoided when superluminal signals propagate in the background which
breaks the Lorentz symmetry (compare with Fig. 1). The observers cannot send a message to themselves
in the past.

First, we would like to recall a well-known paradox sometimes called tachyonic anti-telephone
[50] arising in the presence of the superluminal hypothetical particles tachyons possessing un-
bounded velocity ctachyon > 1. In this case we could send a message to our own past. Indeed, let
us consider some observer, who is at rest at x = 0 with respect to the reference frame (x, t) and
sends along OR a tachyon signal to an astronaut in the spacecraft R (see Fig. 1). In turn, after
receiving this signal, the astronaut communicates back sending the tachyon signal, RP . As this
signal propagates the astronaut proper time t grows. However, if the speed of the spacecraft is
larger than 1/ctachyon, then the signal RP propagates backward in time in the original rest frame
of the observer. Thus, the observers can in principle send information from their future to their
past. It is clear that such situation is unacceptable from the physical point of view.
Now let us turn to the case of the Minkowski space-time filled with the scalar field, which allows
the superluminal propagation of perturbations in its background. For simplicity we consider a
homogeneous time dependant field 0 (t). Its velocity is directed along the timelike vector,
u = (1, 0, 0, 0). Why does the paradox above not arise here? This is because the superluminal
propagation of the signals is possible only in the presence of nontrivial background of scalar field
which serves as the aether for sonic perturbations. The aether selects the preferred reference
frame and clearly the equation of motion for acoustic perturbations is not invariant under the
Lorentz transformations unless cs = 1. In the moving frame of the astronaut the equation for
perturbations has more complicated form than in the rest frame and the analysis of its solutions is
more involved. However, keeping in mind that k-essence signals propagate along the characteristics
which are coordinate independent hypersurfaces in the spacetime we can study the propagation of
sonic perturbations, caused by the astronaut, in the rest frame of the aether and easily find that

7
the signal propagates always forward in time in this frame (see Fig. 2). Hence no closed causal
curves can arise here.
We would like to make a remark concerning the notion of future- and past directed signals.
It was argued in [30] that in order to have no CCCs for the k-essence during the superluminal
stage, ...the observers travelling at high speeds with respect to the cosmological frame must send
signals backwards in their time for some specific direction. One should remember, however, that
the notion of past and future is determined by the past and future cones in the spacetime and has
nothing to do with a particular choice of coordinates. Thus, the signals, which are future-directed
in the rest-frame remain the future-directed also in a fast-moving spacecraft, in spite of the fact that
this would correspond to the decreasing time coordinate t . As we show in Section 5, the confusion
arises because of a poor choice of coordinates, when decreasing t correspond to future-directed
signals and vice versa. The example shown in Fig. 4 illustrates this point: one can see that even
without involving superluminal signals, an increasing coordinate time does not always imply the
future direction.
Another potentially confusing issue is related to the question which particular velocity must
be associated with the speed of signal propagation, namely, phase, group or front velocity. For
example, in [30] an acausal paradox is designed using different superluminal group velocities for
different wavenumbers. One should remember, however, that neither group nor phase velocities have
any direct relation with the causal structure of the spacetime. Indeed the characteristic surfaces
of the partial differential equations describe the propagation of the wavefront. This front velocity
coincides with the phase velocity only in the limit of the short wavelength perturbations. Generally
the wavefront corresponds to the discontinuity of the second derivatives and therefore it moves
off-shell (a more detailed discussion can be found in e.g. [9]). The group velocity can be less or
even larger than the wavefront velocity. One can recall the simple examples of the canonical free
scalar field theories: for normal scalar fields the mass squared, m2 > 0, is positive and the phase
velocity is larger than c while the group velocity is smaller than c; on the other hand for tachyons
(m2 < 0) the situation is opposite. Thus, if the group velocity were the speed of the signal transfer,
one could easily build the time-machine similar to those described in [30] using canonical scalar
field with negative mass squared, m2 < 0. This, however, is impossible because the causal structure
in both cases (m2 > 0 and m2 < 0) is governed by the same light cones. Finally we would like
to mention that the faster-than-light group velocity has been already measured in the experiment
[51].
To prove the absence of the closed causal curves (CCC) in those known situations where the
superluminal propagation is possible, we use the theorem from Ref. [39] (see p. 198): A spacetime
(M, g ) is stably causal if and only if there exists a differentiable function f on M such that f
is a future directed timelike vector field. Here M is a manifold and g is metric with Lorentzian
signature. Note, that the notion of stable causality implies that the spacetime (M, g ) possesses no
CCCs and thus no causal paradoxes can arise in this case. The theorem above has a kinematic origin
and does not rely on the dynamical equations. In the case of the effective acoustic geometry the
acoustic metric G1 plays the role
of g and the function f serves as the global time function of
the emergent spacetime M, G1 . For example, in the Minkowski spacetime filled with the scalar
field ether one can take the Minkowski time t of the rest frame, where this field is homogeneous,
as the global time function. Then we have

g 00
 
cs 00 L,XX
G t t = g 1 + 2X = . (3.1)
L,X L,X L,X cs

Even for those cases when the speed of perturbations can exceed the speed of light, cs > 1, this
expression is positive, provided that L,X > 0, and the hyperbolicity condition (2.5) is satisfied.

8
 1
Thus t is timelike with respect to the effective metric G ; hence the conditions of the theorem
above are met and no CCCs can exist.
Now we consider the Minkowski spacetime with an arbitrary inhomogeneous background 0 (x)
and verify under which conditions one can find a global time t for both geometries g and G1
and thus guarantee the absence of CCCs. Let us take the Minkowski t, t t = 1, and check
whether this time can also be used as a global time for G1 . We have
       
cs L,XX 2 cs L,XX
G t t = 1+ ( t 0 ) = 1+ 20 , (3.2)
L,X L,X L,X L,X

and assuming that cs > 0, L,X > 0 we arrive to the conclusion that t is a global time for emergent
spacetime provided  
L,XX  2
1+ 0 (x ) > 0, (3.3)
L,X
holds everywhere on the manifold M. This inequality is obviously always satisfied in the subluminal
case. It can be rewritten in the following form
  2
L,XX  ~
1 + c2s 0 (x ) > 0, (3.4)
L,X

from where it is obvious that, if the spatial derivatives are sufficiently small then this condition can
also be satisfied even if cs > 1. Note that the breaking of the above condition for some background
field configuration 0 (x) does not automatically mean the appearance of the CCCs. This just tells
us that the time coordinate t cannot be used as the global time coordinate. However it does not
exclude the possibility that there exists another function serving as the global time. Only, if one can
prove that such global time for both metrics does not exist at all, then there arise causal paradoxes.
In the case of the Friedmann universe with superluminal scalar field, one can choose the
cosmological time t as the global time function and then we again arrive to (3.1), thus concluding
that there exist no CCCs. In particular, the k-essence models, where the superluminal propagation
is the generic property of the fluctuations during some stage of expansion of the universe [2, 32],
do not lead to causal paradoxes contrary to the claim by [2, 30].
The absence of the closed causal curves in the Friedmann universe with k-essence can also be
seen directly by calculating of the effective line element (2.13). Taking into account that the
Friedmann metric is given by

ds2 = g dx dx = dt2 a2 (t)dx2 , (3.5)

we find that the line element (2.13), corresponding to the effective acoustic metric, is

L,X 2 2
dS 2 = G1
cs dt a2 (t)dx2 .


dx dx = (3.6)
cs
p
The theory under consideration is generally covariant. After making redefinitions, L,X cs dt dt,
and, a2 (t)L,X /cs a2 (t), the line element (3.6) reduces to the interval for the Friedmann universe
(3.5), where obviously no causality violation can occur. Thus we conclude that both the k-essence
[26] and the superluminal inflation with large gravity waves [27] are completely safe and legitimate
on the side of causality.
When X = 12 g 0 0 is positive everywhere in the spacetime the background field itself
can be used as the global time function. Indeed for general gravitational background g and cs > 0,
L,X > 0 we have
2X
g 0 0 > 0 and G 0 0 = > 0,
L,X cs

9
and due to the fact that X > 0 the sign in front 0 can be chosen so that the vector 0 is
always future directed on M. Therefore 0 (x) or (0 (x) if necessary) can serve as a global time
in both spacetimes (M, g ) and M, G1


, and no causal paradoxes arise.
In particular this is applicable for the accretion of the superluminal scalar field onto the
Schwarzschild black hole [29]. In this case sound horizon is located inside the Schwarzschild radius
and therefore the Schwarzschild time coordinate cannot be used as a global time function. However,
X > 0 outside the acoustic horizon (see [29]) and in accordance with the theorem and discussion
above we can take as the global time coordinate and hence the acoustic spacetime is stably causal.
In all examples above we have considered the superluminal acoustic metric. Thus, if there
exist no CCCs in M, G1


then there are no CCCs with respect to metric g because acoustic cone
is larger than the light cone. It may happen that in some cases it is not enough to prove that there
1


no CCCs separately in M, G and (M, g ) and one has to use the maximal cone or introduce
an artificial cone [31] encompassing all cones arising in the problem. It is interesting to note that, if
the k-essence realizes both superluminal and subluminal speed of sound in the different regions
of the manifold, then there exist hypersurface where the k-essence metric is conformally equivalent
to the g and one can smoothly glue the maximal cones together everywhere on M. After that
one can consider a new artificial metric G as determining the complete causal structure of the
manifold.
We would like to point out that although the theorem on stable causality allowed us to prove
that there is no causal paradoxes in those cases we considered above, it is no guaranteed that CCCs
cannot arise for some other backgrounds. Indeed, in [8] the authors have found some configurations
of fields possessing CCCs: one for the scalar field with non-canonical kinetic term and another
for the wrong-signed Euler-Heisenberg system. In both cases the small perturbations propagate
superluminally on rather non-trivial backgrounds. We will pursue this issue further in Section 6.

