CP T -Symmetric Universe

Latham Boyle1 , Kieran Finn1,2 and Neil Turok1

1
Perimeter Institute for Theoretical Physics, Waterloo, Ontario, Canada, N2L 2Y5
2
School of Physics and Astronomy, University of Manchester, Manchester, UK M13 9PL
(Dated: March 2018)
We propose that the state of the universe does not spontaneously violate CP T . Instead, the
universe after the big bang is the CP T image of the universe before it, both classically and quantum
mechanically. The pre- and post-bang epochs comprise a universe/anti-universe pair, emerging from
nothing directly into a hot, radiation-dominated era. CP T symmetry selects a unique QFT vacuum
state on such a spacetime, providing a new interpretation of the cosmological baryon asymmetry, as
well as a remarkably economical explanation for the cosmological dark matter. Requiring only the
standard three-generation model of particle physics (with right-handed neutrinos), a Z2 symmetry
arXiv:1803.08928v3 [hep-ph] 2 Dec 2018

suffices to render one of the right-handed neutrinos stable. We calculate its abundance from first
principles: matching the observed dark matter density requires its mass to be 4.8×108 GeV. Several
other testable predictions follow: (i) the three light neutrinos are Majorana and allow neutrinoless
double β decay; (ii) the lightest neutrino is massless; and (iii) there are no primordial long-wavelength
gravitational waves. We mention connections to the strong CP problem and the arrow of time.

Introduction. Observations reveal that, seconds af- Spacetime (Background and Perturbations). In
ter the Big Bang, the universe was described by a this Letter we work at the level of QFT on curved space.
spatially-flat radiation-dominated FRW metric (plus tiny Before turning to the state of the QFT, in this section
gaussian, adiabatic, purely-growing-mode scalar pertur- we first consider what (C)P T says about the spacetime
bations described by a nearly-scale-invariant power spec- itself at a purely classical level [3]. Thus, we treat the
trum; and, so far, no primordial vector or tensor per- metric and the radiation fluid using general relativity:
turbations) [1]. This is a clue about the origin of the The line element may be written in standard ADM
universe, but what is it trying to tell us? The conven- form: ds2 = −N 2 dτ 2 + hij (dxi + N i dτ )(dxj + N j dτ ).
tional view is that, in order to explain these simple initial To describe our universe (a flat FRW spacetime plus
conditions, one should imagine that the radiation domi- small scalar, vector and tensor perturbations), we use
nated era we see was preceded by an earlier hypothetical “comoving gauge” so that the xi = const threads are
epoch of accelerated expansion called inflation. normal to the τ = constant slices (N i = 0), and the
threads follow the stress-energy flow so that (for scalar
In this Letter (and a longer companion paper [2]) we
perturbations) the momentum density T0i also vanishes.
take a different view. Ignoring perturbations for the mo-
Then we can write the lapse as N = a[1 + φ], and hij
ment, the metric we see in our past is strikingly sim- (0) (1) (2)
ple and analytic: gµν = a2 (τ )ηµν where ηµν is the flat as hij = a2 [(1 + 2R)δij + 2γij + 2γij + 2γij ], where
Minkowski metric, and the scale factor a(τ ) is just pro- a = a(τ ) is the background scale factor, R is the “co-
portional to the conformal time τ . If we take this metric moving curvature perturbation,” φ is another scalar per-
seriously, and follow a(τ ) ∝ τ across the bang, we find turbation related to R by the Einstein equation, and we
that the analytically extended FRW background with have split the traceless perturbation γij into its scalar,
(0) (1) (2)
−∞ < τ < ∞ suddenly exhibits a new isometry: time vector and tensor parts: γij , γij , γij [4, 5].
reversal symmetry τ → −τ . It thus becomes possible Next, to treat spinors and CP T , just as we switch from
to adopt the natural hypothesis that, contrary to naive the wave operator  to its “square root” (the Dirac op-
appearances, the state of our universe does not sponta- erator D / ), we switch from the line element ds2 to its
neously violate CP T . “square root” (the tetrad ea ). Thus, we write ds2 =
ηab ea eb , where ηab = diag{−1, 1, 1, 1} is the Minkowski
In this Letter we explore the hypothesis that CP T is
metric, and choose a local Lorentz gauge where the tetrad
unbroken, explain how it provides novel explanations for 1/2
one forms ea = eaµ dxµ are e0 = −N dτ and ei = hij dxj .
