A bimetric cosmological model based on

Andreï Sakharov’s twin universe approach

∗1
Jean-Pierre Petit , Florent Margnat†2 , and Hicham Zejli ‡3

1
Manaty Research Group
2
University of Poitiers
3
Manaty Research Group
arXiv:2412.04644v1 [gr-qc] 5 Dec 2024

Contents
1 Introduction 2

2 The physical interpretation of time Inversion (T-Symmetry) 4

3 Geometrical interpretation of electric charge 9

4 Matter-antimatter symmetry (C-symmetry) 10

5 Group associated with A. Sakharov’s model: the Janus group 13

6 Topology of the Janus model 15

7 Introducing negative masses: First approach 18

8 A paradigm shift to escape the crisis of today’s cosmology 20

9 Foundation of the Janus cosmological model 22

10 Construction of a time-dependent, homogeneous and isotropic solution 24

11 Interaction laws and observational consequences 25

12 The mathematical and physical consistency of the model 28

12.a Newtonian approximation of the field generated by a positive mass M . . . . . . 28
12.b Newtonian approximation of the field generated by a negative mass M . . . . . . 31

13 Dipole repeller prediction 33

14 Beyond the Newtonian approximation 34

15 Conclusion 34

References 36

∗
[Link]@[Link]
†
[Link]@[Link]
‡
[Link]@[Link]

1
 Abstract
The standard cosmological model, based on Cold Dark Matter and Dark Energy (ΛCDM),
faces several challenges. Among these is the need to adjust the scenario to account for
the presence of vast voids in the large-scale structure of the universe, as well as the early
formation of the first stars and galaxies. Additionally, the observed matter-antimatter
asymmetry in the universe remains an unresolved issue. To address this latter question,
Andrei Sakharov proposed a twin universe model in 1967. Building upon this idea and
introducing interactions between these two universe sheets through a bimetric model, we
propose an alternative interpretation of the large-scale structure of the universe, including
its voids and the acceleration of cosmic expansion.

1 Introduction
Between 1967 and 1980, the physicist Andreï Sakharov published several papers ([18, 19, 20]) in
which he presented a cosmological model with two universes, connected by an initial singularity:
the Big Bang. The first universe corresponds to ours, while the second is described by Sakharov
as a twin universe. The "arrows of time" of these two universes are antiparallel, and they
are "enantiomorphic", that is, mirrored. Through this model, Sakharov proposed a possible
explanation for the apparent absence of primordial antimatter in our universe.

For more than half a century, cosmology has been unable to solve one of its greatest enigmas:
not only has no convincing explanation been found as to why one particle of matter in a million
escaped total annihilation with antimatter, but no significant observation of a corresponding
amount of primordial antimatter has been made.

Sakharov was interested in the violation of CP-symmetry, a fundamental property of the

laws of physics, and hypothesized that a twin universe, where these violations would be reversed,
could exist. This model would thus restore a generalized symmetry on a large scale. Based on
the fact that matter is formed from the assembly of quarks and antimatter from antiquarks, he
supposed that in our universe, the reaction leading to the formation of matter would have been
slightly faster than the one leading to the formation of antimatter, while the inverse situation
would occur in the twin universe.

Thus, in our matter-dominated universe, there would remain a small surplus of matter,
accompanied by an equivalent amount of free antiquarks. Symmetrically, in the twin universe,
one would find antimatter with a corresponding surplus of free quarks. Although this model may
seem exotic, it nevertheless offers the only theoretical explanation proposed so far to account for
the disappearance of half of the predicted cosmic content. Consequently, it seems legitimate to
examine in detail the aspects and implications of such a model.

This article revisits the pioneering work of Andreï Sakharov and proposes a new cosmological
model, inspired by his approach, in which two folds of the universe are connected by the same
initial singularity, folded over one another and interacting through gravitational effects. It puts
this work in perspective with modern concepts to address some of the challenges posed by the
standard cosmological model, particularly those of the ΛCDM model. This model offers potential
explanations for phenomena such as the acceleration of cosmic expansion or the existence of
large-scale structures like cosmic voids.

Our paper is structured around several key sections. The first explores T-symmetry, which
corresponds to time reversal, based on the mathematical framework of the Poincaré group. This
symmetry is related to the existence of particles with negative mass and energy, at the core of
the bimetrical Janus model, inspired by the work of J.-M. Souriau. It plays a central role in the

2
dynamics of this double spacetime, where time reversal opens the door to a new interpretation of
physical phenomena ([22, 24]).

Next, C-symmetry, associated with charge conjugation, is extended within the framework of
an additional dimension through the Kaluza-Klein model. This extension allows the interpreta-
tion of electric charge as a geometric component, in accordance with Noether’s theorem. This
connection between the extra dimension and charge conservation offers a new perspective on
charged particles in a five-dimensional spacetime, where charge naturally emerges from geometry
([24]).

The model is enriched by the introduction of the Janus restricted group, which extends the
Kaluza space to several compactified dimensions. This dynamic group links the matter-antimatter
symmetry (C-symmetry) to the inversion of quantum charges in a multidimensional framework.
Through this extension, the group’s geometry allows for the understanding of the quantization of
several charges, including electric charge, and opens the way to the emergence of new quantum
charges. This section establishes a connection between Souriau’s work and the Kaluza-Klein for-
malism to explain complex physical phenomena in a higher-dimensional spacetime ([9, 21, 23, 24]).

The Janus dynamic group, which combines PT-symmetry (simultaneous inversion of energy,
time, and spatial coordinates) and C-symmetry (charge conjugation), allows the modeling of
interactions between matter, antimatter, and negative mass particles. Thanks to Noether’s
theorem, this group associates scalar invariants with the observed symmetries, thus clarifying
the interactions between these different entities within a bimetrical framework, and allowing the
extension of Sakharov’s model by adding compactified dimensions for each quantum charge ([16,
18, 24]).

To illustrate this concept, in the context of our study on bimetric models, we proposed a
model of wormhole linking two PT-symmetric folds of the universe via a modified Einstein-Rosen
bridge [10]. This model includes a cross term dr dt in the corresponding metric, leading to a finite
free-fall time to the wormhole’s throat for an external observer. The two folds are CPT-symmetric
for photons, which are neutral particles. This wormhole model allows for unidirectional traversal
through its throat, inducing a space-time inversion. This opens the possibility of interactions
between matter and antimatter, arising from the PT symmetry observed during the transition
between the two universe folds. Thus, the congruent identification of points on the two universe
folds and the reversal of the arrow of time induce an inversion of energy, offering new insights
into the structure of space-time and the potential inversion of particle mass while crossing this
bridge.

The Janus model will also be studied from a topological perspective, with a closed universe
geometry where P and T symmetries naturally emerge. Spacetime is modeled by a compact
universe with the topology of a 4-dimensional sphere S4 , which forms a two-fold cover of the
projective space P4 . In this structure, the antipodal points, representing the Big Bang and the
Big Crunch, coincide. By replacing these singularities with a tubular structure, they disappear,
allowing P and T symmetries to emerge as natural consequences of this closed projective geometry
(P4 ) and be interpreted in a purely topological framework ([3, 13, 15]).

One of the earliest attempts to introduce negative masses into a cosmological model, explored
by H. Bondi in 1957, showed that the coexistence of positive and negative masses, which re-
spectively induce attraction and repulsion, leads to the "runaway effect" ([2]). In this effect, a
positive mass and a negative mass attract gravitationally while moving away from each other,
thus violating the action-reaction principle. This effect has remained a major challenge for

3
integrating negative masses into standard cosmology.

Thus, to resolve the crisis of modern cosmology, the Janus model proposes a paradigm shift.
Since the 1970s, the ΛCDM model has failed to explain certain observed phenomena, such as the
rotation speeds of galaxies and the acceleration of cosmic expansion. The Janus model, based on
a bimetrical geometry with positive and negative masses evolving on distinct geodesics, offers
an alternative. It proposes a new approach to solving anomalies such as the rapid formation
of galaxies after the Big Bang and discrepancies in the measurement of the Hubble constant ([6, 8]).

The Janus model proposes a bimetrical system where gravitational interactions between
positive and negative masses are described by distinct field equations, each associated with its
own metric. The construction of a homogeneous, isotropic, and time-dependent solution in the
Janus model relies on FLRW-type metrics, respectively describing the universes of positive and
negative masses. A common energy conservation relation is established, proposing an exact
solution for dust universes, where the observed cosmic acceleration is interpreted as a negative
total energy. Numerical comparisons confirm the model’s compatibility with observations, as
illustrated by the magnitude-redshift curve. The interaction laws in the Janus model reveal
that masses of the same sign attract, while those of opposite signs repel, thus eliminating the
"runaway effect". The model reproduces local observations of general relativity while replacing
dark matter and dark energy with invisible negative masses. These negative masses form void-like
structures that confine positive mass, accelerating star and galaxy formation in the first few
hundred million years, in agreement with data from the James Webb telescope and observations
of large cosmic voids ([4, 12, 14, 16, 17]).

Finally, the mathematical consistency of the Janus model is demonstrated in the weak field
limit, thanks to the generalized conservation of energy and the Bianchi identities. The calculation
of Schwarzschild metrics for positive and negative masses shows that masses of the same sign
attract, while those of opposite signs repel. The model satisfies the Tolman-Oppenheimer-Volkoff
equations in the Newtonian approximation, while remaining compatible with cosmological ob-
servations. It is also valid in regions dominated by negative masses, such as the dipole repeller,
where it predicts a negative gravitational lensing effect, dimming the luminosity of background
objects ([1, 11, 8, 10]).

In summary, our model proposes an extension of general relativity by introducing two distinct
metrics, each associated with a type of mass, allowing for the explanation of both the acceleration
of the universe’s expansion and certain large-scale structures, while remaining compatible with
local observations of general relativity. This analysis opens new perspectives and places the
Janus model among the approaches that can be tested by modern cosmological observations.

