You are on page 1of 8

The critical geometry of a thermal Big Bang

Niayesh Afshordi
Perimeter Institute for Theoretical Physics, 31 Caroline St. N., Waterloo, ON, N2L 2Y5, Canada and
Department of Physics and Astronomy, University of Waterloo, Waterloo, ON, N2L 3G1, Canada

Joao Magueijo
Theoretical Physics, Blackett Laboratory, Imperial College, London, SW7 2BZ, United Kingdom
(Dated: November 9, 2016)
We explore the space of scalar-tensor theories containing two non-conformal metrics, and find a
discontinuity pointing to a critical cosmological solution. Due to the different maximal speeds
of propagation for matter and gravity, the cosmological fluctuations start off inside the horizon
even without inflation, and will more naturally have a thermal origin (since there is never vacuum
arXiv:1603.03312v2 [gr-qc] 8 Nov 2016

domination). The critical model makes an unambiguous, non-tuned prediction for the spectral
index of the scalar fluctuations: nS = 0.96478(64). Considering also that no gravitational waves are
produced, we have unveiled the most predictive model on offer. The model has a simple geometrical
interpretation as a probe 3-brane embedded in an EAdS2 E3 geometry.

1. Introduction. In spite of its mathematical simplic- to fit the observed nS , from within a range of possibilities.
ity and observational triumphs, the Big Bang model of This is not to say that they entirely lack predictivity; in-
the Universe remains an unfinished work of art. Many deed they do predict a plethora of conditions involving
of its late-time successes can be traced to the initial con- nS and other observables (e.g. [10]).
ditions postulated for its early stages, and these are put In this Letter, we revisit a class of VSL models [11]
in by hand, without justification, other than to retrofit in which there are two non-conformal metrics, one for
the data. The main culprit for this shortcoming is the matter and another for gravity, so that light and other
so-called horizon problem: the cosmological structures massless matter particles travel faster than gravity. Con-
we observe today span scales that lay outside the ever- ditions for the observational success of such models have
shrinking horizons of physical contact that plagued the been identified [10, 1216], considering both a vacuum
early universe. This precludes a causal explanation for and a thermal initial state. In common with other mod-
their initial conditions. els, they do not bypass the criticism voiced above. How-
Several extensions of the Big Bang model have been ever, in this Letter we uncover a remarkable result per-
proposed with the aim of opening up its horizons. An taining to thermal scenarios.
early bout of accelerated expansion [13], a contracting It is known that thermal VSL models require a fast
phase followed by a bounce [4], a loitering early stage [5], phase transition in c so as to produce near-scale-invariant
and a varying speed of light (VSL) [6, 7] have all been fluctuations; however, the scale-invariant limit (nS = 1)
considered. None of these proposals evades the criticism is unreachable. Closer inspection of the space of all pos-
that retrofitting the data is still used to select in detail the sible theories reveals that this is due to a discontinuity,
primordial fluctuations that the model should produce. pointing to a special, critical solution that should be re-
Once primordial causal contact is established, work can garded as the preferential model for a phase transition
start on concrete physical mechanisms for spoiling per- in c. Instead of nS = 1, the thermal fluctuations in
fect homogeneity (e.g. vacuum quantum fluctuations or this model display a running nS < 1. But what is truly
thermal fluctuations). Typically it is found that one can notable is that the model has a single free parameter,
produce a wide range of initial conditions including, but so that the amplitude AS fully fixes the value of nS at
not circumscribed to those explaining the observations. the observationally relevant scales. The predicted value
is within current constraints, but improved observations
Specifically, the primordial density fluctuations can be
would unambiguously prove or rule out the theory. The
characterised by an amplitude AS , measuring their inten-
model also has a simple geometrical interpretation as a
sity at a given scale, and a spectral index nS , measuring
probe 3-brane embedded in an EAdS2 E3 geometry.
how the amplitude varies with scale. Observations [8, 9]
2. The critical model. We start by reviewing the
show that AS = 2.142(49) 109 and nS = 0.9667(40),
general framework of scalar-tensor bimetric theories. In
signalling a very specific slightly red spectrum, i.e. one
these models there are two metrics (or frames): g asso-
with enhanced amplitude for longer wavelengths. Whilst
ciated with the gravitational action (the Einstein frame),
the observed AS probably indicates nothing more than
and g , to which matter is minimally coupled (the mat-
a hierarchy between two energy scales, one might have
ter frame). The action takes the general form:
expected a theoretical forecast for nS . Yet, all theories
effectively adjust their free parameters (e.g. the reheating M2
Z Z p
temperature after inflation, or the number of e-foldings) S= P d4 x gR[g ]+ d4 x gLM (, g )+S
2
2