a number of the observed features of our Universe, and
The spacetime is (C)P T symmetric in the sense that the
point out some predictions that will be tested in the com-
tetrad geometry according to an observer who moves for-
ing years. In particular, we find it yields a remarkably
ward along the xi = const thread is identical to the
economical explanation of the cosmological dark mat-
tetrad geometry according to an observer who moves
ter: if the Universe is in its preferred CP T -symmetric
backward along the thread and reverses the spatial one
vacuum, late-time observers see heavy sterile neutrinos
forms ei → −ei . Equivalently, the tetrad at time τ after
emerging from the bang, for the same reason that distant
the bang is the reverse of the tetrad at the corresponding
observers see Hawking radiation emerging from a black
moment before the bang along the same thread:
hole. In our opinion, this provides the most elegant and
compelling dark matter model currently available. eaµ (τ, x) = −eaµ (−τ, x). (1)
 2

Let us unpack the implications of this simple constraint: [13, 14], it becomes natural to reinterpret the contract-
i) Background geometry: Eq. (1) implies that the scale ing half of our universe as an anti-universe (whose in-
factor is odd, a(−τ ) = −a(τ ), with a ∝ τ near the bang trinsic proper time runs counter to the natural time-like
(as in the radiation era). coordinate in the embedding superspace, i.e., the scale
If this picture is correct, so that the bang is a topolog- factor), so that our CP T -invariant universe is reinter-
ically enforced singularity, cosmological models in which preted as a universe/anti-universe pair (U Ū ), emerging
a(τ ) undergoes a nonsingular bounce at a minimum scale from nothing! This interpretation continues to be useful
factor amin > 0 are misguided. when spinors and anti-particles enter the story: e.g., the
ii) Scalar perturbations: In fourier space, neglecting matter/anti-matter asymmetry on one side of the bang is
anisotropic stress, R satisfies R′′ +2(z ′/z)R′ +c2s k 2 R = 0, the opposite of the asymmetry on the other side [2]. To
where c2s = δp/δρ is the sound speed, k is the comoving convert this suggestive, semi-classical picture into a fully
wave number, z 2 ≡ a2 ǫ and ǫ = (a′ /a)′ /(a′ /a)2 − 1. Near quantum one (as Feynman did with Stueckelberg’s idea)
the bang, where a ∝ τ and c2s = 1/3, the general solu- is an important task, beyond the scope of this Letter.
tion is R(k, τ ) = τ −1 [A(k) sin(cs kτ ) + B(k) cos(cs kτ )]. CP T Invariant Vacuum. Now we turn from the
The condition (1) then sets B(k) = 0, eliminating the spacetime to the state of the QFT living on it.
mode that is singular at the bang, and selecting the well- In Minkowski spacetime, there is a unique vacuum that
behaved mode that approaches a constant as τ → 0. respects the Minkowski isometries (more precisely, space-
This is precisely the boundary condition responsible time translations, Lorentz transformations, and CP T ).
for producing the famous oscillations seen in the CMB But in a generic curved spacetime, the choice of vacuum
power spectrum, with the correct phases. This observed is ambiguous: different observers will naturally define dif-
phenomenon, usually attributed to inflation, is alterna- ferent, inequivalent vacua, so that the zero particle state
tively explained by a symmetry argument. according to one observer will contain particles according
Also note that density pertubations grow as we get to a different observer [15]. In particular, in an ordinary
further from the bang in either direction, and hence the FRW spacetime, the isometries (spatial translations, spa-
thermodynamic arrow of time points away from the bang tial rotations, and parity) are not enough to determine
in both directions (to the future and past). The possibil- a preferred vacuum, and comoving observers at different
ity that the thermodynamic arrow of time might reverse epochs will disagree. But, as we explain in this section,
is an old one (going back to the debates between Boltz- if the FRW background also has an isometry under time
mann and his contemporaries [6]), and has been invoked reversal τ → −τ , then there is a preferred vacuum that
more recently in several interesting contexts [7–12]. respects the full isometry group (including CP T ).
