You are on page 1of 4

Volume 117B, number 3,4 PHYSICS LETTERS 11 November 1982

DYNAMICS OF PHASE TRANSITION IN THE NEW INFLATIONARY UNIVERSE SCENARIO


AND GENERATION OF PERTURBATIONS

A.A. STAROBINSKY
The Landau Institute for Theoretical Physics, The Academy of Sciences, Moscow, 117334, USSR

Received 13 July 1982

Dynamics of non-equilibrium phase transition in the early universe is investigated. The transition is triggered by vacuum
fluctuations of a Higgs scalar field which determine the duration of an intermediate inflationary stage and the amplitude of
adiabatic perturbations. This amplitude ranges from g2 to one and more depending on scale that presents a serious problem
for the inflationary scenario.

It becomes clear now that many different proper- bubbles and the whole presently observed part of the
ties of the observed universe like its homogeneity, iso- universe was, roughly speaking, inside one bubble.
tropy, existence of perturbations sufficient for galaxy The aim of the present paper is to develop a theory
formation etc. may be explained by assuming only one of such a type of scenario and to calculate two most
hypothesis, namely that the early universe went through important quantities: the duration of the de Sitter (in-
the stage of exponential expansion - the de Sitter stage flationary) stage and the amplitude of adiabatic pertur-
(in ref. [1 ] this stage was called "inflationary"). bations generated during the phase transition. It will
Two reasonable ways to produce such a stage are be shown that the phase transition in this scenario pro-
known. The first one is connected with the account ceeds mainly due to zero-temperature vacuum fluctua-
of one-loop quantum-gravitational corrections to the tions of a Higgs scalar field significantly distorted by
Einstein equations. The corresponding cosmological the curvature of the de Sitter space-time. So, gravita-
scenario was constructed in ref. [2]. The other possi- tional effects play an extremely important role here.
bility is due to phase transitions which can take place Following previous numerous investigations, we
in the early universe. The inflationary scenario based shall describe a phase transition by a simple U(1) Higgs
on the phase transition of a scalar Higgs field was first model which can be easily generalized to more complex
considered by Guth [1 ]. However, it appears impossi- cases including, e.g. the SU(5) model by changing nu-
ble to preserve isotropy and homogeneity after the merical values of coefficients. In doing so, it is usually
phase transition following this "old" inflationary sce- assumed that the variation of a multi-component scalar
nario. This takes place because, according to the "old" Higgs field in the course of a phase transition can be
scenario, the de Sitter phase decays into bubbles with represented by some single "master" component [e.g.,
the most part of matter energy density being concen- q~a = ~(1, 1, 1, - ~ , - ~ ) i n the case of SU(5) breaking
trated in the bubbles' walls. As a result, further ther- to SU(3) X SU(2) U(1)] - the point that deserves
realization of matter does not occur that was already special consideration in cases when fluctuations are
pointed out by Guth [1]. important. However, we shall follow this assumption.
Recently a new inflationary scenario with a phase The gauge coupling constant g2/4n is assumed to be
transition was proposed by Linde [3] (see also refs. small and the scalar coupling constant X0 is supposed
[ 4 - 6 ] ) . The phase transition in the new scenario is of to be ~ g 4 .
Coleman-Weinberg type [7J, the matter energy den- To obtain the effection equation for the scalar field
sity is approximately uniformly distributed inside which governs the dynamics of the phase transition it

0 031-9163/82/0000-0000/$ 02.75 1982 North-Holland 175


Volume 117B, number 3,4 PHYSICS LETTERS 11 November 1982

is necessary to sum all one-loop contributions from M2 =_m 2+12H2~


gauge vector bosons with propagators inside the loops
being the functionals of the external classical gravita-
+ (3g2H2/4rg2) [ln(H/gO0 ) + 7 _ 3'] + 0(g4) (5)
tional field. The latter should be accounted for com- The Coleman-Weinberg condition in flat space-time
pletely. The renormalized effective action V[O, gik] is m 2 = k 0 = 0. The necessary condition for the new
for arbitrary a(t) [where a(t) is the scale factor of an inflationary scenario to occur is different:
isotropic homogeneous cosmological model] and O(t)
IM2I ~ H 2 . (6)
can be obtained and expressed as a mode sum and will
be discussed elsewhere. The most interesting case for Generalizing to more complex models and taking
us is a(t) = a0 exp(Ht) - the de Sitter metric (so R/k = also the thermal contribution from vector bosons into
-3H28ik), ]qb] ,~ m a x ( g O 2, OH). Then: account, we have
- 6 P / 6 0 = (3gZO/167r 2) ((g202 + 2 H 2) [~(3 + v) A(V10 + M 2 0 ) +Bg2T20 - C~.O 3 = 0 (7)