where MP is the reduced Planck mass, and S encodes This is nothing but the cuscuton model [18, 19], and in-
the dynamics of the field relating the two metrics. If deed
the speed of sound is infinite in this limit (cs
the metrics are conformally related we have a varying- 2BX ). The model has conformal (Weyl) sym-
G theory, such as Brans-Dicke theory. In varying-c metry [20, 21], so that any scale-factor a(t) is a solution.
theories, rather, the metrics are non-conformally related: This implies that spatial flatness is compulsory and fully
fixes V [1820]. If and p denote density and pressure,
g = g + B( )( ), (1) we have V and p + K [18, 22], where K = / B
is the kinetic energy. The (spatially flat) Friedmann and
so that the light cones spanned by massless matter parti- continuity equations are:
cles and by gravitons do not coincide [13, 14]. In general
B (also known as the warp factor, for reasons to be made 3MP2 H 2 V and V + 3HK 0, (5)
obvious soon) is a function of . If the speed of light is to
be larger than that of gravity, then B > 0 (with signature where H = a/a.
These can be integrated as:
+ ). !2
It may seem that the number of theories of this type is
 
3 d V
Z
endless, but this is not the case. The simplest non-trivial V () = +O , (6)
4MP2 c2s
p
B()
S must consist of two generally non-constant cosmolog-
ical terms, one in the matter frame and the other in the fixing V as a function of B. Here = H/H 2
= 23 (1+w),
Einstein frame: where w = p/, and cs is the speed of sound in the matter
frame. Although any w is possible, it can be shown that
Z Z
S = d4 x g(2m ()) + d4 x g(2g ()) .
p
as c2s = p/|X , the next order corrections yield
(2) w = p/ for generic solutions [23, 24].
Furthermore, only one of the functions m (), g () and So far we have merely reviewed old results. Now we
B() is independent [17]. This has been known for a come to the crucial element of this Letter. It has been
while; here we sketch the proof in two steps. shown [10, 12, 16] that thermal bimetric scenarios are
Firstly, let be canonically normalized in the general- close to scale-invariance whenever B() n , with n
ized sense that it should exhibit a Klein-Gordon equa- close to 2. Then, the potential V is still a power-law, but
tion of motion in the matter frame when no matter its exponent is close to zero (cf. Eqn. (6)). The variation
is present. Then, its action should be a cosmological in cs is abrupt, but one still has constant s = cs /(cs H),
term in the matter frame, due to a simple calculation with s as n 2. The cosmological solutions
in variational calculus in the presence of two metrics are scaling solutions, i.e. they have constant and s ,
(see [13, 14]). Furthermore, if the field dynamics is to leading to thermal fluctuations with constant nS , which
be driven by 2m , regarded as a potential, we should can be tuned to be as close to 1 as wanted. Indeed
have 2m () = 1/B(). This fully fixes the first term +1
of (2) in terms of B, and it is known [14] that it amounts nS 1 = , (7)
s + 1
to postulating a DBI action in the Einstein frame
and although nS = 1 is unreachable, any red spectrum
4 1
Z  
S = d x g 1 + 2BX V , (3) as close to scale-invariance as required can be obtained
B by suitably tuning B.
It should be immediately obvious from Eqn. (6) the
with X = 12 ( )( ) and V = 2g . This can be de- reason why the scale-invariant limit cannot be reached.
rived from simple properties of determinants, and will be Within the space of these theories, there is a discontinuity
important in seeking a geometrical interpretation for our at B 2 , because the potential fails to be a power-law.
critical model. All the theories around it imply power-law potentials,
Secondly, of the two remaining free functions (B and but this critical theory stands out as an exception:
V ) only one is free in the UV limit, which is the limit of  2
interest to us. In the physical situation we are consid-
Bcrit. () = B0 , (8)
ering, the action (3) should more strictly be called anti- MP
DBI, since the sign of B is opposite to the usual one, 3



so that the speed of light is larger, rather than smaller Vcrit. () = ln2 . (9)
4B0 MP
than that of gravity. Thus, the UV limit of the theory is
achieved with X 1 (instead of saturating at an upper It marks a special, crucial boundary in the space of the-
bound, as is the case with the usual DBI theory), so that: ories. The critical model is unique in that it cannot have
r   a constant s , since V is no longer a power-law. This
2X 1 induces natural deviations from scale-invariance, making
L V +O . (4)
B B3X its phenomenology remarkable, as we will show presently.
3

