You are on page 1of 4

Small Cosmological Constant from the QCD Trace Anomaly?

Ralf Schu¨tzhold
*
Department of Physics and Astronomy, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z1
Institut fu¨r Theoretische Physik, Technische Universita¨t Dresden, D-01062 Dresden, Germany
(Received 17 April 2002; published 5 August 2002)
According to recent astrophysical observations the large scale mean pressure of our present Universe is
negative suggesting a positive cosmological constant-like term. The issue of whether nonperturbative
effects of self-interacting quantum fields in curved space-times may yield a significant contribution is
addressed. Focusing on the trace anomaly of quantum chromodynamics, a preliminary estimate of the
expected order of magnitude yields a remarkable coincidence with the empirical data, indicating the
potential relevance of this effect.
DOI: 10.1103/PhysRevLett.89.081302 PACS numbers: 04.62.+v, 12.38.Aw, 12.38.Lg, 98.80.Es
Recent measurements of the cosmic microwave back-
ground [1] suggest that the large scale structure of our
Universe is quite accurately described by the conformally
flat Friedmann-Robertson-Walker (FRW) metric
Js
2
÷ Ji
2

2
(i)Jr
2
÷
2
(¬)(J¬
2
Jr
2
). (1)
with i denoting the proper (comoving wristwatch) time and
¬ the conformal time, respectively. The temporal variation
of the scale factor inducing the cosmological redshift is
represented by the Hubble parameter
H ÷
1

J
Ji
- 10
10
yr
1
. (2)
Inserting the FRW metric in Eq. (1) into the Einstein
equations (with the cosmological constant ¬)
R
µ:

1
2
g
µ:
R ÷ (8rG
N
(
^
TT
µ:
) g
µ:
¬)
ren
. (3)
the aforementioned observations and supernova [2] data
consistently yield the following conclusions: the 00 com-
ponent of the right-hand side (r.h.s.) of Eq. (3) equals (at
least approximately) the critical density n - n
crit
and the
spatial || components—associated with the pressure µ—
are negative: µ¡n
crit
- 2¡3. As a result the Universe is
presently undergoing an accelerated expansion as approxi-
mately described by the de Sitter metric (i) ÷ exp¦Hi¦.
A negative pressure together with a positive energy
density necessarily implies a nonvanishing trace of the
energy-momentum tensor T
µ:
. Since T
µ:
can be derived
via the variation of the action Awith respect to the metric
g
µ:
, i.e., T
µ:
÷ 2(g)
1¡2
uA¡ug
µ:
, its trace corre-
sponds to the change of A under the conformal trans-
formations g
µ:
(x) -
2
(x)g
µ:
(x) by virtue of Euler’s
law: T
µ
µ
÷ (g)
1¡2
uA¡u; see, e.g., [3].
Let us focus on the contribution of the SU(3)-color
gauge field theory of QCD to the r.h.s. of the Einstein
equations (3) in the following. Its dynamics are governed
by the well-known Lagrangian density
L ÷
1
4
G
o
µ:
G
µ:
o
+

!!(|~
µ
·
µ
+g T
o
~
µ
A
o
µ
m)!.
(4)
with G
o
µ:
÷ ·
µ
A
o
:
·
:
A
o
µ
+g]
o
lc
A
l
µ
A
c
:
being the
gluonic field strength tensor. Here ]
o
lc
denote the SU(3)
structure constants, T
o
denote its fundamental generators,
and g is the strong coupling. For simplicity we drop the
ghost fields as well as the gauge fixing terms and consider
only one single flavor, i.e., quark species !. The remaining
electroweak sector of the standard model is discussed at the
end of this Letter.
On the classical level all gauge field theories as de-
scribed by Eq. (4) are conformally invariant (for m ÷ 0).
According to the above arguments this feature implies a
vanishing trace of the classical energy-momentum tensor.
Turning to the quantum field theoretical description the
situation becomes more complicated. In the first place, the
naive expectation value of the operator-valued energy-
momentum tensor diverges due to the infinite zero-point
energy. In order to renormalize this singularity by an
appropriate counterterm one has to interpret the cosmo-
logical constant ¬ in Eq. (3) as a bare quantity [3].
After such a minimal subtraction procedure the trace of
the renormalized expectation value of the energy-momen-
tum tensor (
^
TT
µ:
)
ren
vanishes (again assuming m ÷ 0) for
free (g ÷ 0) fields in flat ( ÷ 1) space-times—but not in
the general case. This phenomenon is called the trace
anomaly and goes along with the dynamical breaking of
the conformal invariance of the classical theory in Eq. (4).
It has been calculated for two limiting cases: first, for self-
interacting quantum fields in flat ( ÷ 1) space-times [4]
and, second, for free (g ÷ 0) fields in curved space-times
[3]. In the first case one obtains [4]
(
^
TT
µ
µ
)
ren
÷
E(g)
2g
(
^
GG
o
µ:
^
GG
µ:
o
)
ren
+(1 +~
m
) m(

!!
^
!!)
ren
. (5)
The Callan-Symanzik or Gell-Mann–Low E function
describes the scale dependence E ÷ µ·g¡·µ of the
renormalized coupling g(µ) and reflects the dynamical
VOLUME 89, NUMBER 8 P HYS I CAL RE VI E W L E T T E RS 19 AUGUST 2002
081302-1 0031-9007¡02¡89(8)¡081302(4)$20.00  2002 The American Physical Society 081302-1
breaking of the conformal invariance of the classical theory
(dimensional transmutation). Similarly, the ~
m
function
corresponds to the running of the renormalized mass.
The expectation values (
^
GG
o
µ:
^
GG
µ:
o
)
ren
and (