2 The physical interpretation of time Inversion (T-Symmetry)

The T-symmetry refers to the inversion of the time coordinate. In 1970, contributing to the
development of symplectic geometry and its application to physics, mathematician J.-M. Souriau
provided the physical interpretation of this inversion of the time coordinate ([24]). The Gram
matrix defining the Minkowski space is :
Ü ê
1 0 0 0
0 −1 0 0
G= . (1)
0 0 −1 0
0 0 0 −1

4
Its isometry group is the Poincaré group:
Å ã
L C
. (2)
0 1

Where L is the matrix representing the Lorentz group Lor which describes how spacetime
coordinates change between different inertial frames. These transformations include rotations in
space as well as Lorentz transformations (boosts), which are changes of reference frames moving
at a constant speed relative to each other. it’s axiomatically defined by:

LT GL = G, (3)

and C is the quadrivector of space-time translations in R1,3 as follows:

Ü ê
∆t
∆x
C= . (4)
∆y
∆z

It acts on points in Minkowski space:

Ü ê
t
x
ξ= . (5)
y
z

This Lie group with 10 independent parameters1 is the isometry group of this space, defined
by its metric:
ds2 = dt2 − dx2 − dy 2 − dz 2 . (6)
The Lorentz group Lor has four connected components:
• Lorn is the neutral component (its restricted subgroup), does not invert either space or
time and is defined by:

Lorn = {L ∈ Lor, det(L) = 1 ∧ [L]00 ≥ 1}

• Lors inverts space and is defined by:

Lors = {L ∈ Lor, det(L) = −1 ∧ [L]00 ≥ 1}

• Lort inverts time but not space and is defined by:

Lort = {L ∈ Lor, det(L) = 1 ∧ [L]00 ≤ −1}

• Lorst inverts both space and time and is defined by:

Lorst = {L ∈ Lor, det(L) = −1 ∧ [L]00 ≤ −1}

And we have:
Lor = Lorn ⊔ Lors ⊔ Lort ⊔ Lorst . (7)
The first two components are grouped to form the so-called “orthochronous” subgroup:

Loro = Lorn ⊔ Lors . (8)

1
Including the 6 independent parameters of the Lorentz group (3 rotations and 3 boosts) and 4 independent
transformations, which are translations in the 4 directions of Minkowski space.

5
It includes P-symmetry, which poses no problem for physicists who know that there are photons
of "right" and "left" helicity whose motions are derived from this symmetry. This corresponds
to the phenomenon of the polarization of light.

The last two components form the subset "retrochronous" or "antichronous", whose compo-
nents invert time:
Lora = Lort ⊔ Lorst . (9)
Thus, we have:
Lor = Loro ⊔ Lora . (10)
Noting that:
Lort = −Lors Lorst = −Lorn . (11)
The Poincaré group inherits the properties of the Lorentz group and thus has four connected
components, it is defined by:
ßÅ ã ™
L C
g := , L ∈ Lor ∧ C ∈ R1,3 , (12)
0 1

acting on Minkowski space as follows:

g(X) = L.X + C. (13)

The action of the group on its space of moments is the action on the dual of the Lie algebra
of the group2 . The element of the Lie algebra is obtained by differentiating the ten components
of the group. Souriau designates by the Greek letter Λ the differential of the square matrix Z
representing the element of the Poincaré group, and by the Greek letter Γ the element of the
subgroup of spatio-temporal translations3 :
ßÅ ã ™
Λ Γ
Z := , Λ̄ = −Λ ∧ Γ ∈ R1,3 . (14)
0 0

The elements of the Lorentz group act on points in spacetime, transforming one motion into
another. By applying an element L of the Lorentz group to a given motion, we obtain a new
motion. The neutral component Lorn is a subgroup containing the identity matrix that inverts
neither space nor time.

Let’s consider the 4-component matrix ω made up of two parameters λ1 and λ2 :

Ü ê
λ1 0 0 0 ®
0 λ2 0 0 λ1 = ±1
ω(λ1 ,λ2 ) = with (15)
0 0 λ2 0 λ2 = ±1
0 0 0 λ2
2
Souriau’s approach, thanks to the Poincaré group which is the isometry group of Minkowski space encompassing
the Lorentz group (with its four connected components), allows the parameters associated with each of these
motions, whose representative points belong to a vector space, the space of moments, to emerge. The dimension
of this space is equal to that of the group: ten. Indeed, the Lorentz group is made up of transformations that
preserve the quadratic form of space-time. It consists of the orthochronous Lorentz transformations and the
translation group. The transformations of the orthochronous Lorentz group have 6 degrees of freedom, while the
translation group has 4 degrees of freedom. This structure leads to 10 independent parameters of the Poincaré
group. By combining them into an antisymmetrical matrix called a torsor, the parameters of the space of motions
can thus be defined.
3
(13.54) of [24]. He then writes µ, an element of the space of motions, in the form (13.57) and expresses the
invariance in the form of the constancy of the scalar (13.58), where M is an antisymmetric matrix.

6
Thus, the four components of the Lorentz group can be easily expressed using the four possible
combinations of these two parameters applied to its neutral component, of which an element
Ln ∈ Lorn is expressed according to the expression L = ωLn :
Ü ê Ü ê
1 0 0 0 1 0 0 0
0 1 0 0 0 −1 0 0
ω(1,1) × Ln = ∈ Lorn ω(1,−1) × Ln = ∈ Lors
0 0 1 0 0 0 −1 0
Ü 0 0 0 1 ê Ü 0 0 0 −1 ê
−1 0 0 0 −1 0 0 0
0 1 0 0 0 −1 0 0
ω(−1,1) × Ln = ∈ Lort ω(−1,−1) × Ln = ∈ Lorst
0 0 1 0 0 0 −1 0
0 0 0 1 0 0 0 −1
(16)
We note that λ1 = −1 inverts time while λ2 = −1 inverts space. The four components
are grouped into two subsets “orthochronous” and “retrochronous” according to the respective
expressions 8 and 9.

The Poincaré group can then be written according to these four connected components as
follows: ßÅ ã ™
ωLn C
g := , ωLn ∈ Lor ∧ C ∈ R 1,3
. (17)
0 1
Thus, the action of this Poincaré group on the spacetime coordinates yields the following space
of motions: ï ò ï ò ï ò
ωLn C ξ ωLn ξ + C
× = . (18)
0 1 1 1
Indeed, this describes the action of the Poincaré group on its space of moments µ, consisting of
ten independent scalar quantities:
• The energy E,
• The momentum p = {px , py , pz },
• The passage f = {fx , fy , fz },
• The spin s = {lx , ly , lz }.
If we consider the motion of an object in space. Such motion is also defined by its moment µ.
The physicist can then apply an element G, for example from the Galilean group, to this moment
µ. This produces a new moment µ′ . This action can be written as follows:
µ′ = GµG T , (19)
where G T represents the transpose of this matrix G. µ is an antisymmetric moment matrix of
size 5 × 54 where the more compact form is defined as follows:
Å ã
M −P
µ= , (20)
PT 0
with5 : Ü ê Ü ê
0 −lz ly fx E
lz 0 −lx fy px
M= , P = . (21)
−ly lx 0 fz py
−fx −fy −fz 0 pz
4
Meaning the symmetric elements with respect to the main diagonal have opposite signs. The elements on the
main diagonal are equal to zero, as each is its own opposite.
5
M is the moment matrix associated with µ with dimensions 4 × 4, and P , a four-vector energy-momentum
with dimensions 4 × 1.

7
 Then, by applying the action of the Poincaré group 12 on the dual of its Lie algebra, i.e., on
its space of moments, we obtain the following action according to 19 :
Å ãÅ ãÅ T ã
L C M −P L 0
µ′ = , (22)
0 1 PT 0 CT 1

LM LT − LP C T + CP T LT −LP
Å ã
′
µ = . (23)
P T LT 0
We can deduce6 :
M ′ = LM LT + CP T LT − LP T C, (24)
and
P ′ = LP. (25)
Therefore, the torsor of Poincaré group is given by the different components of the space of
moments7 as follows:
µ = {M, P } = {l, g, p, E}, (26)
where l is the angular momentum of M , g is the relativistic barycenter of M , p is the linear
momentum of P and E is the energy of P .

Now, let’s consider for example the symmetry T, where there is only a time inversion
(λ1 = −1), without space inversion (λ2 = 1), in a case where there is also no translation in
spacetime (C = 0). We thus have:

ω(−1,1) × Ln = Lt . (27)

Hence: Ü ê Ü ê Ü ê
−1 0 0 0 t -t
0 1 0 0 x x
Lt × ξ = × = . (28)
0 0 1 0 y y
0 0 0 1 z z
Thus, we obtain the action of time inversion in the space of trajectories or in spacetime.

The second equation (25) sheds light on the physical significance of this inversion of the time
coordinate. Indeed, the application of the Lt component of the Lorentz group to the motion of a
particle gives: Ü êÜ ê Ü ê
−1 0 0 0 E -E
0 1 0 0 px px
P ′ = Lt P = = . (29)
0 0 1 0 py py
0 0 0 1 pz pz
Therefore, we can deduce that the application of the Lt component of the Lorentz group to the
motion of a particle induces an inversion of its energy from E to -E.

The T symmetry applied to the motion of a particle inverts its energy which leads to mass
inversion8 following the definition of the mass9 as:
√
m = P T · P sgn(E). (30)
6
(13.107) of [24].
7
(13.57) of [24].
8
page 198-199 of [24].
9
(14.57) on page of [24].

8
A very detailed commentary on the work can be found in reference [22]. The approach is based
on the introduction of the space of motions as a dual of the Lie algebra of the group.

In this context, we uncover the physical interpretation of the model proposed by A. Sakharov:
the second universe in his framework could consist of particles possessing both negative energy
and negative mass.

To further extend the interpretation of fundamental symmetries, we now turn our attention to
C-symmetry, which is associated with charge conjugation. By introducing a higher-dimensional
framework inspired by Kaluza-Klein theory, we can offer a geometrical interpretation of electric
charge, according to Noether’s theorem. This will allow us to explore the relationship between
spacetime transformations and the emergence of electrically charged particles.

3 Geometrical interpretation of electric charge

The geometrical interpretation of C-symmetry, which is synonymous with charge conjugation and
matter-antimatter duality, was provided by J.-M. Souriau in 1964 in chapter V of reference [24].