4. Which initial data are allowed for the well posed Cauchy problem?
Using the theorem on stable causality we have proven that the superluminal k-essence does not
lead to any causal paradoxes for cosmological solutions and for accretion onto black hole. However,
the consideration above is of a kinematic nature and it does not deal with the question how to pose
the Cauchy problem for the background field 0 and its perturbations .
It was pointed out in [8] that in the reference frame of the spacecraft moving with respect
to nontrivial background, where cs > 1, with the speed v = 1/cs the Cauchy problem for small
perturbations is ill posed. This happens because the hypersurface of the constant proper time t
of the astronaut is a null-like with respect to the acoustic metric G1
. Hence t = const is tangential
to the characteristic surface (or sonic cone see Appendix A) and cannot be used to formulate the
Cauchy problem for perturbations which live in this acoustic metric. Intuitively this happens
because the perturbations propagate instantaneously with respect to the hypersurface t = const.
Moreover, for v > 1/cs , the sonic cone deeps below the surface t = const (see Figs. 3 and 2)
and in the spacetimes of dimension D > 2 the Cauchy problem is ill posed as well because there
always exist two directions along which the perturbations propagate instantaneously in time t
(red vectors in Fig. 3). This tell us that not every imaginable configuration of the background can
be realized as the result of evolution of the system with the well formulated Cauchy problem and
hence not every set of initial conditions for the scalar field is allowed.
In this Section we will find under which restrictions on the initial configuration of the scalar
field the Cauchy problem for equation (2.3) is well-posed. For this purpose it is more convenient not
to split the scalar field into background and perturbations and consider instead the total value of
the field = 0 + . The k-essence field interacts with gravity and therefore for consistency one has
to consider the coupled system of equations for the gravitational metric g and the k-essence field

10
. In this case the Cauchy problem is well posed only if the initial data are set up on a hypersurface
which is simultaneously spacelike in both metrics: g and G1 (for details see P. 251 of Ref. [39]
and Refs. [40], [36], [41]). We will work in the synchronous coordinate system, where the metric
takes the form
ds2 = dt2 ik dxi dxk , (4.1)
and select the spacelike in g hypersurface to be a constant time hypersurface t = t0 . The 1-form
t vanishes on any vector R tangential to : R t = 0 (see Fig. 3). This 1-form is timelike with
respect to the gravitational metric g , that is g t t > 0. In case when Lagrangian for k-essence
depends at maximum on the first derivatives of scalar field the initial conditions which completely
specify the unambiguous solution of the equations of motion are the initial field configuration (x)
and and its first time derivative (x) (g t ) . Given these initial conditions one can
calculate the metric G1 and consequently the influence cone at every point on . First we have
to require that for a given set of initial data the hyperbolicity condition (2.5) is not violated. This
imposes the following restriction on the allowed initial values (x) and (x):
 2  2  L
2 ~ ,XX
cs = 1 + (x) (x) > 0, (4.2)
L,X
 2
~
where we have denoted (x) = ik i k . In addition we have to require that the hypersurface
is spacelike also with respect to emergent metric G , that is, for every vector R , tangential to
1
, we have G R R < 0, or
 2 L
~ ,XX
1 + c2s (x) > 0. (4.3)
L,X
If at some point on the vector R becomes null-like with respect to G1 1
, that is, G R R = 0,
the signals propagate instantaneously (red propagation vectors from cone B on Fig. 3) and one
cannot guarantee the continuous dependence on the initial data or even the existence and uniqueness
of the solution, see e.g. [52]. Using (2.6) the last inequality can be rewritten as
  2 L 
,XX
c2s 1 + (x) > 0. (4.4)
L,X
Therefore,
 given Lagrangian L (, X) and hypersurface one has to restrict the initial data

(x), (x) by inequalities (4.2) and (4.3) (or equivalently (4.4)), to have a well posed Cauchy
problem. The condition (4.4) is always satisfied in the subluminal case for which L,XX /L,X 0.
In addition, we conclude that, if these conditions are satisfied everywhere on the manifold M and
the selected synchronous frame is nonsingular in M, then time t plays the role of global time and
in accordance with the theorem about stable causality no causal paradoxes arise in this case.
As a concrete application of the conditions derived let us find which restrictions should satisfy
the admissible initial conditions for the low energy effective field theory with Lagrangian L(X)
X X 2 /4 + ..., where is a cut off scale Ref. [8]. In this case (4.4) imply that not only X 4 ,
 2  2
but also (x) 4 and (x) ~ 4 . Note that these restrictions can be rewritten in the
2
Lorentz invariant way: for example the first condition takes the form (g t ) 4 .
Finally let us note that even well-posed Cauchy problem cannot guarantee the global existence
of the unique solution for nonlinear system of the equations of motion: for example, the solution
can develop caustics [58] or can become multi-valued [61].

5. How to pose the initial conditions in a fast moving spacecraft?

In this section we resolve paradoxes which at first glance seems arising in the case of superluminal
propagation of perturbations [8, 30, 2] when one tries to formulate the Cauchy problem in a fast

11
 A
t
B

Figure 3: The Cauchy problem for the equation of motion of k-essence is set up on the hypersurface :
t = t0 . The vector R is tangential to : R t = 0. For the hyperbolic equation of motion, the Cauchy
problem is well posed provided that t is timelike with respect to G everywhere on , or, equivalently,
the hypersurface is spacelike with respect to G (the cone A in the figure). The cone B represents an
ill-posed Cauchy problem for the hyperbolic equation. In particular the red propagation vectors are tangent
to .

moving spacecraft. To simplify the consideration we restrict ourselves by purely kinetic k-essence,
for which L (, X) = L(X) and assume that for the background solution X0 = const > 0 and
cs > 1. This is a reasonable approximation for more general backgrounds with X0 > 0 on the
scales much smaller than the curvature scale of the emergent geometry G1 . There is always the
preferred reference frame t, xi in which the background is isotropic and homogeneous. We refer

to this frame as the rest frame. In the presence of an external source J equation (2.7) in this frame
takes the following form
t2 c2s x = J, (5.1)
2


where cs /L,X , for details see Appendix B, equations (B.20) and (B.1). Now let us consider
a spacecraft moving in x-direction with velocity v through the k-essence background and denote
the Lorentz boosted comoving spacecraft coordinates by t , xi . As we have already mentioned

above, if the velocity of the spacecraft is larger than c2 /cs then the Cauchy problem for cannot be
well posed on the hypersurface t = const 3 . After Lorentz transformation to comoving spacecraft
frame, equation (5.1) becomes
1 
v2 c2 v 2 c2
    
1 s4 t2 2v 1 2s t x + v 2 c2s x2 c2s J J = J, (5.2)


1 2
c c c
where prime denotes comoving coordinates and index J = 2, 3, ... stands for the spatial directions

1
other than x [note that in Ref. [8] the factor 1 v 2 /c2 in front of squire brackets is missing]. For
v = c2 /cs the second time derivative drops out of (5.2) and the necessary conditions for applicability
of the Cauchy-Kowalewski theorem are not satisfied; hence the existence and uniqueness of the
solution (5.2) are not guaranteed. For v > c2 /cs the necessary conditions of the Cauchy-Kowalewski
theorem are met and the unique solution of (5.2) exists; however, this solution contains exponentially
growing modes in the spatial directions, perpendicular to x . Indeed, substituting

exp i t + ikx x + ikJ xJ ,

3 Throughout this section we explicitly write the speed of light c and without loss of generality we assume v > 0.

12
in (5.2) we find that in the boosted frame:
s
2 2 1
 (  2  )
v2 v2
      2 2
v c s c s 2 v cs 2
= 1 4 kx v 2 1 cs 1 2 kx 1 2 1 k . (5.3)
c c c c c4

2
where we have denoted k = {kJ } and k = kJ kJ . For D = 2, when k = 0, the frequencies
are always real and no instability modes exist (note that v < c). However, if D > 2 and v > c2 /cs
then for
1 v 2 /c2
 
2
k > kx2 , (5.4)
v 2 c2s /c4 1
the general solution of (5.2) contains exponentially growing modes. Note that these are the high
frequency modes and hence the instability would imply catastrophic consequences for the theory.
At first glance, this looks like a paradox, because equation (5.1), which has no unstable solutions in
the rest frame, acquired exponentially unstable solutions in the boosted frame. On the other hand,
any solution of (5.1) after performing the Lorentz transformation with v > c2 /cs does not contain
2
exponentially growing modes with k satisfying (5.4). Indeed, given (kx , k ) in the rest frame one
can perform the Lorentz transformation and obtain:
( )
+ vkx kx + v/c2
{ , kx , k } = p , p , k , (5.5)
1 v 2 /c2 1 v 2 /c2
p
were = cs kx2 + k 2 . Expressing via k and k we again arrive to (5.3). However, it follows
x
from (5.5) that if v > c2 /cs then the components of the Lorentz boosted wavevector satisfy the
condition
1 v 2 /c2
 
2 2
k kx , (5.6)
v 2 c2s /c4 1
and hence unstable modes are not present. This raises the question whether the unstable modes
which cannot be generated in the rest frame of k-essence, can nevertheless be exited by any physical
device in the spacecraft. We will show below that such device does not exist. With this purpose
we have to find first the Greens function in both frames.
Let us begin with two-dimensional spacetime. In this case the retarded Greens function for
(5.1) in the rest frame (rf ) is (see e.g. [52]):
1
Grf
R (t, x) = (cs t |x|) . (5.7)
2cs
In the boosted Lorentz frame it becomes
!
cs t + vx /c2 |x + vt |