+ ~(~ - v) + 21n(H/gO0) ] + ~ H 2} at the de Sitter stage where


+ (3/8~2)g2 ff]Oln [max(H, glOl)/gO0] + .... (1) X = X(O) = -X 0 + (3g 4/87r 2) In [gO 0/max(H, glO[)]
where qJ is the logarithmic derivative of the ['-function, ~g4 >0,
v = ( -g202/H2)l/2 (fi = c = 1 is assumed throughout
this paper). for lOt "~ O 0 ; m 2 =Dg202;A = C = D = 1,B = ~- for
Kinetic O-terms are unimportant. The first term dif- U(1);A = ~ , B = ~-, 7s C = -Tg-,
lSVS D = 2s
~- for SU(5). Here
fers from the result of Shore [8] only by a term = the temperature T ~ a -1 oc exp(_Ht); H 2 =
const. X g2H20 that corresponds to a different nor- nGCk(O) O4/6; in particular, H (GeV) = 4 X 1010 (mw/
malization point of ~ in eq. (4) and is also unimpor- 1015 GeV) 2 for SU(5)..Eq. (7) is also valid for space-
tant. dependent if IO,iO't[ "~g204.
At H "~ gO: The de Sitter stage begins when T "- m w >>H. At
this time interaction is small and the last term in eq.
--6r/6qb = (3g20/87r 2) (g202 In IO/Ool (7) could be omitted. We get:
+ 2H21nlO/Ool + ~-H2 + ...). (2)
= (2rr)-3/zf d3k [Ok(t) exp(ikr) + c.c.],
The first term is just the derivative of the C o l e m a n -
Weinberg potential [7]. At H > gO: ~k + 3Hk +{[( k2 +k2)/a2] e x p ( - Z H t ) +M2) O k = 0,
(8)
- 6 p / 8 0 = (3g2OH2 /4n 2) [ln(H/gOo) + ~ - 3'1 where k 2 = Ikl 2, k g = Bg2a2T 2 = const. This equation
3
+ ( 3 g 4 0 3 / 8 n2) [ln(H/gOo) + ~- - 3'] + .... (3) is easily solved (see, e.g. ref. [9]) and the regularized
expression for (O 2) can be obtained under assumption
where 3' is the Euler constant. So, 8 P / 8 0 is the analytic that before the phase transition there was a complete
at = 0. The resulting equation for has the form: thermal equilibrium:
I-]O - m 2 0 - ~RO + ~k003 = a r / a o , (4) (O 2) = A T 2 / 12 + A 3/2H2 /8rr ~/Bg + (AH 3 t/4rr 2) 0 (t)
where m 2 = ko O2, so = O 0 is a solution at R = 0. + O(H 2 In (gOo/H)), (9)
We assume further that H "~g200, so ln(gOo/H)
is large. Then we obtain a simple prescription that where O(t) = 1, t > 0 and O(t) = 0, t < 0, and t = 0 is
(with a logarithmic accuracy) curvature effects on the chosen to correspond the moment when ko/a =
vector-boson loop can be accounted for by replacing H(T ~H/g). Also t ~H/IM21 is assumed for simplicity.
ln(IOI/qSo) in the Coleman-Weinberg effective poten- The first and second terms in eq. (9) are due to thermal
tial with In[max(H/g, [OI)/O0] and by introducing a fluctuations, the third term describes zero-temperature
new mass fluctuations. If had non-thermal fluctuations at scales
larger than the horizon (H - 1 ) at the beginning of the