Let’s apply an extension of the Poincaré group to form the following dynamic group:
Ñ é 
 1 0 ϕ 
g := 0 L C , ϕ ∈ R ∧ L = λLo ∈ Lor ∧ λ = ±1 ∧ C ∈ R 1,3
. (31)
0 0 1
 

Starting from Minkowski space: Ü ê

t
Å ã
x t
ξ= = , (32)
y r
z
let’s introduce Kaluza space10 that incorporates a 5 × 5 Gram matrix:
à í
1 0 0 0 0 Ü ê
1 0 0 0
0 −1 0 0 0 Å ã
G 0 0 −1 0 0
Γ = 0 0 −1 0 0 = where G = . (33)
0 −1 0 0 −1 0
0 0 0 −1 0
0 0 0 −1
0 0 0 0 −1
In the considered group, we just add a translation ϕ to the fifth dimension ζ. Thus, the
dimension of the group becomes 11. It is the isometry group of Kaluza space, defined by its
metric:
ds2 = dX T Γ dX = dt2 − dx2 − dy 2 − dz 2 − dζ 2 , (34)
with : à í
t
Å ã x
ξ
X= = y . (35)
ζ
z
ζ
According to Noether’s theorem11 , this new symmetry is accompanied by the invariance of a
scalar that we will call q. The torsor of this Kaluza group then incorporates an additional
10
Kaluza space is a hyperbolic Riemannian manifold with signature (+ − − − −).
11
Noether’s theorem states that for every continuous symmetry of a physical action, there exists a conserved
quantity. In our context, if a new symmetry ensures the invariance of a scalar q, this scalar is the conserved
quantity. This means that q remains constant when the symmetry is applied to the system’s action.

9
parameter:
µ = {M, P, q} = {l, g, p, E, q}. (36)
Let’s introduce the action of the group on its Lie algebra:

Z ′ = g −1 Zg. (37)

If we consider an element of the Lie algebra of this group:

δϕ′
Ñ é Ñ é
0 0 δϕ 0 0
Z = 0 Gω γ Z ′ = 0 Gω ′ γ ′ , (38)
0 0 0 0 0 0

we obtain:
δϕ′
Ñ é Ñ é
0 0 0 0 δϕ
Z′ = 0 Gω γ ′
′ = 0 L−1 GωL L−1 GωC + L−1 γ . (39)
0 0 0 0 0 0
This allows us to deduce the action of the following group:

q ′ = q, (40)
′ T
M = LM L − LP C + CP L , T T T
(41)
′
P = LP. (42)

If we identify q as the electric charge, this would show that the motion of a massive particle in a
five-dimensional space would transform it into an electrically charged particle.

The interpretation of C-symmetry within a higher-dimensional framework, as explored, leads

naturally to a broader geometric understanding of symmetries in the Janus model. Specifically,
the notion of charge conjugation extends to encompass the duality between matter and antimatter.
To develop this further, we now introduce the Janus restricted group, which provides a formal
structure to describe these symmetries. This group allows us to explore how quantum charges
can be inverted by compactified dimensions, linking the symmetry properties of spacetime to the
emergence of quantized charges and new quantum numbers.

4 Matter-antimatter symmetry (C-symmetry)

Let’s introduce the Janus restricted group as follows:
Ñ é 
 µ 0 ϕ 
g := 0 L C , µ = ±1 ∧ ϕ ∈ R ∧ L = λLo ∈ Lor ∧ λ = ±1 ∧ C ∈ R1,3 . (43)
0 0 1
 

The action of the group on the coordinates of the 5-dimensional spacetime defined by 35 yields
the space of the following motions:
Ñ éÑ é Ñ é
µ 0 ϕ ζ µζ + ϕ
0 L C ξ = Lξ + C . (44)
0 0 1 1 1

A similar calculation to the previous one yields the action of the group:

q ′ = µq, (45)
M ′ = LM LT − LP C T + CP T LT , (46)
′
P = LP. (47)

10
This group acts on the five-dimensional Kaluza space. We observe that µ = −1 reverses the fifth
dimension ζ and the scalar q.

Through a dynamic interpretation of the group, we find the idea suggested by J.-M. Souriau
[24]: the inversion of the fifth dimension is associated with the inversion of electric charge.
However, this is only one of the quantum charges. Indeed, the C-Symmetry translating the
"matter-antimatter" symmetry introduced by Dirac ([5]), reverses all quantum charges. This
inversion operation is only obtained by adding as many compactified dimensions as there are
quantum charges. The action of the group on the coordinates of n-dimensional spacetime yields
the space of the following motions:

µ 0 0 · · · 0 ϕ1  ζ 1   µζ 1 + ϕ1 
 
 0 µ 0 · · · 0 ϕ2  ζ 2   µζ 2 + ϕ2 
.  . 
 0 0 . . . · · · 0 ..  ..
   
.

. .
  
= . (48)
   
 .. ..
   
 p   p p
 . . · · · µ 0 ϕ  ζ   µζ + ϕ 
p

 0 0 · · · 0 L C  ξ   Lξ + C 
 

0 0 ··· 0 0 1 1 1

The torsor of this group incorporates several additional scalars q p :

p
X
µ = {M, P, q i } = {l, g, p, E, q 1 , q 2 , . . . , q p }. (49)
1

This allows us to obtain the action of the group on its momentum space:
1
q ′ = µq 1 , (50)
′1
q = µq 1 , (51)
... (52)
′p
q = µq , p
(53)
′ T
M = LM L − LP C + CP L , T T T
(54)
′
P = LP. (55)

Moreover, Souriau considers that electric charge can be geometrically quantized into discrete
values (+e, 0, −e) when the associated fifth dimension is closed.

Imagine representing motion in Minkowski space along a simple straight line oriented in time.
At each point, we add a closed dimension, which extends Minkowski space into a bundle. In the
didactic figure 1, it is represented as a cylinder. But in approach [21], these transformations
no longer a priori preserve the electric charge q, which then becomes dependent on the chosen
coordinate system. In reference [21], taking up the approach initiated in [23], the author opts for
a closed fifth dimension, in which the radius of this "universe tube" becomes very small, of the
order of Planck’s length. He then rediscovers the invariance of electric charge and concludes [21],
we quote:

In this paper, we revisit the Kaluza-Klein theory from the perspective of the
classification of elementary particles based on the coadjoint orbit method. The
keystone conjecture is to consider the electric charge as an extra momentum on an
equal footing with the mass and the linear momentum. We study the momentum
map of the corresponding symmetry group Ĝ1 which conserves the hyperbolic metric.
We show that the electric charge is not an invariant, i.e. it depends on the reference
frame, which is in contradiction with the experimental observations. In other words,

11
 Figure 1: Inversion of the winding direction of a particle’s motion reflecting the C-symmetry

it is not the symmetry group of the Universe today as we know it. To avert this
paradox, we scale the fifth coordinate and consider the limit when the cylinder radius
ω vanishes. For the corresponding group Ĝ0 also of dimension 15, the charge is an
invariant then independent of the frame of reference and the observer. On this ground,
we propose a cosmological scenario in which the elementary particles of the early
Universe are classified from the momenta of the group Ĝ1 , next the three former
dimensions inflate quickly while the fifth one shrinks, leading to the 4D era in which
as today the particles are characterized by the momenta of the group Ĝ0 . By this
mechanism, the elementary particles can acquire electric charge as a by-product of the
4 + 1 symmetry breaking of the Universe. This work opens the way to the geometric
quantization of charged elementary particles.

The expression for this characteristic dimension of this universal tube is given in [23] on page
412 :
…
ℏ χ
e , (56)
e 2π
χ being the Einstein constant taken equal to [23]:

8πG
χ=− = 1.856 × 10−27 cm g−1 . (57)
c2
By introducing numerical values, this characteristic length is 3.782 × 10−32 cm. Dividing by 2π
gives us the order of magnitude of Planck’s length. In this view, the quantization of electric
charge and its constancy are derived from the closure of the extra dimension associated with the
decrease in the characteristic dimension associated with it.

This group refers to an extension of the Poincaré group, i.e. to a field-free, curvature-free
universe. This construction of a five-dimensional relativity was suggested in 1964 in reference
[23] and has been taken up again more recently in [21]. Note that it is in [23], page 413, that the

12
link between charge conjugation and fifth-dimensional inversion is first mentioned.

By generalizing [9], we can envisage an extension of space-time to a space with 4 + p di-

mensions, all of which may see their characteristic dimensions reduced, like that of this fifth
dimension, each of these collapses leading to the emergence and quantization of new quantum
numbers, baryonic, leptonic, unique etc., the electric charge being only the first of these.

Thus, the Janus restricted group has provided us with a framework for understanding the
matter-antimatter symmetry (C-symmetry) and the inversion of quantum charges through
additional compactified dimensions. We can now extend it to a broader symmetry group
associated with A. Sakharov’s model, the Janus group, which incorporates both C-symmetry
and PT-symmetry. This extension allows us to explore a dynamic group structure that includes
negative masses and antimatter within the framework of Sakharov’s twin universe model.

5 Group associated with A. Sakharov’s model: the Janus group

If we want to construct a group that translates the T-symmetry invoked by Sakharov, we’ll
replace Lo by λLo with λ = ±1. But, as proposed in [16], we can translate what had already
been proposed [18], we quote:
All phenomena corresponding to t < 0 are, in this hypothesis, assumed to be CPT
images of phenomena corresponding to t > 0.
Then, by introducing a new symmetry to the previous Janus restricted group, which we can call
PT Symmetry allowing the conversion of matter into antimatter with negative mass12 , we thus
combine C-symmetry and PT-symmetry to form the Janus dynamic group as follows:
Ñ é 
 λµ 0 ϕ 
g := 0 λLo C , λ, µ ∈ {−1,1} ∧ ϕ ∈ R ∧ Lo ∈ Loro ∧ C ∈ R1,3 . (58)
0 0 1
 

We can consider that particles of matter and antimatter can coexist in the same space fold.
However, no coexistence is possible for the motion of particles deduced by T-symmetry (or
PT-symmetry). This space is of dimension 4 + p (for p quantum charges). We will therefore
consider the two-fold covering of this manifold Mn+p . In each of these two folds, there remains
a possibility to perform the symmetry corresponding to µ = −1, that is, the inversion of all
quantum charges. In other words, the "matter-antimatter" duality exists in both folds.

To understand the nature of the different components of these folds, we will consider the
motion of a particle of matter with energy and mass:
• By acting on this motion with elements of the group corresponding to (λ = 1 ; µ = 1), we
will obtain other motions of particles of matter with positive mass and energy.
• By acting on this motion with elements of the group corresponding to (λ = 1 ; µ = −1), we
will obtain other motions of antimatter particles with positive mass and energy13 .
• By acting on this motion with elements of the group corresponding to (λ = −1 ; µ = 1), we
will obtain other motions of particles of matter with negative mass and energy14 .
• By acting on this motion with elements of the group corresponding to (λ = −1 ; µ = −1),
we will obtain other motions of antimatter particles with negative mass and energy15 .
12
A concept we could call antimatter in the sense of Feynman ([7]).
13
These are "antimatter in the sense of Dirac" (C-symmetry).
14
CPT-symmetry.
15
These are "antimatter in the sense of Feynman" (PT-symmetry).