1
Grf
R (t , x ) = . (5.8)
2cs
p
1 v 2 /c2

For cs v < c2 , the Fourier transform of (5.8) is the retarded in t Greens function:

(t )  i+ t

Grf
R (t , k ) =
e ei t , (5.9)
2ics k
whereas for cs v > c2 it is given by:

(t )ei+ t + (t )ei t
Grf
R (t , k ) = . (5.10)
2ics k
This Greens function corresponds to the Feynmans boundary conditions in the boosted frame.
Thus, in the fast moving spacecraft, the retarded Greens function (5.10), obtained as a result of

13
Lorentz transformation from (5.7) looks like a mixture of the retarded [proportional to (t )] and
the advanced [proportional to (t )] Greens functions with respect to the spacecraft time t . In
fact, the situation is even more complicated. If from the very beginning we work in the comoving
spacecraft frame (sc), then solving (5.2) we obtain the following expression for the retarded Greens
function,
(t )  i+ t

Gsc
R (t , k ) =
e ei t . (5.11)
2ik cs
which coincides with equation (5.7), only if cs v < c2 . However, for fast moving spacecraft, cs v > c2 ,
formula (5.11) does not coincide with formula (5.10).
The situation is more interesting in the four dimensional spacetime. Similar to the 2d case,
after we apply the Lorentz boost to the retarded (in the rest frame) Greens function (see e.g. [52])

(t)
Grf i
c2s t2 |x|2 ,


R (t, x ) = (5.12)
2cs
and calculate its Fourier transform (see Appendix E for the details) we find that for the slowly
moving spacecraft, vcs < c2 ,
2 2 4 1/2 
(t )
 
2 1 cs v /c

Grf
R (t , k ) = 2
kx + k e i+ t
e i t
. (5.13)
2ics 1 v 2 /c2

That is, the resulting Greens function is also retarded with respect to the spacecraft time t . On
the other hand, for the fast moving spacecraft, vcs > c2 , we obtain:
2 2 4 1/2 
 
1 2 1 cs v /c

Grf
R (t , k ) = kx2 + k 2 2
(t ) ei+ t + (t ) ei t . (5.14)
2ics 1 v /c

Similar to the 2d case formula (5.14) is the Feynman Greens function in the spacecraft frame. Note
that formula (5.14) can be rewritten as:
2 2 4 1/2
 
1 2 1 cs v /c
Grf
R (t , k ) = 2
kx + k (5.15)
2cs 1 v 2 /c2
s !
1 c2s /c2
1 v 2 /c2 2 2
2 1 v /c
exp ikx vt cs |t | k kx2 .
1 c2s v 2 /c4 c2s v 2 /c4 1 cs v /c4 1
2 2

It is obvious from here that the modes with large k are exponentially suppressed and therefore
2
very high frequency source J cannot excite perturbations with k satisfying inequality (5.4).
In the spacecraft frame the retarded Greens function calculated directly for Fourier modes of
(5.2) is:
2 2 4 1/2 
(t )
 
2 1 cs v /c

Gsc
R (t
, k ) = k 2
x + k 2 2
ei+ t ei t .
2ics 1 v /c
It coincides with Greens function (5.13), obtained by applying the Lorentz transformation, only in
the case of slow motion with v < c2 /cs . However, the results drastically differ for the fast moving
spacecraft - compare equations (5.13) and (5.14). The function Gsc
R (t , k ) contains exponentially
growing modes for sufficiently large k and its Fourier transform to coordinate space Gsc
R (t , x ) does
not exist. Physically this means that we have failed to find the Greens function, which describes
the propagation of the signal which the source J in the fast moving spacecraft tries to send in the
direction of growing t . Instead, the response to any source in the spacecraft is always driven by
(5.15) (or the Lorentz transformed Greens function in the rest frame (5.12)). Because we cannot
send a signal in the direction of growing t one cannot associate growing t with the arrow of time
contrary to the claims in [30].

14
 Now we will discuss in more details how the problem of initial conditions for perturbations
must be correctly formulated in the fast moving spacecraft. The first question here whether the
fast moving astronaut can create an arbitrary initial field configurations and at a given moment
of his proper time t1 = const. This hypersurface is not space-like with respect to the metric G1
and therefore as it follows from the consideration in the previous section the Cauchy problem is not
well posed on it. Hence not all possible configurations are admissible on this hypersurface but only
those which could be obtained as a result of evolution of some initial configuration chosen on the
hypersurface which is simultaneously spacelike with respect to both metrics g and G1 . If the
astronaut disturbs the background with some device (source function J) which he/she switches off
at the moment of time t1 , then the resulting configuration of the field on the hypersurface t1 = const
obtained using the correct Greens function (5.15) will always satisfy the conditions needed for
unambiguous prediction of the field configuration everywhere in the spacetime irrespective of the
source J(x). The presence of the advanced mode in this Greens function plays an important role
in obtaining a consistent field configuration on t1 = const. Thus we see that not everything
is in the hand of the astronaut: he has no complete freedom in the choice of the initial field
configuration at time t1 . Nonrecognition of this fact leads to the fictitious causal paradoxes discussed
in the literature [2, 30].
For a slowly moving spacecraft, v < c2 /cs , the retarded Greens function in the rest frame is
transformed in the retarded Greens function in the spacecraft frame. Therefore we can obtain any
a priori given field configuration on the hypersurface t1 = const by arranging the source function
J in the corresponding way. Thus, the choice of the initial conditions for the perturbations at
t1 = const is entirely in the hand of the astronaut. This is in complete agreement with our previous
consideration because in the slowly moving spacecraft the hypersurface t1 = const is spacelike with
respect to both metrics.
The appearance of the advance part in the correct Greens function for the fast moving space-
craft still looks a little bit strange because according to the clocks of the astronaut the head of
the spacecraft can feel signals sent at the same moment of time by a device installed on the
stern of the spacecraft. However, in this case the proper time of the astronaut is simply not a
good coordinate for the time ordering of the events at different points of the space related by the
k-essence superluminal signals. The causality is also preserved in this case but it is determined by
the superluminal k-essence cone which is larger than the light cone and as we have already seen no
causal paradoxes arise in this case. If the astronaut synchronizes his clocks using the superluminal
sonic signals then the new time coordinate t becomes a good coordinate for the time ordering of
the causal events in different points of the space. The hypersurface t = const being spacelike in
both metrics can then be used as the initial hypersurface for the well posed Cauchy problem in
the fast moving spacecraft, that is, any initial configuration of the field can be freely created by


the astronaut on this hypersurface. In the well synchronized reference frame t, x, y, z the equa-
tion of motion for perturbations (5.1) takes the same form as in the rest frame of the k-essence
background:
t2 c2s x = J. (5.16)

It follows from here that q

= cs kx2 + ky2 + kz2 ,

and hence no exponentially growing modes exist for any kx , ky and kz .

The causal Greens function in the spacecraft frame contains only the retarded with respect to
the time t part. For example, in four-dimensional spacetime it is given by


sc i t
c2s t2 |xi |2 .


GR (t, x ) = (5.17)
2cs

15
This result can be obtained either by applying the Lorentz transformation with the invariant speed
cs to (5.7), or directly by solving equation (5.16). Thus, no paradoxes with Greens functions arise
for the superluminal perturbations. The same conclusions are valid in 4d spacetime.

acoustic cone

t
x light cone

x
x
O
xJ

Figure 4: It is shown how one can create a would -be paradox similar to that discussed in this section,
without involving any superluminal signals. The fluid is at rest and the perturbations propagate sublumi-
nally in the fluid, cs < c. The reference frame (t , x ) is connected to the rest frame by the Lorentz boost
with the invariant speed cs . If the boost speed v is such that cs /c < v/cs < 1, then the hypersurface of
constant t is inside the light cone and the Cauchy problem for the electromagnetic field is ill posed in this
reference frame. Instead, one should use the correct frame (t, x), obtained by the Lorentz boost with the
invariant fastest speed c = 1. In this frame the Cauchy problem is well-posed.

To make the consideration above even more transparent we conclude this section by considering
analogous situation with no superluminal signals involved. Namely, we take a fluid at rest with
a subluminal speed of sound, cs < c. Then we can make the Lorentz transformation using the
invariant speed cs :
t vx/c2s x vt
t = p , x = p , xJ = xJ .
2
1 v /cs2 1 v 2 /c2s
If the speed v is such that cs /c < v/cs < 1, then the hypersurface of constant t is inside the
light cone (see Fig. 4) and it is obvious that one cannot formulate the Cauchy problem for the
electromagnetic field on the hypersurface t = const. Instead, the Cauchy problem for the electro-
magnetic field can be well posed on the hypersurface t = const defined by the correct Lorentz
transformation, with the invariant speed c:
t vx/c2 x vt
t = p , x = p , xJ = xJ ,
1 v 2 /c2 1 v 2 /c2
(see Fig. 4). This consideration is fully equivalent to those one above with the only replacement
cs c.
Thus we have shown that no physical paradoxes arise in the case when we have superluminal
propagation of small perturbations on the background.