iii) Vector perturbations: Neglecting anisotropic stress, Consider a spinor Ψ with mass m > 0 on a flat FRW
the gauge-invariant vector metric perturbation σg satis- background ds2 = a2 (τ )[−dτ 2 + dx2 ]. Its Lagrangian is
fies σg′ + 2(a′ /a)σg = 0 [4], so σg (k, τ ) = C(k)/τ 2 near √
(1) L = −g[iΨ̄eµa γ a ∇µ Ψ − mΨ̄Ψ] (2a)
the bang. In our chosen gauge, σg ∝ γij ′ , so Eq. (1) im-
plies that σg (and hence the primordial vorticity, which = iψ̄∂
/ ψ − µψ̄ψ. (2b)
is tied to σg by the 0i Einstein equation) vanishes, again
in agreement with observations. In the first line, we have the usual curved space Dirac
iv) Tensor perturbations: neglecting anisotropic stress, operator [15]; in comoving/conformal coordinates, the
(2) (2) (2) (2) tetrad is eµa = (1/a)δaµ , and γ a are the 4 × 4 Dirac gamma
γij satisfies γij ′′ + 2(a′ /a)γij ′ + k 2 γij = 0 so that
(2)
matrices. In the second line, we introduce the Weyl in-
γij (k, τ ) = τ −1 [Aij (k) sin(kτ ) + Bij (k) cos(kτ )]. Now variant spinor field ψ(τ, x) and its effective mass µ(τ ):
(1) sets Bij (k) = 0, eliminating the mode that is singular
at the bang and selecting the mode that is well behaved. ψ ≡ a3/2 Ψ, µ ≡ a m, (3)
Note that, for each type of perturbation – scalar, vec-
tor, and tensor – the condition (1) “protects” the geome- with ∂/ ≡ γ µ ∂µ the flat-space Dirac operator, and partial
try near the bang by precisely eliminating the dangerous derivatives ∂µ with respect to xµ = {τ, x}. The resulting
singular modes that would cause the breakdown of linear equation of motion is
perturbation theory and destroy the smooth (Weyl) char-
acter of the singularity. In this way of looking at it, the (i∂
/ − µ)ψ = 0. (4)
elimination of the singular modes is not a consequence of
a boundary condition at τ = 0, but is instead enforced Note that, since a(τ ) is an odd function of τ , so is µ(τ ).
by the symmetry between past and future. To quantize, we expand ψ(x) in a basis of solutions of
v) U Ū pair: Eq. (1) implies eaµ (0, x) = 0. If we Eq. (4), ψ(k, h, x) and ψ c (k, h, x):
combine this with Stueckelberg’s observation that an
anti-particle is a particle whose worldline proper time
XZ d3 k
[a(k, h)ψ(k, h, x)+b† (k, h)ψ c (k, h, x)]. (5)
runs counter to the time in the embedding spacetime h
(2π)3/2
 3

Here ψ(k, h, x) ∝ eikx is the solution with momentum satisfying η(k) = −η(−k). The vacuum defined by
k, helicity h, and “positive frequency”; ψ c (k, h, x) ≡ aη |0η i = bη |0η i = 0 is still invariant under the full isome-
−iγ 2 ψ ∗ (k, h, x) is the charge-conjugate (“negative fre- try group of the FRW background including CP T . How-
quency”) solution; and a(k, h) and b(k, h) are parti- ever, among this family of “η vacua” the vacuum |00 i
cle and anti-particle annihilation operators, which sat- is preferred since it minimizes the Hamiltonian in the
isfy the usual fermionic anti-commutation relations: asymptotic “+/−” regions (or the particle number ac-
{a(k, h), a† (k′ , h′ )} = {b(k, h), b† (k′ , h′ )} = δ(k − k′ )δh,h′ , cording to an early or late time observer) [2].
all other anti-commutators vanishing. Now we assume the universe is in the preferred CP T -
But in a general curved spacetime, there is no canoni- invariant vacuum state and consider the consequences:
cal choice for which solutions have “positive frequency,” Neutrino Dark Matter. Consider the standard
and observers in different regions will make inequiva- model of particle physics (including a right-handed neu-
lent choices: e.g. in FRW the positive frequency so- trino in each generation) coupled to Einstein gravity.
lutions ψ− and ψ+ chosen, respectively, by observers There is only one possible dark matter candidate in this
in the far past (τ → −∞) or far future (τ → +∞) model – one particle that has not yet been detected and
exhibit positive frequency behavior in Rthese two lim- can have a lifetime longer than the age of the universe
τ
its, respectively: ψ± (k, h, x) ∼ exp[−i ω(k, τ ′ )dτ ′ ] as – namely, one of the three right-handed neutrinos νR,1 .