13
 Its isometry group is that of Janus space, defined by the same metric as structuring Kaluza
space (34), and its dimension is 1116 . The torsor of the group is also the same as (36).
However, if we consider an element of the Lie algebra of this group:
Ñ é
0 0 δϕ
Z = 0 λGω γ , (59)
0 0 1

we can then calculate Z ′ according to the relation 37 as follows:

δϕ′
Ñ é Ñ é
0 0 0 0 (λµ)δϕ
Z ′ = 0 λGω ′ γ ′ = 0 λ3 L−1 2 −1
o GωLo λ Lo GωC + λLo γ
−1 . (60)
0 0 1 0 0 0

Thus, by identification, we can deduce:

δϕ′ = λµδϕ, (61)

ω ′ = λ2 GL−1
o GωLo , (62)
′
γ = λ2 L−1
o GωC + λL−1
o γ. (63)

We know that:

L−1 T
o = GLo G. (64)

Then:
δϕ′ = λµδϕ,
ω ′ = λ2 LT
o ωLo , (65)
′ 2
γ =λ GLT
o ωC + λGLT
o Gγ.
However, inspired by J.-M. Souriau, we could add as many additional closed dimensions as
quantum charges and write the dynamic group as follows:

ϕ1
 
λµ 0 0 ··· 0
 0 λµ 0 · · · 0 ϕ2 
.. .. 
 
. . 

 0 0 ··· 0
 ..

..
. (66)
 . . · · · λµ p

 0 ϕ  
 0 0 · · · 0 λLo C 
0 0 ··· 0 0 1

The isometry group of this space can be defined by the following metric:

ds2 = (dt)2 − (dx)2 − (dy)2 − (dz)2 − (dζ 1 )2 − (dζ 2 )2 − . . . − (dζ p )2 . (67)

With :
t
 
x
 
y
Å ã z

ξ
X= = ζ 1 . (68)
 
ζ  
ζ 2 
.
 
 .. 
ζp
16
10 + 1 dimension associated with the fifth space dimension ζ that J.-M. Souriau identifies with the electric
charge q.

14
The action of this Janus group on the coordinates of 10 + p independant parameters then yields
the space of the following motions:
ϕ1  ζ 1   λµζ 1 + ϕ1 
 
λµ 0 0 ··· 0
 0 λµ 0 · · · 0 ϕ2  2  2 2 
 ζ   λµζ + ϕ 

. . ..   ..

 .. · · · .. 

 0 0 0 .  .

 = . (69)
 
 .. ..
   
p p p
 . . p  ζ   λµζ + ϕ 

· · · λµ 0 ϕ
  
 
ξ   λLo ξ + C 
 
 0 0 · · · 0 λLo C 
0 0 ··· 0 0 1 1 1

According to Noether’s theorem, this new symmetry is accompanied by the invariance of

additional scalars q p . Therefore, the torsor of the group integrates them according to this relation:
p
X
µ = {M, P, q i } = {l, g, p, E, q 1 , q 2 , . . . , q p }. (70)
1

Thus, the duality relation17 gives us:

p p
1 X 1 X i
Tr (M · ω) + P T · Gγ + δϕ q i = Tr (M ′ · ω ′ ) + P ′T · Gγ ′ + δϕ q′ . (71)
2 2
1 1

This allows us to deduce the action of the group by identification with 65:
p p
′i
X X
q = λµ qi, (72)
1 1
M ′ = LM LT − LP C T + CP T LT , (73)
′
P = LP. (74)

Having established the Janus dynamic group as a natural extension of Sakharov’s model,
which incorporates both PT-symmetry and C-symmetry, we now shift our focus to the topological
implications of the Janus model. In particular, we will explore how the symmetries discussed
earlier can emerge from a closed, higher-dimensional universe. This section delves into the
topological structure of the model, illustrating how P and T symmetries can arise naturally from
the geometry of a closed universe, modeled as a projective space P4 .

6 Topology of the Janus model

Let’s consider a universe closed in all its dimensions, including space and time (see figure 2).
Diametrically opposed, antipodal points can be brought into coincidence. The image is then that
of a P2 projective. The north and south poles, one representing the Big Bang and the other the
Big Crunch, come into coincidence. The sphere cannot be paved without the presence of these
two singularities. The same applies to any sphere S2n if n is even, especially if this dimension is
4. This geometry was proposed in [13].

The figure 3 shows how this coincidence of antipodal regions generates this T-symmetry. On
the S2 sphere, the direction of time is given by the orientation of the meridian curves. This
orientation is shown on the left at the new state of maximum expansion, when space is identified
with the sphere’s equator. During this folding of the S2 sphere, described in reference [15] page
65, the vicinity of this equator is configured as the two-folds cover of a Möbius strip with three
half-turns (see figure 3 on the right).

17
(13.58) from [24].

15
Figure 2: A simplified 2D representation of a closed universe with a spherical topology S2 ,
illustrating the temporal progression from the Big Bang to the Big Crunch, with the universe
reaching maximum spatial extension in between.

In figure 4, we evoke the appearance of T-symmetry by manipulating the vicinity of a merid-

ian line. In addition, we evoke the possible elimination of the Big Bang - Big Crunch double
singularity by replacing them with a tubular passage, which then gives this geometry the nature
of the two-fold cover of a Klein bottle.

For enantiomorphy and P-symmetry to appear, the operation would have to be performed
on a larger sphere. This aspect can be highlighted by considering the conjunction of antipodal
regions in the vicinity of a meridian line, which is then configured according to the two-fold
covering of a half-turn Möbius strip. The figure 5 illustrates this enantiomorphic situation.

By bringing the antipodal points of even-dimensional spheres into coincidence, we locally

create a configuration associating two T-symmetrical folds. By adding further dimensions,
the coincidence of the antipodes creates a two-fold CPT-symmetric coating configuration of a
projective space. In the case of the sphere S2 , which corresponds only to a 2D didactic image,
the image of the projective P2 is its immersion in, which corresponds to the surface described in
1903 by the German mathematician Werner Boy [3]. See figure 6. In this figure, we show how
the coincidence of the antipodal points of the equator of the sphere S2 gives the two-fold covering
of a Möbius ribbon with three half-turns.

In this section, we have demonstrated that the P and T symmetries invoked by A. Sakharov
can arise as consequences of a purely topological structure, specifically the covering of a projective
space P4 .

After exploring the topological structure of the Janus model, we now address a major
consequence of T-symmetry: the introduction of negative masses. According to J.-M. Souriau,
the application of T-symmetry to the motion of a particle inverts its energy, which leads to the

16
Figure 3: How the coincidence of antipodal regions creates T-symmetry. Drawing extracted from
[15], page 65.

Figure 4: Coincidence of antipodal regions on a sphere S2 , according to the two-folds cover of a

half-turn Möbius strip, with the appearance of T-symmetry.

Figure 5: P-symmetry as a consequence of contacting antipodal region neighborhoods on an S2

sphere.

17
 Figure 6: Boy’s surface, immersion of the P2 projective in R3 .

inversion of its mass18 , in accordance with the definition of mass19 . Although this idea is elegant,
it presents significant challenges when integrated into the framework of general relativity. In the
following section, we will propose an initial approach to incorporating negative masses into the
cosmological model, analyzing the implications of their interaction with positive masses and the
resulting geodesics.

7 Introducing negative masses: First approach

Using dynamical group theory, we showed that this T-symmetry was synonymous with the
introduction of negative masses into the cosmological model. A. Sakharov’s primordial antimatter
would therefore be endowed with negative mass. This first step is far from anecdotal since, if
we neglect it, we admit to losing nothing less than half the universe from the outset. Is it then
possible to introduce negative masses into the standard model of general relativity?

A first idea would be to consider that the field comes from two sources, represented by two
tensors, the first referring to a positive mass content and the second to a negative mass content:
1 î ó
Rµν − Rgµν = χ Tµν (+) (−)
+ Tµν . (75)
2
We can then consider the metric solution corresponding to a region where the field is created,
firstly by a positive mass content:
1 (+)
Rµν − Rgµν = χTµν . (76)
2
Geodesics are given by a solution in the form of an external metric:
Ç å
2GM (+) 2 2 dr2
2
− r2 dθ2 + sin2 θ dφ2 . (77)


ds = 1 − 2
c dt − (+)
c r 2GM
1−
c2 r
The geodesics evoke an attraction (see figure 7).
18
pages 198-199 of [24].
19
(14.57) on page 346 of [24].

18
Figure 7: Deflection of positive-energy neutrinos by a positive mass. The trajectories, when
passing near the mass, are deflected more strongly due to the gravitational effect. The angle of
deflection reaches its maximum (C) when the neutrinos graze the edge of the mass. Trajectories
further away, such as (D), experience a weaker deflection, and the deflection angle becomes null
for trajectories passing at a very large distance from the mass. The trajectory passing through
the center of the mass (A) remains undeflected due to the symmetry of the configuration.

Now consider the field created by a negative mass M (−) , the field equation becomes then:
1 (−)
Rµν − Rgµν = χTµν . (78)
2
And the solution corresponds to the metric:
Ñ é
2G M (−) dr2
ds2 = 1 + c2 dt2 − − r2 dθ2 + sin2 θ dφ2 . (79)


c2 r 2G M (−)
1+
c2 r
The geodesics then represent a repulsion (see figure 8).

In this context, our single field equation provides only a single family of geodesics, which the
test particles, with both positive and negative masses, must follow. We deduce that:

• Positive masses attract both positive and negative masses.

• Negative masses repel both positive and negative masses.

• Two masses of identical absolute values but opposite signs are brought together, the positive
mass flees, pursued by the negative mass. Both then accelerate uniformly, but without
any energy input, since the energy of the negative mass is itself negative. This result
was illustrated in 1957 by H. Bondi [2]. This phenomenon is known as "runaway effect".
What’s more, this scheme violates the action-reaction principle. In 1957, the conclusion was
reached that it was physically impossible to include negative masses in the cosmological
model. This would only be possible at the price of a profound paradigmatic shift, not by
denying the achievements of general relativity, but by considering its extension in a wider
geometric context.