6. Chronology protection
It was claimed in [8] that the theories with superluminal propagation are plagued by closed causal

16
curves (CCC). We will argue here that the superluminal propagation cannot be the sole reason for
the appearance of CCC and moreover this problem can be avoided in this case in the same way as
in General Relativity.
It is well know that General Relativity admits the spacetimes with the closed causal curves
without involving any superluminal fields into consideration. Among examples of such spacetimes
are: Gdels cosmological model [21], Stockums rotating dust cylinder [59], wormholes [25], Gotts
solution for two infinitely long strings [22] and others [24]. A prominent time-machine model was
suggested recently by Ori [23]. In this model, made solely of vacuum and dust, the spacetime
evolves from a regular normal asymptotically flat state without CCCs and only later on develops
CCCs without violating the weak, dominant and strong energy conditions. Thus, we see that
initially good spacetime might in principle evolve to a state where the chronology is violated and
the General Relativity does not by itself explains these strange phenomena. Therefore one needs
to invoke some additional principle(s) to avoid the pathological situations with CCCs. With this
purpose Hawking suggested the Chronology Protection Conjecture, which states that the laws of
physics must prohibit the appearance of the closed timelike curves [20]. In [20] it was argued that
in the situation when the timelike curve is ready to close, the vacuum polarization effects become
very large and the backreaction of quantum fields prevents the appearance of closed timelike curves.
Similarly to General Relativity, one might assume that in the case of superluminal propagation
the chronology protection conjecture is valid as well. For example, the chronology protection was
already invoked to exclude the causality violation in the case of two pairs of Casimir plates [19], in
which photons propagate faster than light due to the Scharnhorst effect [16].
Once we employ the chronology protection principle, no constructions admitting CCCs, similar
to those presented in [8], may become possible.
In fact, the first example in [8] with two finite fast moving bubbles made of superluminal scalar
field (see Fig. 2 in Ref. [8]) is quite similar to the time machine involving two pair of Casimir
plates. In the latter case the chronology protection excludes the existence of CCCs. Here the
situation is a little bit more involved. In the example with the bubbles the background is not a free
solution of the equation of motion (2.3). Indeed as it was pointed out in [8] the fast moving bubbles
have to be separated in the direction orthogonal to the direction of motion. On the other hand
they have to be connected by light. However, if this were a free solution, then the bubbles would
expand with the speed of light and collide at the same moment of time, or even before the closed
causal curve would be formed. Thus an external source J(x) of the scalar field is required in order
to produce this acausal background. However, without clear idea about the origin of this source
and possible backreaction effects the physical interpretation of this time machine is obscure. It
is well known that admitting all possible sources of gravitational field one can obtain almost any
possible even acausal solutions in general relativity. Finally, generalizing the Hawking conjecture to
the case of scalar field one can argue that the backreaction of quantum fluctuations of perturbations
around become large before CCC is formed thus destroying the classical solution imposed by
the external source J(x) and preventing the formation of CCC.
The other example considered in [8] involves non-linear electrodynamics. The electromagnetic
field is created by charge currents, serving as a source. Thus, unlike the previous example, one can
control the strength of the field, simply changing the configuration of charges. The electromagnetic
part of the Lagrangian is the wrong-signed Euler-Heisenberg Lagrangian:
1
L = F F (F F )2 + ..., (6.1)
4
with a small positive . For such a system the propagation of light in a non-trivial background is
superluminal. As a consequence, a cylindrical capacitor with the current-carrying solenoid leads to
the appearance of the CCCs, provided the electrical and magnetic fields inside the capacitor are

17
large enough (see Fig. 3 in [8]). In this case one may invoke a simplified version of the chronology
protection conjecture. In fact, let us begin with some good initial conditions in the capacitor,
namely, with electric and magnetic fields being not too large, so that no CCCs exist. Then we
increase the current in the solenoid and the voltage between the plates of a capacitor in order to
increase the strength of the fields. When the causal curves become almost closed, the expectation
value of the energy-momentum tensor for the quasi-photons on this classical electromagnetic
background becomes very large due to the quantum vacuum polarization effects. In the limit when
the causal cone becomes horizontal, the energy density of the field in the capacitor tends to infinity
and the capacitor will be broken before the CCCs will be formed.
Thus we conclude that concerning the causal paradoxes the situation in the theories with super-
luminal propagation on the non-trivial backgrounds is not much worse than in General Relativity.
In fact, in this respect the similarity between these two theories goes even much deeper than it
looks at the first glance. For example, let us imagine a time machine which is constructed with
the help of superluminal propagation in non-trivial background produced by the external source
J (x), e.g., similar to those described in [8]. Then we can identify the effective metric G for
this system with the gravitational metric g of some spacetime produced by an energy-momentum
(J)
tensor T (x). Put differently, once having the effective metric, we can find spacetime where the
gravitational metric is G . In this spacetime the time machine exists as well. Remarkably, now
the gravitation (or light) signals are used to make CCCs. The spacetime with the metric G is
(J)
the solution of Einstein equations with the energy-momentum tensor T calculated substituting
the metric in the Einstein equations. After that one could try to find such theories and such fields
(J)
configurations on which their resulting energy momentum tensor is equal to T (x) consistently
with equations of motion. One can, in principle, argue, that in the case when the CCCs exist
the energy-momentum tensor mights have some undesired properties, for example, it would violate
the Week Energy Condition (WEC). However, in several known examples with CCCs the WEC is
satisfied, see, e.g. Refs. [21, 22, 59]. Moreover, the system found by A. Ori [23] possesses CCCs and
satisfies the week, dominant and strong energy conditions. Thus the violation of the energy condi-
tions is not an inherent property of the spacetimes with CCCs. Therefore the question, whether the
spacetime constructed by the procedure described above requires bad energy-momentum tensor
or not, must be studied separately in each particular case.
(J)
Moreover the correspondence G g , J T can also be used to learn more about
Chronology Protection Conjecture and time-machines in General Relativity with the help of more
simple theory. It is well-known that Analogue Gravity [60] gives more simple and intuitively clear
way to investigate the properties of Hawking radiation, the effects of Lorentz symmetry breaking,
transplanckian problem etc., by using the small perturbations in the fluids instead of direct im-
plication of General Relativity. In a similar way, analogue time-machine or analogue Chronology
Protection Conjecture may provide one with a tool to check Chronology Projection Conjecture and
the possibility of construction of time machines in General Relativity.

7. Is the gravitational metric universal?

It was argued in [3] that the same causal limits apply to all fields independent of the matter present,
thus endowing the gravitational metric with the universal role. However, in the theories under
consideration the causal limit is governed not by the gravitational metric g , but by the effective
acoustic metric G1 and hence the gravity loses its universal role in this sense. Nevertheless, here
we argue, that even in the case of the spontaneously broken Lorentz invariance with a superluminal
propagation the gravitational metric g still keeps its universal role in the following sense. First in
accordance with the discussion in 4 we remind that the Cauchy hypersurface for the field should
anyway be a spacelike one in the gravitational metric. Thus in order to produce a background which

18
breaks the Lorentz invariance one has to respect the usual causality governed by the gravitational
metric g . Moreover, if a clump of the scalar field is created in a finite region surrounded by a
trivial background, then the boundaries of the clump will generically propagate with the speed of
light.
Indeed, let us consider a finite lump of non-trivial
 field
 configuration with smooth boundaries
(see Fig. 5) and assume that the initial data (x), (x) are specified in some finite spatial region
R. These initial data are smooth everywhere (see Fig. 5) and satisfy the conditions (4.4) and (4.2),
in particular the first derivatives of the field are continuous everywhere including the boundaries of
the clump. If the system described by action (2.1) has at least one trivial solution = triv = const
with non-pathological acoustic geometry, then, as it follows from (2.3) and (2.12), the Lagrangian
L(, X) is at least twice differentiable at (, X) = (triv , 0) and moreover L,X (triv , 0) 6= 0. Thus
for the theories of this type we have

L(, X) V () + K1 ()X + K2 ()X 2 + ... (7.1)

in the vicinity of the trivial solution triv 4 . And as expected we conclude that the speed of sound
for the small perturbations is equal to the speed of light in the vicinity of triv because any trivial
solution and in particular a possible vacuum solution = 0 does not violate the Lorentz invariance.
Moreover, close to the boundaries of the clamp the initial data (x), (x) can be considered as
small perturbation around the trivial background and therefore the front of the clump propagates
exactly with the speed of light in the vacuum. Thus, without preexisting nontrivial configuration
of the scalar field the maximum speed of propagation never exceeds the speed of light and the
causality is entirely determined by the usual gravitational metric only.

(x)
(x)

Figure 5: The figure shows that the gravitational metric g keeps its universal meaning even if the small
perturbations on the non-trivial backgrounds propagate superluminally. If in the initial moment of time the
non-trivial configuration of the field is localized in the finite region R on a spacelike in g hypersurface
, and beyond this region the field is in its vacuum state = const, then the front of the solution always
propagates with the speed of light. The blue lines correspond to the light rays. The pink cones represent
the influence cones for k-essence. On the boundary of R the influence cones are equal to light cones.

If we abandon the condition of the regularity of the emergent geometry G1

, but still require
that the Lagrangian is analytic function of X in the neighborhood of X = 0, then the speed of
propagation in vacuum is always smaller than the speed of light. Indeed in this case the speed of

4 In particular, it was required in [57] that for models allowing topological k-defects, the asymptotic behavior near

the trivial vacuum X = 0 (at the spatial infinity) is of the form (7.1)

19
sound cs is:
1
c2s = < 1,
(1 + 2 (n 1))
where n is the power of the first non-zero kinetic term in (7.1).
To demonstrate explicitly the points stated above we will find now exact solitonic solutions in
the purely kinetic k-essence theories with Lagrangian L (X) and verify that these solitons propagate
in the Minkowski spacetime with the speed of light. Assuming that the scalar field depends only
on x + vt and substituting = () in equation (2.3) we find that this equation reduces to

2
L,X , v 2 1 + L,XX , 2, v 2 1 = 0,


(7.2)

This equation is trivially satisfied for v = 1, that is, there exist solitary waves (x t) propagating
with the speed of light. They are solutions corresponding to rather special initial conditions 0 (x) =
(x) and 0 (x) = (x). Note that the general solutions are not a superposition of these solitonic
solutions because the equation of motion is nonlinear. Assuming that v 6= 1 we find that (7.2) is
satisfied by either nonlocalized solution = x vt + const, or it reduces to:

L,X + L,XX 2, v 2 1 = L,X + 2XL,XX = 0,

(7.3)

This algebraic equation is trivially satisfied for all X if L =f () X V () . This is the case
when the perturbations propagate with the infinite speed on any background [63]. For more gen-
eral Lagrangians L (X) equation (7.3) can be solved algebraically to obtain a particular X0 =
1 2 2


2 , v 1 = const. The only solutions of this last equation are either (x t) or trivial solu-
tions triv = const. For the Born-Infeld Lagrangian [56] the exact solutions of this type were found
in [61] (see also [62]). For more complicated Lagrangian, for example, of the form L = K(X) + V ()
there exist solitonic solutions with v < 1 [57].
Thus, we have shown that under reasonable restrictions on the theory the field configurations
localized in trivial vacuum never propagate faster than light. Therefore the causal limit for these
localized configurations is always governed by the usual gravitational metric.