τ → ±∞,pwhere k = |k| is the comoving wave number, This particle appears in two places in the Lagrangian:
c
and ω = k 2 + µ2 > 0 is the comoving frequency. The the Majorana mass term ν̄R,i Mij νR,j (where Mij is the
“−” solutions may then be expressed in the “+” basis: 3 × 3 Majorana mass matrix) and the Yukawa term
l̄L,i Yij νR,j hc (where lL,j is the left-handed lepton dou-
c
ψ− (k, h, x) = α(k)ψ+ (k, h, x) + β(k)ψ+ (−k, h, x). (6) blet, hc = iσ 2 h∗ is the charge conjugate of the Higgs
doublet h, and Yij is a 3 × 3 Yukawa coupling matrix).
We may adjust the phases of ψ+ and ψ− so that α(k) =
The assertion that νR,1 is exactly stable corresponds to
cos λ(k), β(k) = i sin λ(k), and λ(−k) = −λ(k) is real.
the statement that the standard model couplings respect
The “−” observer’s annihilation operators (a− , b− ) are
the Z2 symmetry νR,1 → −νR,1 . This symmetry sets
then related to the “+” observer’s annihilation operators
to zero the first column of the matrix Yij , whose three
(a+ , b+ ) by the Bogoliubov transformation
entries Yi1 would otherwise lead the νR,1 to decay.

a+ (+k, h)
  
cosλ(k) i sin λ(k) a− (+k, h)
 Thus, in the same limit that νR,1 becomes stable, it
† = † . (7) also becomes decoupled from all of the other particles in
b+ (−k, h) i sin λ(k) cosλ(k) b− (−k, h)
the standard model, and so might seem to become a poor
The observer in the far past (resp. far future) defines dark matter candidate since it is not produced by ther-
the vacuum to be the state |0− i (resp. |0+ i) that is an- mal processes in the early universe. But, in our picture,
nihilated by all the operators a− and b− (resp. a+ and these particles have a predictable non-zero cosmic abun-
b+ ): a± (k, h)|0± i = b± (k, h)|0± i = 0. We are in the dance, according to late-time comoving observers like us,
Heisenberg picture, so states do not evolve. Note that, just because the universe is in the CP T -invariant vacuum
unless sin λ(k) is identically zero, |0− i and |0+ i are in- |00 i, which differs from our late-time vacuum |0+ i. If the
equivalent: |0− i has no particles according to its own stable neutrino’s mass has a certain value, it automati-
particle number operator N− = a†− a− , but a non-zero cally has the abundance, coldness and darkness needed
number according to N+ = a†+ a+ . Moreover, since a and to match observations. This yields a strikingly simple al-
b transform as [CP T ]a± (k, h)[CP T ]−1 = −b∓ (k, −h) ternative explanation for the dark matter, different from
and [CP T ]b± (k, h)[CP T ]−1 = −a∓ (k, −h), the inequiv- previous neutrino dark-matter models based on thermal
alent vacua |0+ i and |0− i are exchanged by CP T , |0± i = or resonant production mechanisms [16–20].
CP T |0∓ i, so neither vacuum is CP T invariant. To see this explicitly, note that near the bang, i.e.,
However, if we define new operators a0 and b0 : during the radiation era, above the electroweak phase
     transition, when a(τ ) ∝ τ , the dark matter neutrino has
a0 (+k, h) cos λ(k) ∓i sin λ(k)
a (+k, h) equation of motion:
 = 2 2  ± ,
†
b0 (−k, h) λ(k)
∓i sin 2 cos 2λ(k)
b†± (−k, h) (i∂
/ − µ)N1 = 0 (µ = γτ ). (9)
(8) 3/2 c
Here N1 ≡ a (νR,1 + νR,1 )
and γ is a constant given
√
they transform as [CP T ]a0 (k, h)[CP T ]−1 = −b0 (k, −h) by γ = (mdm /mpl ) ρ1 , where mdm is the mass of the
and [CP T ]b0 (k, h)[CP T ]−1 = −a0 (k, −h), so the cor- right-handed neutrino νR,1 , mpl = (8πG/3)−1/2 ≈ 4 ×
responding vacuum defined by a0 |00 i = b0 |00 i = 0 is 1018 GeV is the Planck scale, and ρ1 = a4 ρ (the radiation
CP T invariant: CP T |00 i = |00 i. In fact, there is density times a4 ) is a constant.