After examining the introduction of negative masses and their implications within the
framework of general relativity, we now turn to a broader cosmological context. The discovery
of anomalies, such as the dipole repeller and the accelerating expansion of the universe, has
revealed significant shortcomings in the standard model ΛCDM. Recent observations, particularly
those made with the James Webb Space Telescope, have intensified the crisis in cosmology by
challenging long-held assumptions about galaxy formation. In the following section, we will

19
Figure 8: Deviation of positive-energy photons by a negative mass. The trajectories, when the
curvature remains moderate, are very close to hyperbolas. The angle of deviation reaches a
maximum (C) when the geodesic is tangent to the limit of the mass. It then decreases steadily
to zero at very large distances (D). The angle of deviation is null, due to symmetry, when the
geodesic passes through the center of the mass (A).

explore how the Janus cosmological model offers a paradigm shift capable of resolving these
issues by proposing a bimetrical structure for the universe, integrating both positive and negative
masses into a broader and more innovative geometric framework.

8 A paradigm shift to escape the crisis of today’s cosmology

In the mid-1970s, the excessive rotation speeds of stars in galaxies had already led specialists to
propose the existence of dark matter, ensuring their cohesion. In 2011, the discovery that the
cosmic expansion was accelerating was attributed to a new, unknown ingredient known as dark
energy. Over the decades, all attempts to assign an identity to these new components ended in
failure.

In 2017 [8], Hélène Courtois, Daniel Pomarède, Brent Tully and Yeudi Hoffman produced
the first very-large-scale mapping of the universe, in a cube of one and a half billion light-years
across, with the Milky Way, our observation point, at the center (see figure 9 on the following
page). By subtracting the radial component of the velocity linked to the expansion motion, they
indicate the trajectories followed by the masses. A dipolar structure appears. One formation, the
Shapley attractor, comprising hundreds of thousands of galaxies, attracts galaxies to itself. But,
symmetrically to this formation, 600 million light-years from the Milky Way, there is an immense
void, some one hundred million light-years across, which, on the contrary, repels galaxies, and
to which we give the name of dipole repeller. To date, no theory has been able to explain the
existence of this vast void. While the idea of a gap in dark matter, positive and attractive, has
been evoked, it doesn’t hold water, as no mechanism has been found to give rise to it. Since
2017, several other such voids have been detected and located.

The launch of the James Webb Space Telescope has only added to this crisis [6]. The
Standard Model ΛCDM proposes a hierarchical mechanism for the birth of stars and galaxies.
Gravitational instability appears as soon as matter and radiation are decoupled. The scenarios
for the formation of both stars and galaxies in this model make use of the attributes conferred
on hypothetical dark matter. But even with these parameters, it’s impossible to imagine galaxies

20
Figure 9: Location of the dipole repeller (highlighted by the red circle) within the large-scale
structure of the universe [8]. The dipole repeller is a hypothesized region of space where galaxies
are pushed away from, counteracting the attractive force of the Shapley Supercluster.

forming before a billion years. The Hubble Space Telescope was already able to obtain images
in the near infrared. Early images of distant objects appeared to show groups of mini-galaxies.
But the James Webb Space Telescope showed that these objects were nothing other than HII re-
gions belonging to barred spiral galaxies, fully formed, hosting old stars, only 500 million years old.

For decades, the Standard Model ΛCDM has relied on its ability to account for CMB
fluctuations as gravito-acoustic oscillations, by adapting the numerous parameters relating to
dark matter, dark energy and, in particular, the value of the Hubble constant. This desire to
match observational data has resulted in a Hubble constant value of 67 km s−1 Mpc−1 . This is
significantly lower than the value of 70 km s−1 Mpc−1 deduced from direct observation of standard
candles.

All these factors are creating a deep crisis within the specialist community, and some voices
are beginning to be heard, suggesting the need to consider a profound paradigm shift. This is
what the Janus cosmological model20 proposes.

Since we are unable to introduce negative masses into the general relativity model, let’s
consider a profound change of geometric paradigm, already evoked in the previous sections
under the aspect of group theory and topology. The motion of positive masses, immersed in
the gravitational field, takes place according to geodesics that we consider to be derived from
a first metric gµν . We will therefore describe the motion of negative masses using a second set
of geodesics, derived from a second metric g µν . We thus have a manifold, whose points are
marked by the coordinates (x0 , x1 , x2 , x3 ), equipped with a pair of metrics (gµν , g µν ). We shall
neglect the action of electromagnetic fields and consider only the field of gravity. From the
metrics and we can construct Ricci tensors Rµν and Rµν and their associated Ricci scalars R and R.

As the Janus model proposes a paradigm shift by introducing a bimetric structure to account
for both positive and negative masses, we will now focus on the foundational mathematical
structure underlying this model. The Janus cosmological model builds upon the interaction
between two entities, i.e. positive and negative mass populations, each associated with its
20
see section 9 on the next page, where this model is developped.

21
respective metric. In the following section, we will explore the formulation of the action and
field equations governing this interaction, and how these coupled systems lead to a coherent
description of cosmic phenomena, offering an alternative to the limitations of the standard model
ΛCDM.

9 Foundation of the Janus cosmological model

To build this model, let us now consider the interaction between two entities: ordinary matter
with positive mass interacting with negative mass through gravitational effects. This model
involving negative mass takes into account the influence of both dark matter and dark energy.

We can describe this system of two entities with respective metrics gµν and g µν . Let R and
R be the corresponding Ricci scalars. We then consider the following two-layer action21 :
Z Å ã» Z Å ã»
1 κ
A= R+S+S 4
|g| d x + R+S+S |g| d4 x. (80)
E 2χ E 2χ̄

The terms S and S will give the source terms related to the populations of the two entities,
while the terms S and S will generate the interaction tensors. χ and χ̄ are the Einstein
gravitational constants for each entity. g and g are the determinants of the metrics gµν and g µν .
For κ = ±1, we apply the principle of least action. The Lagrangian derivation of this action
gives us:

0 = δA,
Z Å ã»
1
= δ R+S+S |g| d4 x
E 2χ
Z Å ã»
κ
+ δ R+S+S |g| d4 x,
E 2 χ̄
Z ñ Ç p å
1 δR R δ |g|
= δ +p
E 2χ δg µν |g| δg µν (81)
p p ô
1 δ( |g|S) 1 δ( |g|S) µν
»
+p µν
+ p µν
δg |g| d4 x
|g| δg |g| δg
Z ñ Ç p å
κ δR R δ |g|
+ δ +p
E 2χ̄ δg µν |g| δg µν
p p ô
1 δ( |g|S) 1 δ( |g|S) µν
»
+p + δg |g| d4 x.
|g| δg µν |g| δg µν
p

For any variation δg µν and δg µν , we locally obtain:

Ç p å p p
1 δR R δ |g| 1 δ( |g|S) 1 δ( |g|S)
+p +p +p = 0, (82)
2χ δg µν |g| δg µν |g| δg µν |g| δg µν
Ç p å p p
κ δR R δ |g| 1 δ( |g|S) 1 δ( |g|S)
+p +p +p = 0. (83)
2χ̄ δg µν |g| δg µν |g| δg µν |g| δg µν
21
Integration over E using the element d4 x is a method for computing the total action in the bimetric spacetime,
reflecting the four-dimensional nature of this bimetric universe. This implies considering the entire spacetime as
the domain of integration, integrating the contributions from each point to the action. The term d4 x represents an
infinitesimal element of hypervolume of this bimetric spacetime, used to measure each segment during integration.
Thus, it is a multiple volume integral performed over the four dimensions of spacetime, accumulating contributions
to the total action from each four-dimensional volume segment corresponding to each metric.

22
Let us then introduce the following tensors:
p
2 δ( |g|S) δS
Tµν = − p µν
= −2 µν + gµν S, (84)
|g| δg δg
p
2 δ( |g|S) δS
T µν = − p µν = −2 µν + g µν S, (85)
|g| δg δg
p
2 δ( |g|S)
Tµν = − p , (86)
|g| δg µν
p
2 δ( |g|S)
T µν = − p . (87)
|g| δg µν

We obtain then from equations (86) and (87):

p p
|g| |g| −2 δ( |g|S) −2 δ( |g|S) δS
Tµν = p µν
=p µν
= −2 µν + gµν S, (88)
|g| |g| |g| δg |g| δg δg
p p
|g| |g| −2 δ( |g|S) −2 δ( |g|S) δS
T µν = =p = −2 µν + g µν S. (89)
|g| |g| δg µν µν
p
|g| |g| δg δg
Introduced into equations (82) and (83), we can thus deduce the coupled field equations describing
the system of the two entities: To obtain the desired interaction laws under the Newtonian
approximation, we must choose κ = −1. The system of equations then becomes:
Ç å
1 |g|
Rµν − gµν R = χ Tµν + Tµν , (90)
2 |g|
Ç å
1 |g|
Rµν − g µν R = κχ̄ T µν + T µν . (91)
2 |g|
To obtain the desired interaction laws under the Newtonian approximation, we must choose
κ = −1. The system of equations then becomes:
Ç å
1 |g|
Rµν − gµν R = χ Tµν + Tµν , (92)
2 |g|
Ç å
1 |g|
Rµν − g µν R = −χ̄ T µν + T µν . (93)
2 |g|
The tensor Tµν is the energy-momentum tensor, which representsqthe source of the field acting
on positive mass entities and positive-energy photons. The term |g|
|g| is the source of this field
attributed to the action of negative masses on these positive masses. The tensor T µν is the
energy-momentum tensor, which represents theqsource of the field acting on negative mass entities
|g|
and negative-energy photons, and the term |g| is the source of this field attributed to the
action of positive masses on these negative masses. Tµν and T µν are the interaction tensors
of the system of the two entities corresponding to the "induced geometry", meaning how each
matter distribution on one layer of the universe contributes to the geometry of the other22 .

General relativity produces only a limited number of exact solutions. We will follow the same
logic.

22
Interaction between populations of positive and negative masses.

23
 Having established the foundation of the Janus cosmological model, with its bimetric structure
and the corresponding field equations, we turn to constructing explicit solutions under the
assumption of homogeneity and isotropy. By considering the FLRW form for both metrics, we
aim to derive a time-dependent solution that accounts for the interaction between positive and
negative mass populations. This section will focus on obtaining these solutions, exploring their
compatibility with observational data, and providing a theoretical framework for the accelerated
cosmic expansion.