8. Discussion
In this paper we have considered the k-essence-like scalar fields with the Lorentz invariant action
(2.1) and have studied the issues of causality and Cauchy problem for such theories. These questions
are non-trivial because small perturbations on backgrounds 0 can propagate faster-than-light.
The perturbations feel the effective metric, G given by (2.12), which is different from the
gravitational metric g , if the Lagrangian L is a non-linear function of X and the background is
nontrivial 0 6= 0. We have derived the action for the perturbations on an arbitrary background
and have shown that these perturbations feel the emergent geometry G1 . The influence cone
determined by G1 is larger than those one determined by metric g provided L,XX /L,X < 0
[6, 36, 33]. Thus perturbations can propagate with the speed exceeding the speed of light. In this
case the background serves as a new aether and preselects the preferred reference frame. This is why
the causal paradoxes arising in the presence of tachyons 5 (superluminal particles in the Minkowski
vacuum) do not appear here. In particular, we have shown that in physically interesting situations,
namely, cosmological solutions and for the case of a black hole surrounded by an accreting fluid,
the closed timelike curves are absent and hence we cannot send the signal to our own past using
the superluminal signals build out of the superluminal scalar field perturbations. Thus, the k-
essence models, which generically possess the superluminal propagation, do not lead to the causal
paradoxes, contrary to the claim in [2, 30].
5 Do not confuse them with field theoretical tachyons with m2 < 0.

20
 We have shown how to pose correctly the Cauchy problem for the k-essence fields with su-
perluminal propagation, which sometimes might seem problematic [8]. The correct initial Cauchy
hypersurface must simultaneously be spacelike with respect to both gravitational metric g and
the effective metric G1 1
. Because the effective metric G itself depends on the values of the field
and its first derivatives, the initial value problem must be set up in a self-consistent manner: in
addition to the usually assumed hyperbolicity condition (2.5), one must require that the field
and its derivative on must satisfy the inequality (4.3). In particular, in the case of spacecraft
which has very large velocity with respect to the homogeneous background of the k-essence, the
latter conditions are violated on the hypersurface of constant astronaut proper time. Therefore no
physical devices are able to produce an arbitrary configuration of perturbations on this hyperspace.
It was found in [8] that in the theories under consideration one can have the backgrounds
possessing the closed causal curves (CCCs). However, as we have argued above, this is not directly
related to the superluminal propagation. In fact, the situation here is very similar to the situation
in General Relativity, where one can also have the manifolds with the closed causal curves although
the speed of propagation is always limited by the speed of light. In this respect the situation in
the theories with the superluminal propagation is not worse than in General Relativity. To avoid
causal paradoxes in General Relativity, Hawking suggested the Chronology Protection Conjecture,
which states that the quantum effects and, in particular, vacuum polarization effects can prevent
the formation of the closed timelike curves [20]. Similarly to Hawking one may argue that in the
case of the superluminal propagation the Chronology Protection Conjecture can be valid as well. In
fact, this conjecture was already invoked to exclude the causality violation in the case of two pairs
of Casimir plates [19]. Once we employ the Chronology Protection Conjecture, no constructions
admitting CCCs, similar to those presented in [8], are possible.
Sometimes the superluminal theories are criticized in the literature on the basis of general, or
better to say, aesthetic grounds. For example, Ref. [3] claims: The spacetime metric is preferred in
terms of clock measurements and free fall (geodesic) motion (including light rays), thus underlying
General Relativitys central theme of gravity being encoded in spacetime curvature. Although this
argument is not more than the matter of taste, we would rather prefer to have General Relativity
as a theory which keeps its (restricted) universal meaning even in the presence of superluminal
propagation. We have argued that under physically reasonable assumptions and without a preex-
isting nontrivial background the causality is governed by metric g . Indeed, if initially the field
is localized within some finite region of space surrounded by vacuum, then the border of this region
propagates with the speed of light and it is impossible to send signals faster than light.

Acknowledgments

We are very thankful to Camille Bonvin, Chiara Caprini, Sergei Dubovsky, Ruth Durrer, Valery
Frolov, Robert Helling, Mattew Kleban, Lev Kofman, Stefano Liberati, Alan Rendall, Sergei
Sibiryakov, Ilya Shapiro, Alexey Starobinsky, Leonard Susskind, Matt Visser, and especially Sergei
Winitzki for very useful discussions. It is a pleasure to thank Sergei Winitzki for helpful com-
ments on the first version of the manuscript. E.B. thanks Alexander von Humboldt foundation and
INFN for support. A.V. would like to thank the theory group of Laboratori Nazionali del Gran
Sasso, INFN, and organizers and staff of Les Houches Summer School for hospitality during the
preparation of this manuscript and during earlier stages of this project respectively.

A. Characteristics and superluminal propagation

Let us consider scalar field interacting with external source J (x). The equation of motion for the

21
scalar field is
G + , = J (A.1)
where metric G is given by (2.4) and for brevity we use the hydrodynamic notation (X, ) =
2XL,X L (see Appendix D). Suppose 0 is the background solution of (A.1) in the presence
of source J0 (x) and gravitational metric g (x). Let us consider a slightly perturbed solution
= 0 + of (A.1) with the source J = J0 + J and the original unperturbed metric g (x). The
equation of motion for is then
G + , + ,X X + G 0 = J, (A.2)
where
G G
X = 0 and G = + .

This equation can be written as
G + V + M 2 = J, (A.3)
where
G
V (x) 0 + ,X 0 , (A.4)

and
G
M 2 (x) 0 + , . (A.5)

Considering the eikonal (or short wavelength) approximation [53] we have (x) = A (x) exp iS (x),
where is a large dimensionless parameter and the amplitude A (x) is a slowly varying function.
In the limit the terms containing no second derivatives, V (x) and M 2 (x) , become
unimportant and (A.3) becomes
G S S = 0. (A.6)
The equation of motion in the eikonal approximation (A.6) is conformally invariant. The surfaces
of constant eikonal S (constant phase) correspond to the wave front (characteristic surface) in
spacetime. Thus the 1-form S is orthogonal to the characteristic surface. The influence cone at
point P is formed by the propagation vectors N tangential to the characteristic surface N S = 0
and positive projection on the time direction. Using (A.6) one can chose N = G S and verify
that this vectors are tangential to the characteristic surface. The metric G has an inverse G 1

due to the requirement of hyperbolicity (Lorentzian signature of G ). Therefore S = G 1

N
and we obtain the equation for the influence cone in the form
G1
N N = 0.

Thus the metric G1 governs the division of acoustic spacetime into past, future and inaccessible
spacelike regions (or in other words this metric yields the notion of causality). It is well known
that this division is invariant under conformal transformations. From action (2.10) for perturbations
, which we derive in Appendix B, it follows that in four dimensions it is natural to consider a
conformally transformed metric G1 2

1
= L,X /cs G . Using this metric from (2.12) one obtains

L,X 2
G1
N N = g N N cs L,XX ( N ) .
cs
Therefore 
L,XX 2
g N N = c2s ( N ) ,
L,X
and if L,XX /L,X is negative, then g N N < 0, that is, N is spacelike and the cone of influence
on this background is larger than the light cone: the wave front (or signal) velocity is larger then
the speed of light. Note that this is a coordinate independent statement.

22
B. Action for perturbations
Here we sketch the derivation of action (2.10) for in the spacetime of arbitrary dimension N > 2.
First of all we would like to investigate whether there exists a metric G for which the equation
of motion for perturbations takes a canonical (Klein-Gordon) form

G D D + Meff
2
= I, (B.1)

where D is a covariant derivative with associated with the new metric G : D G = 0. Note
that the equations of motion (A.2) and (B.1) should have the same influence cone structure. Thus
the metrics G and G must be related by conformal transformation and if it is really possible
to rewrite (A.2) in canonical form, then there must exist (0 , X0 ) , such that

G = G . (B.2)

Therefore our first task is to find (0 , X0 ). Note that this method makes sense for the dimensions
D > 2 only. That happens because in D = 2 all metrics are conformally equivalent to and
the wave equation is conformally invariant, see e.g. Ref. [39], P. 447. Let us define the following
covariant derivative
D A = A L A (B.3)
which is compatible with the new metric whereas A = A A denotes the standard
covariant derivative associated with the gravitational metric: g = 0, as usual. Note, that
the tensor L introduced in (B.3) is the difference of the Christoffel symbols corresponding to the
effective and gravitational metrics. Comparing (A.2) and (B.1) we infer that

G D D + Meff
2
= G G L + Meff
2

must be equal (up to a multiplication by a scalar function ) to the l.h.s of (A.3). These can be
true only if the following condition holds

G L = V , (B.4)

where V is defined in (A.4). When this condition is satisfied we can always make the redefinition
2
Meff = M 2 and I = J,

where M 2 is defined in (A.5). The connection L depends on the unknown function (and its
derivatives) which has to be obtained form (B.4). To solve (B.4) it is convenient to multiply its
both sides by . Then using (A.4) and (B.2) this condition takes the form:
!
G
G L = 0 + ,X 0 . (B.5)

Let us now solve (B.5) with respect to . In complete analogy with the formula (86,6) from Ref. [53]
we have 
1 
G L = GG , (B.6)
G
q q
where G = detG1 = D/2
detG1
, and D is the number of dimensions of the space-
time. Using the formula (B14) from Ref. [33] one obtains
D 1 D 2
detG = (L,X ) c2
s det (g ) , and detG = (L,X ) cs det (g ) . (B.7)

23
Finally we arrive to the relation,

G = cs g (L,X )D/2 . (B.8)

It is convenient to introduce the auxiliary function

D/2
F = cs (L,X ) . (B.9)

and then using (B.6), we can rewrite equation (B.5) as:

!