3. Geometrical interpretation. Before embarking upon MP z, subject to dynamical equation:


the phenomenology of the critical model we reinforce its
z
 
special status by uncovering an elegant geometrical in- 2 2
v + cs k v = 0, (14)
terpretation. It is known that the DBI action can be z
derived from the induced metric on a probe 3-brane em- where k is the comoving wave-number. The central quan-
bedded in a higher dimensional geometry, with the B tity to be computed is cs , and this is given by [14, 18, 22]:
function interpreted as a geometrical warp factor. For
example, in the celebrated DBI action associated with 2 2 4
q
B0
cs = 1 + 2BX B B0 e 3 , (15)
the motion of a probe 3-brane in AdS5 S5 geometry, 3 3
one finds B 4 , with interesting cosmological im-
where the first identity is generic for (anti-)DBI models,
plications [25, 26].
in the second step we used 32 = K K
V and K
Likewise, B() 2 follows from embedding a 3-brane p
in the EAdS2 E3 geometry given by: 2X/B, and in the third we used Eqns. (8) and (9). The
fact that V (and so ) is not a power-law in , explains
r 2 2 R2 2 why the model has a cs which is not a power-law in a or
d52 = dt + 2 dr dx2 dy 2 dz 2 , (10) . Even if the background scales (constant ), the speed
R2 r
of sound does not, with a varying s given by:
where R is the radius of the Euclidean AdS2 . Ignoring r !
the gravitational backreaction, the induced action on a cs B0
uniform probe 3-brane at r(t) is given by s = = 2 1 + 2 . (16)
cs H 3
r
r2 R2 r 2
Z
4 Thus, nS is expected to run, a property that can be
S3B = T3 d x 2
+ 2 , (11)
R r guessed from (7). However that formula is incorrect for
varying s , indeed many standard formulae in the liter-
where T3 is the brane tension (with mass units M 4 ). ature [14, 16, 22] break down. A full derivation of nS
Field is a redefinition of r that renders (11) canoni- can be found in the Supplementary Material [23] (SM)
cal in the IR limit, and a Taylor expansion shows that (in this version of our paper included in appendix). Here
this is given by r = 4R3 T3 /2 . Straightforward algebra we present an approximate calculation, good enough to
shows that this brings (11) to the anti-DBI form (3) with extract all the salient features.
B matching the critical model (8) and As usual, Eq. (14) has two regimes, an acoustic one
 2 and a gravitational instability one, depending on which
MP of its two terms in v dominates. The two regimes are
B0 = . (12)
2RT3 separated by the sound horizon scale, kh , where these
terms become of the same order:
Turning on gravity for the effective 4D geometry, the
potential is fixed by Eq. (6), as a result of the conformal z
c2s kh2 (aHs )2 (aHs ) (aHs )2 , (17)
invariance of the theory in the UV [20, 21]. z
A crucial novelty here is that the extra dimension, r, (for simplicity, we have assumed a constant , but in fact
is time-like rather than space-like, something also dis- this is not necessary). The sound horizon scale therefore
cussed in string theory literature [27]. While this may satisfies cs kh aHs , and we note the extra factor of s
raise alarm about ghost instabilities for the bulk, the with regards to the usual formula. Matching the 2 types
ghost degrees of freedom may be made arbitrarily heavy of solution is sufficient to derive to a good approximation
and thus decouple from the 4D low-energy effective field the power spectrum frozen-in outside the horizon. For
theory [24]. k kh the solutions should be normalized as [14, 16, 22]:
4. Density fluctuations. We now come to the core of R
cs k d
this Letter, the evaluation of the thermally induced fluc- ei
v= , (18)
tuations for the critical solution. This can be done fol- 2cs k
lowing well-known methods developed for theories with a
varying speed of sound cs [14, 16, 22], since that is what whereas for k kh the growing mode takes the form
our theory is in the Einstein frame. The second order v = F (k)z. By means of simple algebra F (k) can be
action for the curvature fluctuation is: found by matching the two expressions at k kh .
The square of F (k) is nothing but the frozen-in power
1 2 spectrum of , up to a factor representing the expectation
Z
S2 = MP dd3 x z 2 2 c2s ()2 ,
 