176
Volume 117B, number 3,4 PHYSICS LETTERS 11 November 1982

de Sitter stage, then (qb2) gains additional constant estimate [ l 0]. Inspection of other cases shows that eq.
contribution depending on initial conditions for the (12) presents an upper bound on the duration of the
scalar field at earlier times. The third term becomes de Sitter stage (in caseM 2 > g 2 H 2 , t O means the time
the greatest one if l i t ~ g - 1 . After that time all tem- of rolling down from the top of the potential = the
perature effects can be neglected but the situation time of the exponential expansion of one bubble). In
does not resemble the zero-temperature behaviour in the case o f SU(5), eq. (12) givesHt 0 ~ 2 - 3 - too
flat space-time because space-time curvature changes small. So our conclusion is that the phase transition
equations of motion. At t ~ H / I M 2 I >>H -1, (@2) in the standard SU(5) model with m w ~ 1015 GeV
-+ const = 3AH4/87r2M 2 i f M 2 > 0 and (cb2) cannot give rise to the "new" inflationary scenario.
( H 4 / I M 2 I ) exp(21M21t/3H) i f M 2 < 0 :~1. Com- In a number of cases it is possible to go beyond the
paring these results with ref. [6], we see that in this one-scalar-loop approximation and account for the
paper fluctuations were underestimated at least by interaction completely in the infrared region. Consider,
two orders o f magnitude that changes final results e.g., the caseM 2 < 0, x/-XH 2 ,~ IMZl , ~ H 2. Then the
completely. interaction is not yet important at Ht ~ H2/IM21, but
But a t i l t ~ 1/g 2 >> 1, when (4~2) ~ H2 /~/~ ~ H2/g 2, since this time the contribution of momenta k ~ aH
self-interaction becomes significant. As a first step, let to (~2) is negligible, So one can omit spatial and sec-
us try the one-scalar-loop approximation though we ond time derivatives in eq. (7) after that this equation
have no small parameter of 1IN type in this case, so can be solved exactly even if cb is an operator (non-
we can obtain only qualitatively correct results. In commutativity o f ~ and is ~ e x p ( - 3 H t ) and can be
this approximation, we have eq. (8) for k modified neglected):
by the additional term 3C~qbk(~2) in the right-hand
(r, t)
side, and <~2) = ((27r)-3 f d3k [~kl2)reg. The main
contribution to (~2) is made by the infra-red region ~l(r) exp(lM2lt/3H)
k 0 ~ k ~ aH where renormalization terms are small {1 - (C'h/A {M21) qb~(r) [exp(2 {M21t[3I-I) - 1] } 1/2
compared to classical ones, so (q~2) can be correctly
approximated by (1/2r2)f0k =all dk Iq~kl2. Also small (13)
in this region are the second derivatives in [] ~k as Here ~l(r) is an initial condition for at some t
compared to the 3Hob k term. Then the following H/IM21 - some realization of quantum fluctuations
equation for u = (~2) can be obtained: of a free scalar field. It is a stochastic function (or an
operator) which distribution is gaussian in the first ap-
= - ( 2 M 2 / 3 H ) u + (2CX/AH)u 2 +AH3/4rr 2 , (10)
proximation with
with the initial condition u = u 0 at t = 0. I f u 0
<@l(r)) = o ,
, H2/X, then one can safely put u 0 = 0. Let, in par-
ticular, IM2[ ~ x / ~ H 2, then the linear term in eq. (qb2(r)) = 3dlq418rr21M2l +A312H2/87rx/rBg. (14)
(10) is unimportant and
The first term is due to vacuum fluctuations, the sec-
u = [AH2/2rr(2CX) 1/2] tg((CX/2)l/2Ht/rr). (1 1)
ond - due to thermal ones. The de Sitter stage ends
Returning back to eq. (7), one can estimate that the when formally #; -+ oo (really there is a cutoff at
second time derivative becomes important again when ~ H / x / ~ when the second time derivatives comes into
u ~ H2/~ (spatial derivatives are still negligible). At play). This takes place at
that moment the de Sitter stage ends. So the duration
to(r ) = (3H/21M21) ln[AIM21/C'acb2(r)], (15)
of the de Sitter stage t o is:
e.g., at different time at different points of space [eq.
Ht 0 = rr2/(2C3,) 1/2 ~ 1/g 2 . (12)
(13) and (15) are not applicable if b2 ~ H 4 / I M 2 1 ] .
This result is in a qualitative agreement with Linde's Typically, to(r ) ~ (3H/2 IM21) ln(M4/M/4)
>>3H/2 IM21, but fluctuations of ~ are not small
1 These results were also independently obtained recently [(~4> = 3((~2))2] so the values of to(r ) can differ very
by Linde [ 10] and Vilenkin and Ford [ l I ]. significantly. The same idea can be used in other cases