 G
F G =F 0 + ,X 0 . (B.10)

Differentiating the metric G from the l.h.s. of the last equation in accordance with the chain rule
we find:
! ! !
G G G
G F = F 0 ,X g 0 . (B.11)

Further we obtain

G
0 = (L,X + 2XL,XX ) 0 = ,X 0 . (B.12)

For the first term in the brackets in (B.11) we have:

G
= L,XX g 0 + g 0 + g 0 + L,XXX 0 0 0 ,


(B.13)

and therefore
G G
= 0. (B.14)

Thus the r.h.s. of (B.11) identically vanishes. Note that there exists the inverse matrix G1 to
G . Therefore from (B.11) we conclude that F = 0 or F = const on all backgrounds and for all
D/2
theories. Considering the linear case, L (, X) = X V (), we infer that F = cs (L,X ) =1
or   1/(D/21)
D/2
= cs L,X . (B.15)

Having calculated we can formulate the main result of this Appendix as follows: the action from
which one can obtain the equation of motion in the canonical Klein-Gordon form (B.1) is

1
Z
dD x G G Meff 2 2
 
S = + 2I , (B.16)
2

where the emergent metric G is the conformally transformed eikonal metric G , defined in (2.4),
1/(D/21)  1/(D/21)    

D/2 cs L,XX
G cs L,X G = g + . (B.17)
L,X L,X

The inverse metric G1 1

can be easily calculated using the ansatz G = g + 0 0 and
is given by the formula
 1/(D/21)    
cs L,XX
G1
= g c2s 0 0 . (B.18)
L,X L,X

24
Finally the effective mass is
" #
2

N/2
1/(D/21) G
Meff = cs L,X 2XL,X L, + 0 , (B.19)

and the effective source for perturbations is given by

 1/(D/21)
D/2
I = cs L,X J. (B.20)

For the reference we also list the formula

!1/(D2)
LD,X
G = g . (B.21)
c2s

C. Action for the cosmological perturbations

Here we compare the action (2.10) with the action for scalar cosmological perturbations from
Refs. [34, 28]. In particular we show that cosmological perturbations propagate in the metric (2.12)
but have an effective mass different from (2.9). Finally we derive the generally covariant action for
the scalar cosmological perturbations.
To begin with let us consider the action (2.10) for a perturbations (, x) around a homogeneous
background () in the spatially flat Friedmann universe

ds2 = g dx dx = a2 () d 2 dx2 = a2 () dx dx


(C.1)
R
where is the conformal time = dt/a (t) and is the standard Minkowski metric. Using
Eq. (2.6) and Eq. (C.1) one can calculate the effective line element (2.13):
   
2 1 L,X 2 2 2 L,XX 2
dS = G dx dx = ds a cs 2Xd =
cs L,X
L,X 2 2 2
a cs d dx2 cs A2 c2s d 2 dx2 .



= (C.2)
cs
where we have introduced the convenient variable

A ,X a. (C.3)

Note that for the models respecting the NEC (L,X 0) the hyperbolicity condition (2.5) requires

,X > 0 and therefore A is always well defined. The factor G can be then calculated either from
the last expression above (C.2) or from the general expression (B.21):

L2,X 4
G = a = c3s A4 . (C.4)
cs

Using the formulas (2.8) and (2.6) we calculate the kinetic term

1  2 
G = cs a2 L,X ~ 2 .
( ) c2s () (C.5)

Thus in the case when the perturbations do not influence the metric g the action (2.10) takes
the form " #
2
1 L
Z  
2 ~ 2 M2 ,X
S = d3 xd a2 ,X ( ) c2s () eff a4 2 , (C.6)
2 cs

25
here we have used the definitions of the sound speed (2.6) and energy density (D.2). It is convenient
to introduce the canonical normalization for the perturbations. This is achieved by the following
field redefinition:

= ,X a = A. (C.7)
Finally integrating by parts and dropping the total derivative terms we obtain the following canon-
ical action
1
Z h i
S = ~ 2 m2eff 2
d3 xd ( )2 c2s () (C.8)
2
where the new effective mass meff is given by the following expression
" #

2 2 G A a2 G ,X a
meff = Meff = , + 0 . (C.9)
A2 A ,X ,X a

or in other terms


1  ,X a
m2eff 2

= ,X + H (3p,X ,X ) + , a . (C.10)
,X ,X a
Now let us consider the case of cosmological perturbations in the case where the field is responsible
for the dynamics of the Friedmann universe. Following [28, 34] one introduces a canonical variable

   

,X a + = A + , (C.11)
H H
and a convenient auxiliary variable z

z ,X a = A, (C.12)
H H
where is the gauge invariant perturbation of the scalar field, H a /a and = is the gauge
invariant Newtonian potential. Using this notation the action for scalar cosmological perturbations
takes the form:
1
Z h i
Scosm = ~ 2 m2 2
d3 xd ( )2 c2s () (C.13)
cosm
2
where
z
m2cosm . (C.14)
z
It is easily to check that for all cases besides canonical field without potential L(, X) X

m2cosm 6= m2eff . (C.15)

However, comparing the action (C.8) with C.13 one arrives to conclusion that the cosmological
perturbations propagate in the same metric (2.8), (2.12). Further one can introduce the notation
for the sometimes so-called scalar perturbations on the spatially flat slicing

+ . (C.16)
H
For this scalar field the action for cosmological perturbations (C.13) takes the form
1
Z
2
h i
Scosm = d4 x G G Mcosm
2
, (C.17)
2
thus the cosmological perturbations live in the emergent acoustic spacetime with the metric
(2.8), (2.12). Similarly as we have calculated in (C.9) we have


2 G A 2 2
,X a z
Mcosm 2
= Mcosm a L,X cs = (C.18)
A A ,X a z

26
after some algebra the last expression reduces to
 
A 2
+ A2 Mcosm c3s = 0


+2 (C.19)
A
where we have introduced a new auxiliary field
1/2 r


3 2X
() = . (C.20)
H 8GN
The equation (C.19) is in turn the Klein-Gordon equation
2
c3s = 0



g + Mcosm (C.21)

for the field in the metric g A2 = ,X g conformally related to the gravitational metric
g . Thus we have
2
Mcosm = c3s
1
g . (C.22)
One can rewrite this formula in terms of the gravitational metric g . Using the rules of the
conformal transformations we have
1 p  1 1
gg = 2 ,X gg =


g = (C.23)
g ,X g
1
= ,X + 1
,X g (C.24)

Thus the effective mass for cosmological perturbations is

r r r !
2 3 1 X X
Mcosm = cs ,X g + ln (,X ) ln . (C.25)
X
2
Note that in the case of canonical kinetic terms L(, X) = X V () the last expression for Mcosm
the simplifies to
2 1/2 1/2
Mcosm,canonical = (w + 1) g (w + 1) . (C.26)
where w = p/ is the equation of state parameter. In particular for the universe filled with the
massless canonical scalar field Mcosm, = 0.

D. Effective Hydrodynamics
It is well-known that for timelike (X > 0 in our signature) one can employ the hydrodynamic
approach to describe the system with the action (2.1). To do this one need to introduce a four-
velocity as follows:

u . (D.1)
2X
Using (D.1) the energy momentum tensor (2.2) tensor can be rewritten in the perfect fluid form:

T = ( + p) u u pg ,

where the pressure coincides with the Lagrangian density, p = L(X, ), and the energy density is

(X, ) = 2Xp,X p. (D.2)

The sound speed (2.6) can be expressed [28] as

 
p,X p
c2s = = . (D.3)
,X

27
In what follows we restrict ourselves to the class of Lagrangians which do not depend of explicitly,
p = p (X) and in addition we require that X > 0. This class of models is equivalent to perfect
fluid models with zero vorticity and with the pressure being a function of the energy density only,
p = p(). Then the expressions (2.6) or (D.3) coincide with the usual definition of the sound speed
for the perfect fluid: c2s = p/. Apart from the energy density and pressure p one can also
formally introduce the concentration of particles:

Z 
d
n exp = Xp,X .
+ p()
and the enthalpy
+p
h = 2 X.
n
In particular the equation of motion (2.3) takes the form of the particle number conservation law:
(nu ) = 0. Using these definitions we can rewrite the induced metric metric G and its inverse
in terms of hydrodynamic quantities only:
hcs
G = g 1 c2


s u u , (D.4)
2n
2n
G1 g 1 c2s u u .


= (D.5)
hcs
To our best knowledge these metrics (D.4) along with an action for the velocity potentials were
introduced for the first time in [55], where the accretion of the perfect fluid onto black hole was
studied. As it follows from the derivation in Appendix B, the metric (B.17) and the action (B.16)
derived in our paper are applicable in the more general case of arbitrary nonlinear scalar field
theories L (X, ) and for all possible (not only timelike X0 > 0) backgrounds produced by any
external sources. Note that the scalar field theory with Lagrangian L (X, ) , which explicitly
depends on , is not equivalent to the isentropic hydrodynamics, because and X are independent
and therefore the pressure cannot be expressed though only.