a continuous family of CP T -invariant vacua, obtained To understand the behavior p of Eq. (9), consider the
by defining (aη (+k, h), b†η (−k, h)) to be a real SO(2) comoving frequency ω(τ ) = k 2 + µ2 . If ω were inde-
rotation of (a0 (+k, h), b†0 (−k, h)) through an angle η pendent of τ , the solutions would be N1 ∝ ei(kx−ωτ ) ,
 4

just as in Minkowski space. But since ω does depend leptogenesis [24, 25]. (iii) Since gravitational waves are
on τ , we turn to the WKB method. Consider the di- massless, the corresponding “+” and “−” vacua agree.
mensionless WKB parameter |ω ′ (τ )/ω 2 |: for fixed co- Thus, no long wavelength gravitational waves are pro-
moving wave number k, this vanishes near the bang (as duced by our mechanism.
|τ | → 0) and far from the bang (as |τ | → ∞), but Discussion. Let us end with a few remarks:
reaches a maximum value ∼ γ/k 2 at an intermediate i) Here we assumed a flat, radiation dominated FRW
conformal time |τmax | ∼ k/γ. Thus, for wave numbers background. In a forthcoming paper, we explain how this
k ≫ γ 1/2 , the WKB parameter is always ≪ 1, WKB re- background arises [26].
mains good and the Bogoliubov transformation between ii) In this Letter, we have described the background
(a− , b− ) and (a+ , b+ ) is trivial: sin λ(k) ≈ 0. On the spacetime geometry and radiation fluid purely classically,
other hand, for wave numbers k ≪ γ 1/2 , WKB is badly according to general relativity. A fuller treatment of the
violated, and the Bogoliubov transformation pis maximal: singularity to include the trace anomaly [15] and quan-
|sin λ(k)| ≈ 1. To see this, consider ω = µ2 + k 2 : in tum back-reaction requires semiclassical methods, involv-
the limit k ≪ γ 1/2 , µ2 dominates and k 2 is negligible ing complex classical solutions along the lines of [27–29].
unless the mode is far outside the Hubble horizon; so we iii) A fascinating open question is whether current ob-
can neglect the spatial gradient terms in (9) and solve servations allow the standard model or, more properly, its
Rτ
(iγ 0 ∂τ − µ)ψ = 0 to find ψ = ψ̂ exp[ikx + i µ(τ ′ )dτ ′ ] minimal extension incorporating neutrino masses, to re-
where ψ̂ = (ξ, −ξ), with ξ a constant 2-spinor. Since main valid all the way up to the Planck scale, or whether
µ(τ ) is odd, the solution switches from purely positive new physics is required below this scale. With the mea-
frequency in the far past to purely negative frequency in sured central values of the Higgs and top quark masses,
the far future (corresponding to |sin λ(k)| = 1). the Higgs quartic self-coupling λ runs to negative values
Thus, for late time observers like us, the num- at an energy scale below the Planck mass [30, 31]; how-
ber density ever a recent analysis suggests that a strictly positive λ
P Rndm of dark matter particles is ndm = all the way is only disfavored at the 1.5 or 2 σ level [32].
(2πa)−3 h d3 kh00 |N+ |00 i where the matrix element
is |sin (λ(k)/2)|2 , so that ndm ∼ (2πa)−3 γ 3/2 ∼ Even if the Higgs effective potential runs negative at large
(2π)−3 (mdm /mpl )3/2 ρ3/4 , where ρ3/4 ∼ s, the radiation vev, finite temperature corrections are sufficient to stabi-
entropy density. Since the ratio ndm /s is conserved dur- lize the Higgs field at zero vev in the very early universe.
ing the subsequent expansion, it must match the present There would only be an instability (to a negative-Higgs-
day value ndm,0 /s0 , where s0 ∼ 2.3 × 10−38 GeV3 [21], potential bulk phase) at late cosmological times, far to
ndm,0 = ρdm,0 /mdm is the present dark matter number our future. We find it intriguing that the most economi-
density, and ρdm,0 ∼ 9.7 × 10−48 GeV4 is the present cal possibility, of no new physics, may be viable [33], and
dark matter energy density [1]. Thus, we estimate might even explain the dark matter.