10 Construction of a time-dependent, homogeneous and isotropic

solution
Given the symmetry assumptions, the metrics then have the FLRW form. The variable x0 is the
common chronological coordinate (time marker).

du2
ï ò
02
gµν = dx − a 2 2 2 2 2 2
+ u dθ + u sin θ dφ , (94a)
1 − ku2
du2
ï ò
02
g µν = dx − a 2 2 2 2 2 2
+ u dθ + u sin θ dφ . (94b)
1 − ku2
The determinants of the two metrics are
g = −a6 sin2 θ, g = −a6 sin2 θ. (95)
As shown in reference [16] the treatment of the two equations leads to the compatibility relation:
ρc2 a3 + ρc2 a3 = E = cst. (96)
This translates into conservation of energy, extended to both populations. The exact solution,
referring to two dust universes, corresponds to:
k = k = −1 (97)
and:

d2 a 4πG
a2 =−E, (98a)
dx0 2 c2
d2 a 4πG
a2 2 = + 2 E. (98b)
dx0 c
A theoretical model loses interest if it cannot be compared with observational data. The
evolution of the positive species will correspond to an acceleration if the energy E of the system
is negative. This provides a physical interpretation of the acceleration of the cosmic expansion
([12, 17]), which then follows from the fact that the energy content is predominantly negative.
Numerical data have been successfully compared with observational data [4]. The corresponding
curve is shown in figure 10 on the following page.

To complete the model, we now need to provide exact stationary solutions. We will restrict
ourselves to so(3) symmetry.

We now focus our attention to the interaction laws and their observational consequences.
These interaction laws, derived from the coupled field equations of the Janus model, govern
how positive and negative mass entities influence each other. The next section explores these
laws in detail and examines how they provide explanations for various cosmological phenomena,
including the formation of large-scale structures and the resolution of issues related to dark
matter and dark energy.

24
Figure 10: Comparison of observed and theoretical magnitudes as a function of z redshift [4].

11 Interaction laws and observational consequences

In the system of coupled field equations (92) and (93), the terms on the left-hand side involve
the Ricci tensors Rµν and Rµν and the corresponding Ricci scalars R and R. These terms are
calculated from the two metrics gµν and g µν . Using these two metrics, we then calculate the
form of two operators known as covariant derivatives ∇µ and ∇µ . It turns out that, due to their
form, the two left-hand sides of both equations identically satisfy the following relation:
Å ã
1
∇µ Rµν − Rgµν = 0, (99)
2
Å ã
1
∇µ Rµν − Rg µν = 0. (100)
2
The corresponding covariant derivatives of the two second members must therefore also be zero,
which corresponds to the Bianchi identities, implying:

∇µ Tµν = 0, (101)
∇µ T µν = 0. (102)

We should also have:

ñ ô
g
∇µ Tµν = 0, (103)
g
ï… ò
g
∇µ T µν = 0. (104)
g
In stationary conditions, the square roots of the ratios of the determinants behave like
constants, reflecting an “apparent mass effect”. Conditions (103) and (104) can therefore be
replaced by:

25
 ∇µ Tµν = 0, (105)
and

∇µ T µν = 0. (106)
Let’s write the system of equations in mixed notation, replacing the square roots, which have
2
become constant, by the positive constants b2 and b :
1
Rµν − Rgµν = χ Tµν + b2 Tµν , (107a)
 
2
ν 1 ν î ν 2 ν
ó
Rµ − Rg µ = −χ T µ + b T µ . (107b)
2
Using the Newtonian approximation, in both populations the non-zero tensor terms reduce to:
0 0
T00 = ρc2 > 0 T00 = ρc2 < 0 T 0 = ρc2 < 0 T 0 = ρc2 > 0. (108)

In our system of coupled field equations, the presence of a minus sign in front of the second
member of the second equation gives the following interaction laws:

• Masses of the same sign attract each other;

• Masses of opposite signs repel each other.

We have thus eliminated the runaway effect.

The first conclusion to be drawn is that where one of the two types of mass is present, the
other is absent, as immediately confirmed by simulations [14]. This is the case in the vicinity of
the Sun, and under these conditions the first equation is identified with Einstein’s 1915 equation.
The model is therefore in line with all the classical local observational data of general relativity:
Mercury’s perihelion advance, deflection of light rays by the Sun. The model therefore does not
invalidate that of general relativity, but presents itself as its extension, made essential to integrate
the new observational data, which can no longer be managed by introducing the hypothetical
components of dark matter and dark energy.

We have seen, in our construction of the unsteady solution, that negative energy dominates.
The model is thus profoundly asymmetrical. The negative mass component is proposed as a
substitute for the combined roles traditionally attributed to dark matter and dark energy. By
the way, going back to the original idea, inspired by the work of Andreï Sakharov, this allows us
to attribute a well-defined identity to these components. They are invisible, insofar as negative
masses emit photons of negative energy that our optical instruments cannot capture. They are
therefore simply copies of our own antimatter, assigned a negative mass. We then have a new
distribution of contents (see figure 11 on the next page).

At the moment of decoupling, when the gravitational instability can play its role (we must
then speak of joint gravitational instabilities), the characteristic Jeans time is shorter for negative
masses:
1 1
tJ = p ≪ tJ = √ . (109)
4πG|ρ| 4πGρ
The result will be a regular distribution of negative-mass conglomerates of spheroidal antihy-
drogen and negative-mass antihelium. These will behave like immense negative-mass protostars.

26
 Figure 11: Comparative contents of the ΛCDM and Janus models.

Figure 12: Early rapid star and galaxy formation.

As soon as their temperature causes hydrogen reionization, their contraction will cease. These
formations will then radiate in the red and infrared wavelengths. But their cooling time is
then large compared to the age of the universe, which means that these objects will no longer
evolve. The history of this universe fold associated with negative masses is totally different from
our universe fold of ordinary matter. It will not give rise to stars, galaxies or planets. It will
contain no atoms heavier than negative-mass antihelium. And there will be no life. And, as we’ll
see later: these negative formations are deliberately situated within the Newtonian approximation.

But there’s another very important point. When these spheroidal conglomerates form, they
confine the positive mass to the residual space, giving it a lacunar structure, comparable to
joined soap bubbles. The negative mass is thus distributed in the form of thin plates, sandwiched
between two negative conglomerates that exert a strong back pressure on it. The positive mass
is thus violently compressed and heated. However, due to its plate-like arrangement, it can cool
down very quickly through the emission of radiation (see figure 12).

The result is a pattern of first-generation star and galaxy formation totally different from
the standard one. This configuration had been the subject of simulations [14] since the first,
heuristic, approach to the model, and the fact that objects all form within the first hundred
million years was one of its predictions, largely confirmed by JWST data.

The lacunar structure, advocated as early as 1995 [14], predicted the existence of large voids,
which the discoveries of the dipole repeller and other similar large voids have also confirmed.
Once this lacunar structure has been formed, matter tends to concentrate along the segments
common to three gaps, forming filaments (see figure 13 on the next page). The nodes of this
distribution will only develop into galaxy clusters.

27
 Figure 13: Structure of positive mass in contiguous bubbles.

After establishing the interaction laws and exploring their observational consequences, it’s
essential to verify the mathematical and physical consistency of the Janus model. This requires
demonstrating that the system of coupled field equations respects the Bianchi identities and
provides consistent solutions in the weak field limit. In the following section, we will examine the
conditions necessary to ensure this consistency, particularly in regions dominated by ordinary
matter, such as near the Sun, as well as in regions dominated by negative masses, such as near
the dipole repeller.

12 The mathematical and physical consistency of the model

This is ensured in an isotropic, homogeneous and unsteady situation, the required condition
being the generalized conservation of energy expressed by equation (96). We now turn to the
case of stationary solutions, limiting ourselves to those that satisfy so(3) symmetry. Bianchi
identities must then be satisfied, i.e. relations (101), (102), (105) and (106).

First, we’ll show the existence of asymptotic consistency in Newtonian approximation situa-
tions. The key aspects of this approximation are as follows:

• Velocities must be negligible compared to the speed of light. This is the case for velocities
⟨v⟩ and ⟨v⟩ of thermal agitation in both media, which are involved in the definition of
pressures and in both media. After decoupling:

ερ⟨v⟩ ερ⟨v⟩
εp = and εp = . (110)
3 3
• Curvature effects must be neglected, meaning that the radial coordinate must be much
larger than the characteristic length scale associated with curvature, i.e., the Schwarzschild
radius.

12.a Newtonian approximation of the field generated by a positive mass M

Let’s introduce the Schwarzschild radius RS as follows:
2GM
εRS = ε , (111)
c2
where ε being a small parameter. so(3) symmetry imposes the shapes of the two metrics:

28
 2
ds2 = eν dx0 − eλ dr2 − r2 dθ2 − r2 sin2 θ dφ2 , (112a)
2
ds2 = eν dx0 − eλ dr2 − r2 dθ2 − r2 sin2 θ dφ2 . (112b)

The construction of a stationary solution then requires to calculate the functions:

ν(r), λ(r), ν(r), and λ(r). (113)

To locate this solution, we need to consider the shapes of the field source tensors:
ν ν
Tµν , Tµν , T µ, and T µ . (114)
Let’s start by considering a situation where only positive mass is present. The tensors Tµν
ν
and T µ are then null and the two field equations 92 and 93 become in mixed-mode form:

1
Rµν − gµν R = χTµν , (115)
2
ν 1 |g| ν
Rµ − g νµ R = −χ̄ T . (116)
2 |g| µ

The form of the tensor Tµν in its classical mixed-mode form is given by23 :
Ü 2 ê
ρc 0 0 0
0 −εp 0 0
Tµν = . (117)
0 0 −εp 0
0 0 0 −εp
As we are in the Newtonian approximation, ε is very small. With the introduction of the metric
(112a) and the tensor (117) in the first field equation, we are led to introduce the function m(r)
such that:

2m(r) Ä ä
e−λ = 1 − =⇒ 2m(r) = r 1 − e−λ . (118)
r
Similarly to equation (14.18) from [1], the classic calculation leads to the relationship:
Gρ r
Z
4 G
m(r) = 2 4πr2 dr = πr3 ρ 2 . (119)
c 0 3 c
We then obtain the classical Tolman–Oppenheimer–Volkoff (TOV) equation ([11]). Relation
(119) places the small quantity in front of any quantity that will be neglected in the Newtonian
approximation:
3
1 dp m + 4πεGpr
c4
 p
= − ρ + ε . (120)
c2 dr r(r − 2mε) c2
When ε tends to zero (or c tends to infinity) we get:

dp ρmc2 Gρ 4πr3 ρ
=− 2 =− 2 . (121)
dr r r 3
3
The quantity 4πr3 ρ represents the amount of matter µ(r) contained inside a sphere of radius r.
We know that the force of gravity exerted inside a mass of constant density is equivalent to that
exerted by the mass located at the center of the sphere, and that the mass located outside this
sphere gives a force of zero. So the quantity − Gρµ(r)
r2
is the force of gravity, per unit volume,
23
(13.1) p.425 of [1].