(13) + 1, where N is the number operator (note
2 value of 2N
that upon quantization v 2 is multiplied by a a
+a ,
a
where z = cas 2 and is conformal time. Therefore where a is an annihilation operator). For vacuum fluc-
we have a standard quantum field theory in variable v = tuations, this factor is simply 1, whereas for a thermal
4

state it is twice the thermal occupation number of mode within the allowed observational range of (6.5 7.6)
k in the Rayleigh-Jeans limit [14, 16]. This is 2Tc /k, 103 (see [8]). As for non-Gaussianity we find an am-
where Tc = T cs /a is a conformal temperature which plitude for the bispectrum of order unity, fN L = O(1),
remains constant during the varying-c phase [14, 16]. The comparable to similar models [10] but with a very dif-
frozen-in dimensionless power spectrum of the thermally ferent and unique shape (to be reported elsewhere [24]).
induced fluctuations is therefore: We have unveiled the most predictive model on offer.
k3 1 2s Tc What are the provisos of our claims? From the above
2 th
Pth (k) h|| i , (19) we can work out that B0 MP4 6.6 1013, so that the en-
2 2 24 2 cs Mp4 k
ergy scale at the end of the transition is 3.5 104 MP ,
where the right hand side is to be evaluated at horizon with the current horizon scale leaving the sound hori-
crossing (now cs k aHs ). We stress the extra factors zon less than 3 orders of magnitude above this. So we
in s found in Eq. (17) and (19), in relation to standard never exceed the Planck scale (in common with other
formulae [22]. They are irrelevant if s is a constant, but thermal varying-c scenarios [14, 16]), allaying the first ob-
not in our case. Eq. (19) is valid up to factors of order vious criticism. Then, there are model uncertainties. The
one (fully restored in the SM [23]; see appendix). equation of state can have an effect on the final result (for
Combining Eqs. (17) and (19) and using the chain rule example, = 2 would push nS down to nS = 0.95292);
we find for the spectral index: however we have arguments for why 1 is generic in
 12 our model(see [23] and Appendix). Furthermore, in eval-
d ln Pth
  
1 + 2 B0 1 uating Tc at horizon crossing we have assumed entropy
nS 1 = = +O ,
d ln k 4 3 B0 conservation in the constant c phase (the change in g
where is the density when k = kh . Thus, nS runs from drops out of the final result), but more importantly we
very red, at the largest scales, to almost scale-invariant, have assumed no reheating at the end of the varying-c
at the smallest. However, as announced, the observed phase. This is because in our scenario any such process
amplitude AS fixes where we are in this running flow. would be ad hoc and unnecessary, since the universe is al-
Note that the model has a single free parameter, the 4- ways hot. Nonetheless, we note that a reheat by a factor
volume scale B0 . Although Eq (19) seems to depend both of, for example, 1010 would push nS up to nS = 0.96838.
on B0 MP4 and B0 , the former can be eliminated by us- An isothermal gluing of the two phases remains the most
ing cs k = aHs , the Friedman equation, and some basic minimal assumption.
thermodynamics, to recast it in the descriptive form: 5. Discussion. In summary, we built upon previous
work on thermal fluctuations in bimetric scenarios which
g0 T 3 2
 p 
showed that a sufficiently fast phase transition in cs leads
Pth (k) 30 = C(B0 )1+ exp 4 3B0 (20)
k to fluctuations as close to scale-invariance as seen in the
where C is a numerical constant [23]. The left hand data [14, 16]. In such scenarios, fitting the observed nS
side can be evaluated from observations. For a given requires fine-tuning the warp factor B(). Here we im-
mode (say, k = 0.05 Mpc1 ) the first factor is the ob- proved on this by discovering that the reason why exact
served amplitude (P (k) = 2.142(49) 109 , [8]), and scale-invariance is never achieved is that the limit is dis-
the second is the dimensionless entropy inside the scale continuous, pointing to a critical solution with quadratic
k nowadays (with g0 = 3.91 the effective number of warp factor, but a non-power-law potential (fully deter-
relativistic degrees of freedom). As stated above (and mined by the Bianchi identities and UV conformal sym-
in [23]), generically , so we can solve (20) to get metry). The critical solution has a simple geometrical in-
B0 583.03(16), where the uncertainties are both ob- terpretation as the (anti-)DBI action of a probe 3-brane
servational and arise from the fact that the model is only embedded in an EAdS2 E3 geometry. The non-power-
reliable to O( B10 ). Using (20) we thus obtain: law nature of the potential induces a non-scaling speed
of sound, which in turn produces a natural red tilt and
nS = 0.96478(64), (21) running of the power spectrum.
But what makes the model remarkable is that the am-
well within the most stringent current observational con- plitude AS for a given scale fixes its location on this over-
straints (viz. nS = 0.9667(40), cf. [8]). arching structure, leading to a single prediction for the
This is a remarkable result. But the model makes fur- observed nS . The model does not require reheating, and
ther predictions. It produces no tensor modes (since the this is the ultimate reason why it is more predictive than
horizon problem is not solved for gravitons), and so sin- inflation, even if factors external to cosmology were to
gles out a point in the {nS , r} diagram, with r = 0. It pre-select one of inflations many models. Inflationary
also predicts (cf. Eqs. (17), (19) and (20)) the running models invariably predict a range of nS , depending on
of the spectral index to be: the number of e-foldings, or the reheating temperature
dnS 3 (even for a fixed choice of inflaton action). Adding to
= (ns 1)2 1.8 103 , (22) this the fact that our model makes precise predictions
d ln k 2
5