3/2 iv) We have seen that stability of the dark matter
mdm ∼ [(ρdm,0 /s0 )(2π)3 mpl ]2/5 ≈ few × 108 GeV. A
neutrino νR,1 implies that the Lagrangian has a symme-
more precise calculation [2] yields mdm = 4.8 × 108 GeV. try under νR,1 → −νR,1 . This symmetry suffers from
We emphasize that the definition of |00 i, and the re- no anomalies – not even gravitational anomalies [34].
sulting estimate of ndm , is controlled by CP T symmetry, It is well known that in the standard model, the lep-
not by the detailed physics of the bang itself. In particu- ton representations {lL , νR , eR } echo the quark represen-
lar, we have seen that the Bogoliubov transformation is tations {qL , uR , dR }. (This observation underlies Pati-
insensitive to the behavior of a (or µ) near τ = 0, where Salam grand unification [35], in which the leptons are a
the WKB parameter vanishes, and is instead dominated fourth color.) The parallel symmetry in the quark sector,
by the WKB bump experienced by modes of wave num- uR,1 → −uR,1 , is interesting for other reasons. Naively,
ber k ∼ γ 1/2 at a proper time t ∼ m−1 dm before or after it forces the bare mass of the up quark to zero which,
the bang (when the temperature is already orders of mag- in turn, solves the strong CP problem [36]. Unlike the
nitude below the Planck scale, and the usual radiation- symmetry we are using, this Z2 symmetry is anomalous
dominated Friedmann equation should be reliable). due to the strong interactions; however, if it holds at any
Other Predictions. Several other predictions follow energy and, in particular, at a very high energy scale, this
[2]: (i) The three light neutrino mass eigenstates are Ma- may be sufficient to solve the strong CP problem [37]. A
jorana particles (which will be tested by future neutri- deeper understanding of these symmetries will likely re-
noless double β-decay searches [22]), and one of them quire new insights into the origin of the three generations
is exactly massless (which will be tested by future cos- in the standard model.
mological constraints on the sum of the light neutrino Note added. Shortly after our Letter appeared on
masses [23]). (ii) We have focused on the stable right- the arXiv, a follow-up paper [38] pointed out that the
handed neutrino, but the other two (unstable) right- ANITA experiment may have already seen evidence for
handed neutrinos are thermally coupled and can explain our dark matter candidate.
the observed matter/anti-matter asymmetry by thermal Acknowledgements. We thank Claudio Bunster,
 5

Job Feldbrugge, Angelika Fertig, Steffen Gielen, Jaume 023527 (2006) [astro-ph/0605271].
Gomis, David B. Kaplan, Ue-Li Pen, Laura Sberna and [19] A. Boyarsky, O. Ruchayskiy and M. Shaposhnikov, Ann.
Edward Witten for discussions. Research at Perime- Rev. Nucl. Part. Sci. 59, 191 (2009) [arXiv:0901.0011
[hep-ph]].
ter Institute is supported by the Government of Canada
[20] L. Canetti, M. Drewes, T. Frossard and M. Shaposh-
through Innovation, Science and Economic Development, nikov, Phys. Rev. D 87, 093006 (2013) [arXiv:1208.4607
Canada and by the Province of Ontario through the Min- [hep-ph]].
istry of Research, Innovation and Science. [21] E. Kolb and M. Turner, The Early Universe (1990).
[22] J. J. Gomez-Cadenas, J. Martin-Albo, M. Mezzetto,
F. Monrabal and M. Sorel, Riv. Nuovo Cim. 35, 29 (2012)
[arXiv:1109.5515 [hep-ex]].
[23] K. N. Abazajian et al., Astropart. Phys. 35, 177 (2011)
[1] P. A. R. Ade et al. [Planck Collaboration], Astron. Astro- [arXiv:1103.5083 [astro-ph.CO]].
phys. 594, A13 (2016) [arXiv:1502.01589 [astro-ph.CO]]. [24] M. Fukugita and T. Yanagida, Phys. Lett. B 174, 45
[2] L. Boyle, K. Finn and N. Turok, arXiv:1803.08930 [hep- (1986).
ph]. [25] W. Buchmuller, R. D. Peccei and T. Yanagida, Ann. Rev.
[3] R. M. Wald, Phys. Rev. D 21, 2742 (1980). Nucl. Part. Sci. 55, 311 (2005) [hep-ph/0502169].