177
Volume 117B, number 3,4 PHYSICS LETTERS 11 November 1982

assuming that a sufficiently large fluctuation arises in 10 -1 cm at the present time). Eq. (18) is valid if
the infrared region. 1 ~ l n ( k l / k ) ~ min(H2/[M21, 1/x/~) and Lfhkl ~ 1.
The stochastic function to(r ) is directly related to So the spectrum of adiabatic perturbations, in contrast
the amplitude of adiabatic perturbations generated in to that of gravitational waves [ 13], is not exactly flat
the course of the phase transition. Let our clocks be (e.g., of Zeldovich type) but grows slowly to large
synchronized at the moment when the de Sitter stage scales.
begins. The de Sitter stage ends at t = to(r). We have To obtain 5h <~ 10 - 4 at a scale ~ 1028 cm [ln(kl/k )
from eq. (13): 67] one needs an exceedingly small CX ~< 10 -10.
So our second conclusion is that the C o l e m a n -
b2(r, t) = 3AH/2CX[to(r ) - tl (16) Weinberg potential seems to be inappropriate for the
at 1 ~ H [ t o ( r ) - t] ~ H 2 / I M 2 1 . After that, q5 a n d a construction of an inflationary scenario. One should
are functions of t - to(r ). So the s p a c e - t i m e metric either change the form of the potential or look for
changes in the following way: some realistic damping mechanism.

ds 2 = d t 2 - exp(2Ht) dl 2
The author is thankful to J.M. Bardeen, G.W.
= d t 2 - exp[2Hto(r)] e x p { 2 H [ t - to(r)] } d/2 Gibbons, A.H. Guth, J.B. Hartle, S.W. Hawking, A.D.
Linde, P.J. Steinhardt and M.S. Turner for many valu-
d t 2 - exp[2Hto(r)]a 2 [t - t0(t)] d/2 able discussions.

-+ d t 2 - exp[2Ht0(r)] t dl 2 , (17)
References
at t >> t 0. We obtain a particular type of the so-called
quasi-isotropic metric [12]. Next order corrections to [1] A.H. Guth, Phys. Rev. D23 (1981) 347.
the metric are proportional to Vto(r ) and small as [2] A.A. Starobinsky, Phys. Lett. 91B (1980) 99.
[3] A.D. Linde, Phys. Lett. 108B (1982) 389.
long as X >> t. Therefore, though locally the metric
[4] S.W. Hawking and I.G. Moss, Phys. Lett. ll0B (1982)
(17) looks like an isotropic one, globally it is very 35.
anisotropic and inhomogeneous. [5] A. Albrecht and P.J. Steinhardt, Phys. Rev. Lett. 48
F o r small wavelengths which are inside the present (1982) 1220.
horizon, we can decompose q5 = qb(r, t) + ~k [~k [61 A. Albrecht, P.J. Steinhardt, M.S. Turner and F. Wilczek,
Phys. Rev. Lett. 48 (1982) 1437.
e x p ( i k r ) + c.c.] where cb(r, t) contains only large [71 S. Coleman and E. Weinberg, Phys. Rev. D7 (1973)
wavelengths and satisfies eq. (7) and qbk satisfies eq. 1888.
(8) with M 2 _+M 2 _ 13C X/A) cb2 (r, t). Using eqs. (16) [8] G.M. Shore, Ann. Phys. 128 (1980) 376.
and (17), the final result for the Fourier components [9] T.S. Bunch and P.C.W. Davies, Proc. R. Soc. A360
o f adiabatic perturbations can be found: (1978) 117.
[ 10] A.D. Linde, Lebedev Physical Instiute preprint (1982).
(6p/p)kLX= t ~ 5 h k = (2H~to) k [ 11 ] A.V. Vilenkin and L.H. Ford, Tufts University preprint
TUTP - 82 - 7 (1982).
= (!~CX)1/2 ln3/2(kl/k)/k3/2 , (18) [ 12] L.D. Landau and E.M. Lifshitz, The classical theory of
fields (Pergamon, Oxford, 1975).
where k 1 = a H at the end of the de Sitter stage (a/k 1 [13] A.A. Starobinsky, Pis'ma Zh. Eksp. Teor. Fiz. 30 (1979)
719.

178

You might also like