for the level of primordial gravitational waves (r = 0), [9] P. A. R. Ade et al. (Planck), (2015),
the running of nS , and non-Gaussianity, we can conclude arXiv:1502.02114 [astro-ph.CO].
without prejudice that we have in hand a very predic- [10] J. Magueijo, J. Noller, and F. Pi-
azza, Phys. Rev. D82, 043521 (2010),
tive model indeed. The fact that its main prediction (for
arXiv:1006.3216 [astro-ph.CO].
nS ) lies spot in the middle of the Planck results should [11] J. Magueijo, Rept. Prog. Phys. 66, 2025 (2003),
not beguile us into a false sense of security. Improved arXiv:astro-ph/0305457 [astro-ph].
observations will soon vindicate or disprove this model. [12] J. Magueijo, Phys.Rev.Lett. 100, 231302 (2008),
One may wonder about the status in our model of the arXiv:0803.0859 [astro-ph].
[13] M. A. Clayton and J. W. Mof-
other cosmological problems, such as the flatness, homo-
fat, Phys. Lett. B477, 269 (2000),
geneity and isotropy problems. Firstly, the view may be arXiv:gr-qc/9910112 [gr-qc].
held that such historically motivating problems are now [14] J. Magueijo, Phys. Rev. D79, 043525 (2009),
considered to be of lesser importance than explaining the arXiv:0807.1689 [gr-qc].
structure of our Universe, or may even be misguided [28]. [15] D. Bessada, W. H. Kinney, D. Stojkovic,
Nonetheless we remark that it is possible to solve them and J. Wang, Phys.Rev. D81, 043510 (2010),
using the VSL mechanism before the phase transition arXiv:0908.3898 [astro-ph.CO].
[16] A. Agarwal and N. Afshordi,
(e.g. [7, 13]). In other scenarios it may also happen that
Phys. Rev. D90, 043528 (2014),
their solution takes place in a different phase to struc- arXiv:1406.0575 [astro-ph.CO].
ture formation. Furthermore, we find that at least the [17] While these functions, in general, can depend ()2 and
flatness problem can be solved, in a single package, dur- higher derivatives, we assume that such dependence is
ing the phase transition. The conformal symmetry of the suppressed by a UV scale (e.g., Planck energy) and can
theory in the UV [20, 21] not only fixes the potential but be neglected in the regime of validity of effective field
requires exact flatness (Eqs.(5) lead to a contradiction in theory.
[18] N. Afshordi, D. J. H. Chung, and G. Gesh-
the presence of spatial curvature). A full investigation of
nizjani, Phys. Rev. D75, 083513 (2007),
these matters is deferred to [24]. arXiv:hep-th/0609150 [hep-th].
Acknowledgments We would like to thank Robert Bran- [19] N. Afshordi, D. J. H. Chung, M. Doran, and
denberger, Keith Copsey, Giulia Gubitosi and Sarah G. Geshnizjani, Phys. Rev. D75, 123509 (2007),
Shandera for helpful discussions. The work of NA was arXiv:astro-ph/0702002 [astro-ph].
[20] N. Afshordi, Phys. Rev. D80, 081502 (2009),
partially supported by the Natural Science and Engineer-
arXiv:0907.5201 [hep-th].
ing Research Council of Canada, the University of Wa- [21] P. Horava, Phys.Rev. D79, 084008 (2009),
terloo and by Perimeter Institute for Theoretical Physics. arXiv:0901.3775 [hep-th].
Research at Perimeter Institute is supported by the Gov- [22] J. Garriga and V. F. Mukhanov,
ernment of Canada through Industry Canada and by the Phys.Lett. B458, 219 (1999),
Province of Ontario through the Ministry of Research arXiv:hep-th/9904176 [hep-th].
& Innovation. JM acknowledges support from the John [23] See Supplemental Material at [web site].
[24] N. Afshordi, J. Magueijo, and J. Noller, in preparation.
Templeton Foundation, an STFC consolidated grant and
[25] J. M. Maldacena, Int. J. Theor. Phys. 38, 1113 (1999),
the Leverhulme Trust, and thanks the Perimeter Insti- [Adv. Theor. Math. Phys.2,231(1998)],
tute for hospitality. arXiv:hep-th/9711200 [hep-th].
[26] E. Silverstein and D. Tong,
Phys. Rev. D70, 103505 (2004),
arXiv:hep-th/0310221 [hep-th].
[27] R. Dijkgraaf, B. Heidenreich, P. Jefferson, and C. Vafa,
[1] A. H. Guth, Phys. Rev. D23, 347 (1981). (2016), arXiv:1603.05665 [hep-th].
[2] A. D. Linde, Second Seminar on Quantum [28] S. M. Carroll, (2014), arXiv:1406.3057 [astro-ph.CO].
Gravity Moscow, USSR, October 13-15, 1981, [29] V. F. Mukhanov, H. Feldman, and R. H. Brandenberger,
Phys. Lett. B108, 389 (1982). Phys.Rept. 215, 203 (1992).
[3] A. Albrecht and P. J. Steinhardt,
Phys. Rev. Lett. 48, 1220 (1982).
[4] P. J. Steinhardt and N. Turok, Science 296, 1436 (2002).
[5] A. Nayeri, R. H. Brandenberger, and APPENDIX: SUPPLEMENTAL MATERIAL
C. Vafa, Phys. Rev. Lett. 97, 021302 (2006),
arXiv:hep-th/0511140 [hep-th].
[6] J. W. Moffat, Int. J. Mod. Phys. D2, 351 (1993), A. Cosmological Perturbation Theory
arXiv:gr-qc/9211020 [gr-qc].
[7] A. Albrecht and J. Magueijo,
The FRW metric with scalar linear perturbations in
Phys. Rev. D59, 043516 (1999),
arXiv:astro-ph/9811018 [astro-ph]. the longitudinal gauge is given by:
[8] P. A. R. Ade et al. (Planck), (2015),
ds2 = a2 () (1 + 2)d 2 (1 2)dx dx .
 