[4] H. Kodama and M. Sasaki, Prog. Theor. Phys. Suppl. [26] L. Boyle and N. Turok, in preparation (2018).
78, 1 (1984). [27] S. Gielen and N. Turok, Phys. Rev. Lett. 117, no.
[5] V. F. Mukhanov, H. A. Feldman and R. H. Branden- 2, 021301 (2016) doi:10.1103/PhysRevLett.117.021301
berger, Phys. Rept. 215, 203 (1992). [arXiv:1510.00699 [hep-th]].
[6] L. Boltzmann, Wissenschaftliche Abhandlungen, Vol. I, [28] S. Gielen and N. Turok, Phys. Rev. D 95, no.
II, and III, F. Hasenohrl (ed.), Leipzig: Barth (1909); 10, 103510 (2017) doi:10.1103/PhysRevD.95.103510
reissued New York: Chelsea (1969). [arXiv:1612.02792 [gr-qc]].
[7] A. Aguirre and S. Gratton, Phys. Rev. D 67, 083515 [29] J. Feldbrugge, J. L. Lehners and N. Turok, Phys. Rev. D
(2003) [gr-qc/0301042]. 97, no. 2, 023509 (2018) [arXiv:1708.05104 [hep-th]].
[8] S. M. Carroll and J. Chen, hep-th/0410270. [30] G. Degrassi, S. Di Vita, J. Elias-Miro, J. R. Espinosa,
[9] J. Barbour, T. Koslowski and F. Mercati, Phys. Rev. G. F. Giudice, G. Isidori and A. Strumia, JHEP 1208,
Lett. 113, no. 18, 181101 (2014) [arXiv:1409.0917 [gr- 098 (2012) [arXiv:1205.6497 [hep-ph]].
qc]]. [31] D. Buttazzo, G. Degrassi, P. P. Giardino, G. F. Giu-
[10] J. Barbour, T. Koslowski and F. Mercati, dice, F. Sala, A. Salvio and A. Strumia, JHEP 1312,
arXiv:1507.06498 [gr-qc]. 089 (2013) [arXiv:1307.3536 [hep-ph]].
[11] J. Barbour, T. Koslowski and F. Mercati, [32] A. V. Bednyakov, B. A. Kniehl, A. F. Pikelner and
arXiv:1604.03956 [gr-qc]. O. L. Veretin, Phys. Rev. Lett. 115, no. 20, 201802
[12] S. Goldstein, R. Tumulka and N. Zanghi, Phys. Rev. D (2015) [arXiv:1507.08833 [hep-ph]].
94, 023520 (2016) [arXiv:1602.05601 [astro-ph.CO]]. [33] M. Shaposhnikov and C. Wetterich, Phys. Lett. B 683,
[13] E. C. G. Stueckelberg, “La signification du temps propre 196 (2010) [arXiv:0912.0208 [hep-th]].
en méchanique ondulatoire,” Helv. Phys. Acta 14, 322 [34] E. Witten, private communication.
(1941). [35] J. C. Pati and A. Salam, Phys. Rev. D 10, 275 (1974)
[14] E. C. G. Stueckelberg, “Remarque à propos de la création Erratum: [Phys. Rev. D 11, 703 (1975)].
de paires de particules en théorie de relativité,” Helv. [36] D. B. Kaplan and A. V. Manohar, Phys. Rev. Lett. 56,
Phys. Acta 14, 588 (1941). 2004 (1986).
[15] N. D. Birrell and P. C. W. Davies, Quantum fields in [37] M. Srednicki, Phys. Rev. Lett. 95, 059101 (2005)
curved space, Cambridge University Press (1984). [hep-ph/0503051].
[16] S. Dodelson and L. M. Widrow, Phys. Rev. Lett. 72, 17 [38] L. A. Anchordoqui, V. Barger, J. G. Learned, D. Mar-
(1994) [hep-ph/9303287]. fatia and T. J. Weiler, LHEP 1, no. 1, 13 (2018)
[17] X. D. Shi and G. M. Fuller, Phys. Rev. Lett. 82, 2832 doi:10.31526/LHEP.1.2018.03 [arXiv:1803.11554 [hep-
(1999) [astro-ph/9810076]. ph]].
[18] K. Abazajian and S. M. Koushiappas, Phys. Rev. D 74,