E. Green functions for a moving spacecraft

Here we calculate the retarded Greens function for a moving spacecraft in the case of three spatial
dimensions. First we calculate the retarded Greens function in the preferred (rest) frame and then
we perform the Lorentz boost (with the invariant speed c) for the solution. We compare the result
with one obtained by the direct calculation of Greens function for the Eq. (5.2). We will need the
following formulas (Gradshtein, Ryzhik, p.750):
Z  p

 cos a c 2 b2
J0 b x2 a2 sin (cx) = , for 0 < b < c (E.1)
a c2 b 2
= 0, for 0 < c < b (E.2)
Z  p

 sin a c2 b2
J0 b x2 a2 cos (cx) = , for 0 < b < c (E.3)
a c2 b 2


exp a b2 c2
= , for 0 < c < b (E.4)
b 2 c2
Z a  p

 sin a c2 + b2
J0 b a2 x2 cos (cx) = , for 0 < b (E.5)
0 c2 + b 2
In the preferred frame the Green function is (see e.g. [52])
(t)
Grf c2s t2 |x|2 .


R (t, x) = (E.6)
2cs

28

1/2
Performing the Lorentz transformation x = (x + vt ), t = (t + vx ), where = 1 v 2
we find the Green function in the moving frame:
(t + vx ) h 2  2 2 2
 i
Grf
R (t
, x
) = c s (t + vx ) (x
+ vt ) y 2
z 2
. (E.7)
2cs
We need to calculate the Fourier transform to the function (E.7). It is convenient to shift x as
follows:
1 c2s
 

x = x vt . (E.8)
1 c2s v 2
Then the argument of the delta-function in (E.7) can be rewritten as
 
2 2
2 c2s (t + vx ) (x + vt ) y 2 z 2 = c2s t2 1 x2 y 2 z 2 ,

where
1 v2
= . (E.9)
1 c2s v 2
Now we are ready to proceed with the Fourier transform of (E.7):
Z
ei
Grf
dxdydz (t + vx ) c2s t2 1 x2 y 2 z 2 eikx x+iky y+ikz z


R (t , k ) = (E.10)
2cs
where we introduced the notation:
1 c2s
 
= kx vt . (E.11)
1 c2s v 2
Step-function in the integral implies that the integration over x is made from x to +:
Z Z Z
ei
Grf
dz c2s t2 1 x2 y 2 z 2 eikx x+iky y+ikz z , (E.12)


R (t , k ) = dx dy
2cs x

1 c2s t t 1 v2
   

x = vt = = t . (E.13)
1 c2s v 2 v v 1 c2s v 2 v
p
Introducing
q r y 2 + z 2 , as the angle between the vectors {ky , kz } and {y, z} and k
ky2 + kz2 we obtain:

2
ei
Z Z Z
Grf
d c2s t2 1 x2 r2 eikx x+ik r cos .


R (t , k ) = dx drr (E.14)
2cs x 0 0

Integrating over r first gives:

Z + Z 2
ei
 q 
Grf
R (t
, k ) = dx d exp ikx x + ik c 2 t2 1 x2 cos
s (E.15)
4cs x 0

for
c2s t2 1 x2 > 0, (E.16)
otherwise it is zero. Integrating (E.15) over we find:

ei +
Z  q 
rf 2 2 1 2
GR (t , k ) = k
dxJ0 cs t x exp (ikx x) , (E.17)
2cs x

where J0 (x) is the Bessel function of the zeroth order. Now we need to integrate the expression
(E.17) taking into account the condition (E.16).We consider two cases separately: the case of slow
spacecraft, v 2 c2s < 1 ( > 0), and the case of rapid spacecraft, v 2 c2s > 1 ( < 0).

29
 For the slow spacecraft we easily obtain from (E.17) and (E.16):
cs t
ei
Z  q 
Grf
R (t , k ) = (t ) dx J0 k cs t x eikx x
2 2 1 2
2cs cs t
cs t
ei
 
k
Z q
= (t ) dx J0 2 c2s t2 x2 cos (kx x) .
cs 0

Using (E.5) we then find the Greens function for slow moving spacecraft:

iei  2 2 2 2 
ics t kx +k / ics t kx +k /
Grf
R (t , k ) = (t ) p e e
2cs kx2 + k 2 /

2 2 1/2 
 
1 2 1 cs v

= (t ) kx2 + k 2
ei+ t ei t . (E.18)
2ics 1v

In the case of rapid spacecraft, v 2 c2s > 1 ( < 0), one can verify that 2 c2s t2 > x2 for any t .
Thus (E.17) along with (E.16) can be rewritten as:
!
ei + k
Z q
rf 2
GR (t , k ) = dx J0 p 2 c2s t2 x (cos (kx x) + i sin (kx x)) . (E.19)
2cs |cs t | ||
2
Using (E.2) and (E.4) for k > || kx2 and (E.1) and (E.3) for k 2
< || kx2 we obtain in both cases:
 q 
exp |c t
| k 2 ||1 k 2
s x
ei
Grf
R (t , k ) = . (E.20)
2cs
q
k 2 ||1 k 2
x

The last expression can be written as

2 2 1/2 
 
1 2 1 cs v i+ t i t

Grf
R (t , k ) = 2
kx + k (t ) e + (t ) e . (E.21)
2ics 1 v2

Thus the modes propagating in with

1 v2
 
2
k > kx2 || = kx2 (E.22)
cs v 2 1
2

2
are exponentially suppressed. The singular directions k = kx2 || are unphysical because they have
measure zero in the integral. This directions correspond to the sufficient but integrable singularities
in the Green function.
If the Greens function is calculated directly from the Eq. (5.2) by means of standard approach
then one can find, that the solution is:
2 2 1/2 
 
1 2 1 cs v i+ t i t

Gsc
R (t , k ) = (t ) kx2 + k e e , (E.23)
2ics 1 v2

which coincides with the Greens function (E.18) we calculated by applying the Lorentz transfor-
mation to the rest Greens function in the case of slow motion. Note, however, that the results
differs for the case of fast moving spacecraft - compare (E.23) and (E.21). The function Gsc
R (t , k )
from (E.23) contains exponentially growing modes for sufficiently high k , while correct way of
calculation gave us a sensible result (E.21) - it contains only exponentially suppressed modes. This
makes sense because the late time solution approaches the free wave which do not contain these
high k .

30
References
[1] S.W. Hawking, G.F.R. Ellis, The Large scale structure of space-time, Cambridge University Press,
Cambridge, (1973).
[2] C. Bonvin, C. Caprini, R. Durrer, Phys.Rev.Lett. 97:081303, (2006), [astro-ph/0606584].
[3] G. F. R. Ellis, R. Maartens, M. MacCallum, gr-qc/0703121.
[4] G. W. Gibbons, hep-th/0302199.
[5] G. Velo, D. Zwanziger, Phys.Rev.188:2218-2222, (1969);G. Velo, D. Zwanziger, Phys.Rev. 186, 1337
- 1341 (1969)
[6] Y. Aharonov, A. Komar, L. Susskind, Phys. Rev. 182:1400-1403, (1969).
[7] Antonio De Felice, Mark Hindmarsh, Mark Trodden, JCAP 0608:005, (2006), [e-Print:
astro-ph/0604154]; Gianluca Calcagni, Beatriz de Carlos, Antonio De Felice,
Nucl.Phys.B752:404-438,2006. [e-Print: hep-th/0604201]; A. Gruzinov, M. Kleban, hep-th/0612015;
[8] A. Adams, N. Arkani-Hamed, S. Dubovsky, A. Nicolis, R. Rattazzi, JHEP 0610:014, (2006),
[hep-th/0602178].
[9] G.M. Shore, e-Print: hep-th/0701185.
[10] Timothy J. Hollowood, Graham M. Shore, arXiv:0707.2303; Timothy J. Hollowood, Graham M.
Shore, arXiv:0707.2302.
[11] D. Blochinzev, Space and time in microworld (in Russian), Nauka, 1970.
[12] A. Hashimoto and N. Itzhaki, Phys.Rev. D63 (2001) 126004, [hep-th/0012093]; K. Landsteiner, E.
Lopez and M. H. G. Tytgat, JHEP 0106 (2001) 055, [hep-th/0104133]; Horatiu Nastase,
hep-th/0601182.
[13] T. Jacobson, D. Mattingly, Phys.Rev.D70:024003 (2004), e-Print: gr-qc/0402005.
[14] I.T. Drummond and S. J. Hathrell, Phys. Rev. D 22 (1980) 343; R. D. Daniels and G.M. Shore,
Nucl. Phys. B425 (1994) 634; R. D. Daniels and G.M. Shore, Phys. Lett. B367 (1996) 75.
[15] Y. Ohkuwa, Progr. Theor. Phys. 65 1981 1058.
[16] K. Scharnhortst, Phys. Lett. B236 (1990) 354; G. Barton, Phys. Lett. B237 (1990) 559;
J. I. Lattorre, P. Pascual, and R. Tarrach, Nucl. Phys. B437 (1995) 60; S. Ben-Menahem, Phys.
Lett. B250 (1990) 133.
[17] G. Shore, Nucl. Phys. B460 (1996) 379, [gr-qc/9504041]; A.D. Dolgov and I.B. Khriplovich, Zh.
Eksp. Teor. Fiz. 58 (1983) 1153; [English translation: Sov. Phys. JETP, 58 (1983) 671]; I. B.
Khriplovich, Phys. Lett. B 346 (1995) 251.
[18] A. D. Dolgov, I. D. Novikov, Phys.Lett. B442: 82 (1998), [gr-qc/9807067].
[19] S. Liberati, S. Sonego, M. Visser, Annals Phys. 298:167-185, (2002), [gr-qc/0107091].
[20] S.W. Hawking, Phys.Rev. D46, 603-611, (1992).
[21] K. Gdel, Rev. Mod. Phys. 21, 447 (1949).
[22] J. R. Gott, Phys. Rev. Lett. 66, 1126 (1991).
[23] A. Ori, gr-qc/0701024.
[24] A. Ori, Phys. Rev. Lett. 71, 2517 (1993); Y. Soen and A. Ori, Phys. Rev. D54, 4858 (1996); A. Ori,
Phys. Rev. Lett. 95, 021101 (2005); R.L. Mallett, Found. Phys. 33, 1307 (2003); W. B. Bonnor and
B.R. Steadman, Gen. Rel. Grav. 37:1833-1844 (2005); O. Gron and S. Johannesen, gr-qc/0703139.
[25] M. S. Morris, K.S. Thorne and U. Yurtsever, Phys. Rev. Lett. 61, 1446 (1988).
[26] C. Armendariz-Picon, V. F. Mukhanov, P. J. Steinhardt, Phys.Rev.Lett. 85:4438-4441, (2000),
[astro-ph/0004134]; C. Armendariz-Picon, V.F. Mukhanov, P.J. Steinhardt, Phys. Rev. D63:103510,
(2001), [astro-ph/0006373].