arXiv:1502.01589 [astro-ph.CO]. (23)
6

Observational constraints on scalar adiabatic perturba- where a k and ak are the creation and annihilation opera-
tions are often described in terms of the gauge-invariant tors for particles (or phonons) of momentum k around a
Bardeen variable [29]: gaussian vacuum state |0i, which, by definition, has zero
particles.
H
(H + ). (24) The adiabatic vacuum state |0iad. is defined by the
H condition that mode functions vk (y) approach the posi-
Following [22], we shall adopt the following quadratic tive frequency (flat space) limit, when y , which is
action for the Bardeen variable: also where adiabatic approximation in Eq. (34) becomes
exact:
1
Z
S = MP2 dyd3 x q 2 2 ()2 ,
 
(25) exp(iky)
2 vk (y) , when y , (36)
2k
for acoustic waves of speed cs , where
while its subsequent evolution follows from exactly solv-
a 2 ing the mode equation (34). This ensures that the adi-
q , (26) abatic vacuum coincides with the ground state of the
cs
Z
cs dt
Z Hamiltonian at infinite past.
y = cs d, (27) It turns out that Eq. (34) with the initial condition
a
(36) has an exact solution in terms of the Hankel function
H of the 2nd kind (or Bessel functions of 1st and 2nd kind):
, 2, (28)
y H
y i (2)
MP (8GN )1/2 = 2.435 1018 GeV. (29) vk (y) = e H (ky), (37)
2
We shall call y the tachyo-conformal time (which is also where
equal to the comoving sound horizon), and MP is the
= (2 + 1). (38)
reduced Planck mass. 4
We can change to the Mukhanov-Sasaki variable: The late-time power spectrum of in a thermal state
of temperature T is given by:
v MP q, (30)
2
k 3 |vk |
which is canonically normalized: hP (k)iT = lim [2hnk iT + 1]. (39)
y0 2 2 q 2 MP2
q
 
1
Z
S= dyd3 x v 2 (v)2 + v 2 , (31) Here, the thermal particle occupation number is given by
2 q the Bose-Einstein distribution:
leading to the mode functions that obey the field equation 1
hnk iT = kcs
 . (40)
in the Fourier space: exp aT 1
q We can also use the asymptotic form of Hankel function
 
vk + k 2 vk = 0. (32) for small arguments:
q
2 41 ()2
If we have |vk | = + O(y 22 ). (41)
y 21 k 2

q 2 1/4 Combining Eqs. (33) with (39)-(41) yields:


q = Q(y)1/2 = , (33) " #
q y2 []2 2
hP (k)iT = 3 2 2 + 1 (k/2)32 .
MP Q exp kc

aT 1
s
our mode equation becomes
(42)
Notice that, in the Rayleigh-Jeans limit kcs aT , we
2 1/4
 
vk + k 2 vk = 0. (34) have hnk iT k 1 , and thus the scalar spectral index is
y2 given by:
So far, we have only considered the classical equations ns 1 = 2 2 0 1. (43)
for linear perturbations. Following the standard canon-
ical quantization procedure, we can decompose the free Therefore, close to scale-invariance in the tachyacoustic