29
acting on the matter contained in an elementary volume around a point at distance r from the
center. Thus the relation (117), which follows from the Newtonian approximation, expresses that
the force of gravity balances the force of pressure. This is the classic Euler relationship.

Hence, the Schwarzschild interior metric built is given by:

" ã#2
2 2 dr2
Å ã Å
3 r n 1 r 2
ds2 = dx0 − 2
dθ2 + sin2 θdϕ2 . (122)


1− 2 − 1− 2 r 2 − r
2 r̂ 2 r̂ 1 − r̂2

This metric connects with the Schwarzschild exterior metric :

dr2
Å ã
2GM 2 02
ds2 = 1 − 2 − r2 dθ2 + sin2 θdϕ2 , (123)


c dx − 2GM
c r 1 − c2 r
where rn is the radius of the star and r̂ is a stellar constant as a function of its density ρ. It’s the
characteristic radius of a neutron star, defined under the assumption of constant density ρ. It
establishes a critical threshold for the star’s radius, beyond which the internal pressure becomes
infinite at the center, indicating a physical singularity or instability. This radius is derived from
the balance between gravitational forces and the internal pressure gradients within the star ([10]).
It’s given by:
3c2
r̂ = . (124)
8πGρ
We can thus deduce, according to the classical theory of general relativity, that a particle of
ordinary matter will undergo an attractive gravitational field due to the effect of a distribution
of positive masses.

To ensure the mathematical consistency of the system of two field equations 115 and 116, we
ν
therefore need to consider a form of the tensor T µ that gives back this same Euler relation when
the Newtonian approximation is also applied to this solution. This is guaranteed with the form
ν
of the interaction tensor T µ of the field equation 116 as this choice can stem from a Lagrangian
derivation: Ü 2 ê
ρc 0 0 0
ν 0 +εp 0 0
Tµ= . (125)
0 0 +εp 0
0 0 0 +εp
On the right-hand side of the second field equation (116), the ratio of determinants will be con-
sidered almost unity insofar as we perform this calculation within the Newtonian approximation24 .

Then, if we consider that:

|g| eν eλ r4 sin2 θ
= ≈ 1, (126)
|g| eν̄ eλ̄ r4 sin2 θ
the calculation leads to the Tolman–Oppenheimer–Volkoff (TOV) solution for the population of
negative masses managed by the second field equation:
3
1 dp m − 4πεGpr
c4
 p
= − ρ − ε . (127)
c2 dr r(r + 2mε) c2
The two solutions, (120) and (127), asymptotically approach the Euler equation in the Newtonian
approximation as ε tends to zero. This also corresponds to the asymptotic satisfaction of the
24
In stationary conditions, the square roots of the ratios of the determinants behave like constants, reflecting an
“apparent mass effect”.

30
Bianchi identities in the same context25 .

Consequently, it is possible to build the Schwarzschild interior metric associated with the
population of negative masses by applying the same calculation scheme as for the population of
positive masses, thus constituting the solution to the second field equation 116 as follows:
" ã#2
2 2 dr2
Å ã Å
2
¯ = 3 r n 1 r 2
dx0 − 2 2 2 2
(128)


ds 1+ 2 − 1+ 2 2 − r dθ + sin θdϕ .
2 r̂ 2 r̂ 1 + r̂r2

This metric must join the Schwarzschild exterior metric:

dr2
Å ã
¯ 2 2GM 2 02
− r2 dθ2 + sin2 θdϕ2 . (129)


ds = 1 + 2 c dx − 2GM
c r 1 + c2 r

We can deduce that a particle with negative mass will undergo a repulsive gravitational field due
to the effect of a distribution of positive masses.

The Janus model presents a new paradigm, extending general relativity by describing the
universe as a four-dimensional manifold M4 , endowed with two distinct metrics. These metrics
are solutions to the system of coupled field equations (92) and (93).

Let’s now consider the case, still in the Newtonian approximation, where the geometry is
determined by the presence of negative mass, corresponding to regions of space dominated by
negative masses, such as near the dipole repeller ([8]).

12.b Newtonian approximation of the field generated by a negative mass M

In regions where negative masses dominate, such as near the dipole repeller ([8]), the system
becomes in mixed-mode form:

1 |g| ν
Rµν − gµν R = χ T , (130)
2 |g| µ
ν 1 ν
Rµ − g νµ R = −χ̄T µ . (131)
2
If we consider the impact of the presence of negative masses on the geometry of spacetime
structured by the metric tensor of the first field equation 130 associated with the population of
positive masses, we can define the corresponding interaction tensor 132 as follows:
Ü 2 ê
ρ̄c̄ 0 0 0
0 −p̄ 0 0
Tµν = . (132)
0 0 −p̄ 0
0 0 0 −p̄

Thus, the impact of the pressure gradient of negative masses on the geodesics followed by
ordinary matter and positive-energy photons according to the field equation 130 translates into
the following Tolman–Oppenheimer–Volkoff equation:
3
p̄′ m − 4πGp̄r
c̄4
 p̄ 
= − ρ̄ − . (133)
c̄2 r(r + 2m) c̄2
25
The inequality r ≫ 2m (where m is often replaced by GM c2
to obtain a dimension of length, M being the mass
of the object and G the gravitational constant) indicates that we are sufficiently far from the gravitational source
for the effects of general relativity to be negligible. Indeed, at great distances, the length 2GMc2
is completely
negligible.

31
 Therefore, it is possible to build the Schwarzschild interior metric solution in this manner:
" ã#2
rn 2 r2 dr2
Å ã Å
3 1 2
ds2 = dx0 − 2
dθ2 + sin2 θdϕ2 . (134)


1+ 2 − 1+ 2 r 2 − r
2 r̂ 2 r̂ 1 + r̂2

This metric can be connected to the Schwarzschild exterior metric:

dr2
Å ã
2GM 2 02
ds2 = 1 + 2 − r2 dθ2 + sin2 θdϕ2 . (135)


c dx − 2GM
c r 1 + c2 r

We can deduce that a particle of ordinary matter will undergo a repulsive gravitational field due
to the effect of a distribution of negative masses.

Then, when the source of the gravitational field of the second field equation (131) is created
by a negative mass, we can freely define the following energy-momentum tensor as follows:
Ü 2 ê
ρ̄c̄ 0 0 0
ν 0 p̄ 0 0
Tµ = . (136)
0 0 p̄ 0
0 0 0 p̄

We can therefore deduce the following Tolman–Oppenheimer–Volkoff equation:

3
p̄′ m + 4πGp̄r
c̄4
 p̄ 
= − ρ̄ + . (137)
c̄2 r(r − 2m) c̄2

Hence, the interior Schwarzschild metric can be constructed as follows:

" ã#2
r¯n 2 r2 dr2
Å ã Å
¯ = 3
2 1 2
dx0 − 2 2 2 2
(138)


ds 1− 2 − 1− 2 2 − r dθ + sin θdϕ .
2 r̂ 2 r̂ 1 − r̂r2

This metric matches the exterior Schwarzschild metric:

Ç å
2
¯ = 1− 2GM 02 dr2
2
− r2 dθ2 + sin2 θdϕ2 . (139)


ds 2
c̄ dx −
c̄ r 1 − 2GM
c̄2 r

We can deduce that a particle of negative mass will undergo an attractive gravitational field due
to the effect of a distribution of negative masses.

Both solutions (133) and (137) reduces to the Euler equation approximately equal to
− GM (r)ρ̄(r)
r2
in the Newtonian limit, reflecting hydrostatic equilibrium26 .

The form of these two source tensors satisfies the Bianchi identities. This would obviously
not be the case if the negative mass were to fall outside of this framework. For that, there would
need to exist neutron stars of negative mass. However, the characteristic time of evolution of
conglomerates of negative mass, their "cooling time", exceeds the age of the universe. These
spheroidal conglomerates cannot evolve, so the content of this negative spacetime will be limited
to a mixture of negative mass anti-hydrogen and anti-helium. Since nucleosynthesis cannot occur,
there can be no anti-galaxies or anti-stars, regardless of their mass ([25]). Consequently, there
cannot exist anti-neutron stars.

26
Where the pressure at the center of this negative mass spheroid is balanced by the negative gravitational
force depending on density and mass.

32
 Figure 14: Attenuation of the magnitude of objects in the background of the dipole repeller.

Moreover, in the case where this negative spacetime would generate hyperdense stars through
an as-yet-unknown mechanism, it would then be necessary to reconsider the form of these tensors.
However, the current configuration satisfies all currently available and potentially available
observational data.

After verifying the mathematical and physical consistency of the Janus model, we now turn
to its predictive capabilities. One of the most striking predictions concerns the existence of
large voids and structures such as the dipole repeller. The Janus model not only accounts for
these features but also offers novel predictions regarding the effect of negative gravitational
lensing on the magnitudes of background sources. In the following short section, we explore the
observational signatures of this phenomenon, with a particular focus on the implications for the
dipole repeller.

13 Dipole repeller prediction

The Janus model is essentially falsifiable in Popper’s sense. It predicted a large-scale twin
structure with large voids. This has been confirmed ([8]). It predicted a very early birth of
first-generation stars and galaxies. A new prediction this time concerns the magnitude of sources
located in the background of the large void. According to the model, the magnitude of the light
emitted by these distant sources will be attenuated by the negative gravitational lensing effect.
This is a novel aspect, since it has been assumed that the two entities, positive and negative,
interact only through antigravitation. Photons from these distant sources can then freely pass
through the negative-mass conglomerates. This means that both external and internal geodesics
must be used. The deflection effect of light rays will be greatest when they graze the surface of
the object, with radius. This effect weakens as you move deeper into the object, becoming zero
when the photons pass through its center (see figure 8). Eventually, we will be able to map the
magnitudes of objects in the background of the dipole repeller. Schematically, their luminosity
will be attenuated in a ring-shaped pattern (see figure 14).