31
[27] V. F. Mukhanov, A. Vikman, JCAP 0602:004, (2006), [astro-ph/0512066]; A. Vikman,
astro-ph/0606033.
[28] J. Garriga, V. Mukhanov, Phys. Lett. B458:219-225 (1999), [hep-th/9904176].
[29] E. Babichev, V. Mukhanov, A. Vikman, JHEP 0609:061,(2006),[hep-th/0604075]; E. Babichev,
V. Mukhanov, A. Vikman, arXiv:0704.3301v1 [hep-th] .
[30] C. Bonvin, C. Caprini, R. Durrer, arXiv:0706.1538 [astro-ph];
[31] J.-P. Bruneton, Phys. Rev. D75:085013, (2007), [gr-qc/0607055]; J.-P. Bruneton hep-th/0612113;
J.-P. Bruneton, G. Esposito-Farese, arXiv:0705.4043v1 [gr-qc].
[32] Jin U Kang, Vitaly Vanchurin, Sergei Winitzki, arXiv:0706.3994 [gr-qc];
[33] C. Armendariz-Picon, E. A. Lim, JCAP 0508:007, (2005), [astro-ph/0505207].
[34] V. Mukhanov, Physical foundations of cosmology. Cambridge. Univ. Pr. (2005)
[35] James M. Cline, Sangyong Jeon, Guy D. Moore, Phys.Rev. D70 (2004) 043543, [hep-ph/0311312];
Sean M. Carroll, Mark Hoffman, Mark Trodden, Phys.Rev. D68 (2003) 023509
[arXiv:astro-ph/0301273v2]; I. Ya. Arefeva, I.V. Volovich, arXiv:hep-th/0612098v1; R. P. Woodard,
arXiv:astro-ph/0601672v2.
[36] A.D. Rendall, Class. Quant. Grav. 23, 1557-1570, (2006), [gr-qc/0511158].
[37] M. Novello, M. Makler, L. S. Werneck, C. A. Romero, Phys.Rev.D71 (2005) 043515,
[astro-ph/0501643].
[38] G. W. Gibbons, C. A. R. Herdeiro, Phys. Rev. D63:064006, (2001), [hep-th/0008052]. G. Gibbons,
Koji Hashimoto, Piljin Yi, JHEP 0209:061,2002, [hep-th/0209034];
G. W. Gibbons, Class. Quant. Grav.20:S321-S346, (2003), [hep-th/0301117]; G. W. Gibbons,
Rev.Mex.Fis.49S1:19-29, (2003), [hep-th/0106059].
[39] R. Wald, General relativity, The University of Chicago Press, (1984).
[40] J. Leray, Hyperbolic differential equations, mimeographed notes. Institute for Advanced Study,
Princeton (1953).
[41] H. Friedrich, A. Rendall, The Cauchy problem for the Einstein equations. In B. G. Schmidt (ed)
Einsteins Field Equations and Their Physical Implications. Lecture Notes in Physics 540. Springer,
Berlin (2000), [gr-qc/0002074]; I. G. Petrovsky, Lectures on partial differential equations,
Interscience Publishers, New York-London, (1954).
[42] T. Damour, I. I. Kogan, Phys. Rev. D 66 (2002) 104024, [arXiv:hep-th/0206042]; D. Blas, C.
Deffayet, J. Garriga, Class. Quant. Grav. 23 (2006) 1697, [arXiv:hepth/ 0508163]; D. Blas, C.
Deffayet, J. Garriga, arXiv:0705.1982.
[43] Paolo Creminelli, Markus A. Luty, Alberto Nicolis, Leonardo Senatore, JHEP 0612:080, (2006).
e-Print: hep-th/0606090; Paolo Creminelli, Leonardo Senatore, e-Print: hep-th/0702165; Evgeny I.
Buchbinder, Justin Khoury, Burt A. Ovrut, e-Print: hep-th/0702154; Evgeny I. Buchbinder, Justin
Khoury, Burt A. Ovrut, e-Print: arXiv:0706.3903 [hep-th].
[44] A. Vikman, Phys.Rev.D71:023515, (2005), [astro-ph/0407107].
[45] Robert R. Caldwell, Michael Doran, Phys.Rev.D72:043527, (2005), [astro-ph/0501104]; Anjan
Ananda Sen, JCAP 0603:010, (2006), [astro-ph/0512406]; Luis Raul Abramo, Nelson Pinto-Neto,
Phys.Rev.D73:063522, (2006), [astro-ph/0511562]; Gong-Bo Zhao, Jun-Qing Xia, Mingzhe Li, Bo
Feng, Xinmin Zhang, Phys.Rev.D72:123515, (2005), [astro-ph/0507482]; Martin Kunz, Domenico
Sapone, Phys.Rev.D74:123503, (2006), [astro-ph/0609040].
[46] C. Armendariz-Picon, T. Damour, V. Mukhanov, Phys.Lett. B458:209-218, (1999),
[hep-th/9904075].
[47] Ferdinand Helmer, Sergei Winitzki, Phys.Rev.D74:063528,(2006); [gr-qc/0608019].
[48] M. Alishahiha, E. Silverstein, D. Tong, Phys. Rev. D70:123505 (2004) [hep-th/0404084];
E. Silverstein, D. Tong, Phys. Rev. D70: 103505, (2004) [hep-th/0310221].

32
[49] E. Fradkin and A. Tseytlin Phys. Lett. 15BB (1985), p. 316; David J. Gross, Edward Witten,
Nucl.Phys.B 277:1, (1986); B. Zwiebach, Phys.Lett. 156B (1985) 315; S. Deser and A.N. Redlich,
Phys.Lett. 176B (1986) 350; A. A. Tseytlin, Phys.Lett. 176B (1986) 92; D.R.T. Jones and A.M.
Lowrence, Z.Phys. 42C (1989) 153.
[50] R. C. Tolman, The Theory of the Relativity of Motion, Berkeley, Univ. of California Press, (1917).
[51] D. Mugnai, A. Ranfagni, and R. Ruggeri, Phys. Rev. Lett. 84, 4830-4833 (2000); L. J. Wang, A.
Kuzmich, and A. Dogarlu, Nature 406, 277-279 (2000).
[52] V.S. Vladimirov, Equations of mathematical physics, MIR (1984).
[53] L. Landau, E. Livshitz, Course of Theoretical Physics, Vol. 2, The classical Theory of Fields,
Butterworth-Heinemann, (1987).
[54] M. Visser, Phys.Rev. D46:2445-2451 (1992) [hep-th/9203057].
[55] V. Moncrief, Astrophysical Journal, Part 1, 235, 1038-1046 (1980).
[56] M. Born, L. Infeld, Proc. Roy. Soc. Lond. A144:425-451, (1934).
[57] E. Babichev, Phys.Rev.D74:085004 (2006), [hep-th/0608071]; D. Bazeia, L. Losano, R. Menezes,
J.C.R.E. Oliveira, hep-th/0702052; X. Jin, X. Li, D. Liu, Class.Quant.Grav.24:2773-2780 (2007),
arXiv:0704.1685 [gr-qc].
[58] Gary N. Felder, Lev Kofman, Alexei Starobinsky, JHEP 0209:026, (2002), [hep-th/0208019].
[59] W. J. van Stockum, Pros.R.Soc.Edinb. 57, 135 (1937); F. J. Tipler, Phys. Rev. D 9, 2203 (1974).
[60] W. G. Unruh, Phys.Rev.Lett.46:1351-1353, (1981); Carlos Barcelo, Stefano Liberati, Matt Visser,
Living Rev.Rel.8:12, (2005), [gr-qc/0505065]; Matt Visser, Carlos Barcelo, Stefano Liberati,
Gen.Rel.Grav.34:1719-1734, (2002), [gr-qc/0111111]; Artificial black holes. M. Novello, (ed.) , M.
Visser, (ed.), G. Volovik, (ed.), River Edge, USA: World Scientific (2002).
[61] B. Barbashov, N. Chernikov, Soviet Physics JETP 5, 50, 1296-1308 (1966); B. Barbashov, N.
Chernikov, Soviet Physics JETP 51, 658 (1966).
[62] G. B. Whitham, Linear and nonlinear Waves (Pure & Applied Mathematics), John Wiley & Sons
Inc; (1974)
[63] Niayesh Afshordi, Daniel J.H. Chung, Michael Doran, Ghazal Geshnizjani, Phys.Rev.D75:123509,
(2007), [astro-ph/0702002]; Niayesh Afshordi, Daniel J.H. Chung, Ghazal Geshnizjani,
Phys.Rev.D75:083513, (2007), [hep-th/0609150].

33