quantum fields in the Heisenberg picture as: phase, the power spectrum takes the form:
Tc Tc dq 2
hP (k)iT 3 2 2 3 2
Z
d3 k h i MP Q MP dy
v(x, y) = vk (y) ak eikx ,
ak eikx + vk (y)  
(2) 3 d ln(cs /) HTc
= +2 , (44)
(35) d ln a 2 3 MP2 a
7

where we used the definitions of q and Q (Eq. 26 and 33), which can be combined to give:
while Tc aT /cs . Furthermore, using the definition 
P 5 62/ (B0 )2+2/ exp 4 3B0
(Eq. 28), and for a rapidly decaying speed of sound s = 2 . (53)
k3 9 5 g0 T03

ddln cs q
ln a 1, we can further simplify the expression for 1+2 B0
3
the power spectrum:
d ln(cs /) Tc Subsequently, for scalar spectral index we get:
hP (k)iT . (45)
dH 1 2 3 MP2 a
q
2(1+2) B0
d ln P 1 + 1+ 3
The comoving wavenumber at which the power spectrum nS 1 = = q
d ln k
 
B0
1+ 2+4
+ 8B0
freezes to this value is similarly given by: 1+ 3 3(1+)
 1/2
q2 d(/cs ) 1 + 2 B0
k |y|1 = = a (46) = + O(B0 )1 ,(54)
Q 2 dH 1 4 3
while its running is given by:
We now evaluate the thermal fluctuations for the crit-
ical solution discussed in the main text. The model has dnS 3 (1 + 2) 3/2
= + O (B0 ) . (55)
a single free parameter, the 4-volume scale B0 . Its speed d ln k 16 B0
of sound is: Up to here, we have provided results for general .
However, we shall argue below that is the only
r !
2 4 2 2 2 4 2 2 B0
cs 2BX B (B0 ) exp 8 . expected consistent asymptotic behavior as we approach
9 9 3
the critical bimetric model in the UV limit, and thus we
(47) shall focus our predictions to .
Conservation of entropy relates temperature to expansion Planck 2015 TT,TE,EE+lowP+lensing+ext (1-sigma)
and acoustic history: Table 4 [8] gives:
 3  1/3
Ta g0 P = (2.142 0.049) 109 , (56)
g0 T03 = g T c = T a = T 0 , (48)
cs g at k = 0.05 Mpc1 = 3.198 1031 eV. (57)
where For these values, Eq. (53) can be solved iteratively to
give:
g0 = 3.91, T0 = 2.73 K = 2.35 104 eV, (49)
B0 = 580 for = (58)
are the effective number of degrees of freedom today, and 6
the CMB temperature respectively. T , a , and g & 107 We see that the spectral index as becomes:
are the temperature, scale factor, and the effective rela-
tivistic degrees of freedom at the end of the tachyacoustic nS ( = ) 0.96478(64). (59)
2
phase, where c2s 13 , 2 6B0 = 6B0 30 g T4 The theoretical uncertainty is estimated by difference be-
1, using Eq. (47). Furthermore, Eq. (47) implies: tween the first and second lines in Eq. (54), as the ana-
r ! lytic model is only reliable to leading order in (B0 )1/2 .
d ln(cs /) B0 The prediction 59 is well within the observed range
= 2 1 + 2 . (50)
d ln H 3 (Table 4 in [8]):