33
 This measurement will immediately give us the value of the radius r0 of this formation.

After exploring the implications of the Janus model in the Newtonian approximation and
its predictions for large-scale structures, such as the dipole repeller, we now move beyond these
limitations. In a universe dominated by positive masses, certain astrophysical objects, such as
neutron stars and supermassive black holes, exhibit strong gravitational effects that require a
relativistic treatment. The following section addresses the challenges of extending the model to
these extreme cases.

14 Beyond the Newtonian approximation

These objects are absent in the universe fold associated with negative masses. In our universe fold
of ordinary matter, objects that deviate from the Newtonian approximation are neutron stars and
hypermassive objects located at the center of galaxies, which early images show to be the seat
of a strong gravitational redshift effect, darkening their central part. These objects are a priori
manageable using the classic pair of outer and inner metrics, taking rotation into account. It
should be remembered that we are under no obligation to provide the form of the source tensor of
the other sector, in this case an interaction tensor, whose form would be precisely imposed by the
Bianchi identities. It’s conceivable that one day someone will provide the exact form of this tensor.

But even in the absence of such an object, there is no a priori inconsistency.

15 Conclusion
The genesis of the Janus model spanned several decades. The starting point, in 1967, was Andreï
Sakharov’s attempt to provide an initial explanation for the absence of observations of primordial
antimatter, which remains a significant flaw in the Standard Model ΛCDM. This model offers no
explanation for the loss of half of the universe’s content. Sakharov therefore proposed a universe
structure with two sectors, the second being T-symmetrical to our own. A few years later,
in 1970, through the application of symplectic geometry, mathematician Jean-Marie Souriau
demonstrated that this inversion of the time coordinate, i.e., T-symmetry, is synonymous with the
inversion of energy and mass. Pushing this idea of global symmetry further, Sakharov envisioned
a twin universe that is CPT-symmetrical to ours. In this scenario, the invisible components of
the universe reduce to negative-mass antimatter.

In 1994, we proposed that this universe structure corresponds to a two-fold cover of a projec-
tive P4 , by a compact universe with the topology of a S4 sphere. The two singularities of this
spherical universe, the Big Bang and the Big Crunch, then coincide. By introducing a tubular
structure, these singularities disappear. This configuration consists of two PT-symmetrical folds.
These adjacent sectors are assumed to interact solely through gravity. Therefore, the interaction
between positive masses in one sector and negative masses in the other sector must be taken into
account.

However, the introduction of negative masses is not feasible within the framework of general
relativity, as it would result in interaction laws that are incompatible with known physical princi-
ples. Thus, a bimetric model is proposed. A system of coupled field equations is then constructed
from an action, whose form eliminates the problematic runaway effect. The interaction laws in
the model dictate that masses of the same sign attract each other according to Newton’s law,
while masses of opposite signs repel each other following an anti-Newtonian law. Since these
masses are mutually exclusive, the negative mass can be neglected in the vicinity of the Sun, and

34
the first field equation then aligns with Einstein’s equation.

In this way, the model remains consistent with local relativistic observations, such as the
advance of Mercury’s perihelion and the deflection of light by the Sun. Therefore, the Janus
model can be considered an extension of general relativity. An exact, time-dependent solution is
constructed, revealing a generalized energy conservation law that applies to both sectors. When
adapting the model to observations, it becomes evident that an accelerating expansion is present,
imposing a fundamental dissymmetry between the two entities involved.

In this framework, the vast majority of negative mass replaces the hypothetical components
of dark matter and dark energy. As a result, the matter distribution is approximately 5% visible
matter and 95% negative mass, which is invisible because it emits photons of negative energy
that elude detection by our observation instruments. This dissymmetry implies that, following
decoupling, the negative masses form a regular network of spheroidal conglomerates, while the
positive mass, confined to the remaining space, adopts a patchy distribution.

The model also accounts for the existence of large voids, with the dipole repeller being
the first identified among them. At the centers of these large voids are invisible spheroidal
conglomerates that behave like giant protostars, with cooling times exceeding the age of the
universe. These objects, which emit negative-energy photons corresponding to light in the red
and infrared regions, do not evolve and do not give rise to stars, galaxies, or atoms heavier
than helium. Life, therefore, is absent from this negative sector, which consists of a mixture of
negative-mass antihydrogen and antihelium.

Furthermore, the model explains the very early formation of first-generation stars and galaxies,
as recently demonstrated by the James Webb Space Telescope. We then examine the issue of the
model’s mathematical consistency, specifically whether the Bianchi identities are satisfied. We
show that they can be asymptotically satisfied under conditions corresponding to the Newtonian
approximation.

Lastly, we address the question of objects that do not fit within this approximation, primarily
located on the positive-mass side. We assert that we are not required to provide the exact form
of the interaction tensor in such cases, as it is determined by the zero-divergence condition. The
lack of definition of this tensor does not invalidate the consistency of a non-linear solution.

35
References
[1] R. Adler, M. Bazin, and M. Schiffer. Introduction to General Relativity. McGraw-Hill, 1975.
[2] Hermann Bondi. “Negative Mass in General Relativity”. In: Reviews of Modern Physics
29.3 (July 1957), pp. 423–428. doi: 10.1103/RevModPhys.29.423.
[3] Werner Boy. “Über die Curvatura integra und die Topologie geschlossener Flächen”. In:
Mathematische Annalen 57.2 (1903), pp. 151–184. doi: 10.1007/BF01444342.
[4] Gilles D’Agostini and Jean-Pierre Petit. “Constraints on Janus Cosmological model from
recent observations of supernovae type Ia”. In: Astrophysics and Space Science 363.7 (2018),
p. 139. doi: 10.1007/s10509-018-3365-3.
[5] P.A.M. Dirac. “The quantum theory of the electron”. In: Proceedings of the Royal Society
of London. Series A, Containing Papers of a Mathematical and Physical Character 117.778
(1928), pp. 610–624.
[6] Leonardo Ferreira et al. “Panic! at the Disks: First Rest-frame Optical Observations of
Galaxy Structure at z > 3 with JWST in the SMACS 0723 Field”. In: The Astrophysical
Journal Letters 938.1 (Oct. 2022), p. L2. doi: 10.3847/2041-8213/ac947c.
[7] Richard P. Feynman. “The Theory of Positrons”. In: Physical Review 76 (1949), pp. 749–759.
url: [Link]
[8] Yehuda Hoffman, Daniel Pomarède, R. Brent Tully, and Hélène M. Courtois. “The dipole
repeller”. In: Nature Astronomy 1.2 (Jan. 2017), p. 0036. doi: 10.1038/s41550-016-0036.
[9] Ryszard Kerner. “Generalization of the Kaluza-Klein theory for an arbitrary non-abelian
gauge group”. In: Annales de l’institut Henri Poincaré. Section A, Physique Théorique 9.2
(1968), pp. 143–152. url: [Link]
[10] Pascal Koiran et al. “PT -symmetry in one-way wormholes”. In: Annals of Physics (2024),
p. 169765. issn: 0003-4916. doi: [Link] arXiv:
2409.02951 [gr-qc].
[11] J.R. Oppenheimer and H. Snyder. “On Continued Gravitational Contraction”. In: Phys.
Rev. 56.5 (1939), pp. 455–459.
[12] S. Perlmutter et al. “Measurements of Ω and Λ from 42 High-Redshift Supernovae”. In:
The Astrophysical Journal 517.2 (June 1999), pp. 565–586. doi: 10.1086/307221.
[13] Jean-Pierre Petit. “The missing-mass problem”. In: Nuovo Cimento B Serie 109.7 (July
1994), pp. 697–709. doi: 10.1007/BF02722527.
[14] Jean-Pierre Petit. “Twin universes cosmology”. In: Astrophysics and Space Science 226.2
(1995), pp. 273–307. doi: 10.1007/BF00627375.
[15] Jean-Pierre Petit. Topo The World (The Adventures of Archibald Higgins).
[16] Jean-Pierre Petit and Gilles D’Agostini. “Negative mass hypothesis in cosmology and the
nature of dark energy”. In: Astrophysics and Space Science 354.2 (2014), pp. 611–615. doi:
10.1007/s10509-014-2106-5.
[17] Adam G. Riess et al. “Observational Evidence from Supernovae for an Accelerating Universe
and a Cosmological Constant”. In: The Astronomical Journal 116.3 (Sept. 1998), pp. 1009–
1038. doi: 10.1086/300499.
[18] Andrei D. Sakharov. “Violation of CP Invariance, C Asymmetry, and Baryon Asymmetry
of the Universe”. In: ZhETF Pisma Redaktsiiu 5 (Jan. 1967), p. 32.
[19] Andrei D. Sakharov. “Baryon asymmetry of the universe.” In: Zhurnal Eksperimentalnoi i
Teoreticheskoi Fiziki 76 (Apr. 1979), pp. 1172–1181.

36
[20] Andrei D. Sakharov. “Cosmological models of the universe with reversal of time arrow”. In:
Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki 79 (Sept. 1980), pp. 689–693.
[21] Géry de Saxcé. Which symmetry group for elementary particles with an electric charge
today and in the past? Mar. 2024. arXiv: 2403.14846 [math-ph].
[22] Géry de Saxcé and Charles-Michel Marle. Presentation of Jean-Marie Souriau’s book
“Structure des systèmes dynamiques”. June 2023. arXiv: 2306.03106 [[Link]].
[23] Jean-Marie Souriau. Géométrie et Relativité. Paris: Hermann, 1964.
[24] Jean-Marie Souriau. Structure des systèmes dynamiques: maîtrises de mathématiques.
Collection Dunod Université. Dunod, 1970. isbn: 9782040107529.
[25] Hasnae Zejli and Hicham Zejli. “Nature of the Dipole Repeller”. In: Journal of Natural
Sciences Research 2 (2023). Discusses the concept of negative mass and the role of the
dipole repeller in cosmology, pp. 2–14. url: [Link]
57604/view.

37