Using Eq. (45)and Friedmann equation 3MP2 H 2 = nS = 0.9667 0.0040, (60)


a2 (assuming constant ) leads to an expression for the Planck 2015 + lowP + lensing + ext.
power spectrum: For, (and g 107), we can further find the
d ln(cs /) Tc density and temperature at the end of the tachyacoustic
P = phase:
dH 1 2 3 MP2 a
1 1 1
r ! 1
51/4 6 2 1/4 (B0 ) 2 + 2 B0 = = 9.0 1014 , (61)
= 1/2 7/2 g 1+2 ,(51) MP4 6B0 MP4
3 (B0 MP4 )3/4 3
T = 2.24 104 MP = 5.5 1014 GeV. (62)
and the sound horizon crossing wavenumber:
Eqs. (54-55) also give the running for the scalar spectral
d(/cs ) 31/2 1/2 1/3 1/12 index for :
k = a = 1 g0 g
dH 1 6 251/4  dnS 3
  = (nS 1)2 = 1.8 103 , (63)
d ln k 2
q q
T0 1 + 2 B30 exp 4 B30
, (52) well within the allowed observational range of (6.5
7.6) 103 (Eq. 42 in [8]).
1 1
(B0 MP4 )1/4 (B0 ) 2 + 2
8

3/2
B. Approaching the UV limit, and where we have extracted the pre-factor B0 for W (),
as subleading corrections are suppressed by powers of
1/2
While we have used symmetry and consistency prin- B0 . Note that the kinetic term is fully fixed by the
ciples to construct the UV limit of the critical bimetric EAdS2 E3 symmetry in the no-gravity limit, and thus
theory in the main text, the subleading UV behavior re- the subleading correction, W (), only appears in the
mains unconstrained, and is reflected by the freedom to potential. Now, combining Friedmann and homogenous
choose . Here, we carefully study this subleading behav- field equations for this action, and expanding in powers
1/2
ior and prove that the only consistent choice is to have of B0 , we find:
as cs .
Equations (3), (8), and (9) in the main text provide
the action for the critical UV tachyacoustic model. Let
us write the sub-leading corrections to the Lagrangian
(setting MP = 1 for simplicity):
"p #
4 1 + B0 2 ()2 3(ln )2 W ()
Z
S = d x g 3/2 ,
B0 2 4B0 B0
(64)

3/2
B0 n h io
ln + 2 1 + 3 ln + (W () W () ln ) 3 + O(B 2 ) = 0.
0 (65)
3 4 ln

This equation is trivial at O(B01 ), demonstrating the Requiring


nature of the UV cuscuton limit which allows for arbi- n6
trary expansion (or field) history at this order. However, cs t n4 , (67)
3/2  2
at O(B0 ) we have dynamical field equations. 1
H 2 (ln )2 ln t n4 , (68)
For W () to be subdominant at early times, it should
drop (or grow more slowly than ln2 ) at large . If as t 0, implies n > 6.
we assume a power-law asymptotic fall-off, i.e. W () = Finally,
W0 n with n 0, Eq. (65) has a power-law asymptotic
solution: H 1
, as t 0. (69)
H2 t(ln t)2
  1
W0 n4
(t) = . (66) Q.E.D.
(n 4)t

You might also like