!!
^
!!)
ren
occurring
in Eq. (5) represent the so-called gluonic and quark con-
densates, respectively; see, e.g., [5]. These inherently non-
perturbative quantities again reflect the dynamical
breaking of the classical scale invariance. Both are of great
experimental relevance and their values have been con-
firmed within several contexts; see, e.g., [5]. Since the
symmetry breaking scale
QCD
is (for m ÷ 0) the only
scale in the theory (4), it yields (
^
GG
o
µ:
^
GG
µ:
o
)
ren
÷ O(
4
QCD
)
and (

!!
^
!!)
ren
÷ O(
3
QCD
) —at least for m <
QCD
. As
is well known, the E function occurring in Eq. (5) can
be calculated within the framework of perturbation the-
ory, and it turns out to be negative. Consequently, it is
now commonly accepted (cf. [5]) that the QCD trace
anomaly gives rise to a negative energy density (since
(
^
GG
o
µ:
^
GG
µ:
o
)
ren
> 0 and (

!!
^
!!)
ren
<0) of the QCD vacuum
in the Minkowski space-time.
However, such a huge amount of negative energy density
of order O(
4
QCD
) blatantly contravenes our observations.
This drastic and global violation of the (weak and domi-
nant) energy conditions (see, e.g., [3]) in the Minkowski
space-time goes along with a fundamental contradiction if
one includes gravity since the r.h.s. of the Einstein equa-
tions (3) associated with a flat space-time vanishes.
Consequently, regarding the Einstein equations (3), one
is led to absorb the aforementioned energy density by
renormalizing the cosmological constant ¬, in complete
analogy with the case of the zero-point energy (which
determines the divergent part of ¬ only). In the same
manner as one adjusts the mass counterterm in the self-
energy renormalization of the electron, for example, one
has to fix the bare cosmological constant by demanding
that the r.h.s. of the Einstein equations (3) vanishes for the
Minkowski vacuum
(8rG
N
(
^
TT
µ:
) g
µ:
¬)
Minkowski vacuum
ren
÷ 0. (6)
On the other hand, a nontrivial geometry of the space-
time may also induce a nonvanishing trace—even for free
fields (second limiting case g ÷ 0). In this case (
^
TT
µ
µ
)
ren
is
given by the sum of a bilinear form of the curvature tensor
(such as R
µ:
R
µ:
or R
2
) and second derivatives of it
(ᮀR); cf. [3]. For the free QCD field (with g ÷ m ÷ 0)
within the de Sitter space-time (i) ÷ exp¦Hi¦, for ex-
ample, one finds ( h ÷ c ÷ 1)
(
^
TT
µ
µ
)
ren
÷
281
120r
2
H
4
. (7)
In contrast to the contribution in Eq. (5), there is no reason
to absorb this term by renormalization of ¬. In view of
its potentially space-time dependent character such a pro-
cedure would be rather strange. However, here the associ-
ated energy density is far too small to explain the
observations [1,2].
In summary, the (renormalized) expectation value of the
energy-momentum tensor acquires an anomalous trace for
self-interacting quantum fields in flat space-times (5) on
the one hand, as well as for free fields in curved space-
times (7) on the other hand. However, both effects taken
alone are not capable of explaining the negative pressure as
suggested by the observations [1,2]. But this is just what
one might expect, since realistic investigations have to
involve both contributions simultaneously, i.e., the (non-
perturbative) effects of self-interacting fields in curved
space-times. A rigorous derivation of the renormalized
expectation value of the energy-momentum tensor for
this scenario appears to be rather involved and is not the
aim of the present Letter. Here we give just a preliminary
estimate of the expected order of magnitude of the effect.
To this end we employ an adiabatic approximation (cf. [3])
by exploiting the huge difference of the involved time
scales. The cosmic evolution—governed by H—is ex-
tremely slow compared to the typical fluctuations of the
quantum field as determined by
QCD
. Consequently the
adiabatic approximation is an expansion in the small pa-
rameter H¡
QCD
÷ O(10
40
). So the zeroth-order term is
the pure Minkowski (flat space-time) contribution, whereas
the first-order term represents the lowest correction in-
duced by the cosmic expansion.
In order to calculate the renormalized expectation value
of the energy-momentum tensor it is essential to specify
the correct vacuum state associated with our expanding
Universe; cf. [3]. We adopt the Schro¨dinger picture
J
Ji
í) ÷ |
^
HH
FRW
(i)í). (8)
where
^
HH
FRW
(i) denotes the Hamilton operator, i.e., the
generator of the time evolution, with respect to the coor-
dinates (i. r) in Eq. (1). Within the adiabatic approxima-
tion, the explicitly time-dependent Hamiltonian
^
HH
FRW
(i)
of an expanding universe can be related to the (time-
independent) Minkowski Hamiltonian
^
HH
Min
via
^
HH
FRW
(i) ÷ exp¦|(i)
^
SS¦
^
HH
Min
exp¦+|(i)
^
SS¦. (9)
with
^
SS being the generator for the spatial conformal trans-
formations g
|]
-
2
g
|]
in the Schro¨dinger picture. In
terms of a dynamically scaled state defined via í
~
) ÷
exp¦+|(i)
^
SS¦í), the Schro¨dinger equation reads
J
Ji
í
~
) ÷ |(
^
HH
Min

_

^
SS)í
~
). (10)
Treating
^
HH
1
÷
_

^
SS ÷ O(H) as a perturbation and
switching to the interaction representation
^
SS(i) ÷
exp¦+|
^
HH
Min

^
SS exp¦|
^
HH
Min
i¦ we may solve the above
equation in linear response, i.e., first-order adiabatic
expansion
VOLUME 89, NUMBER 8 P HYS I CAL RE VI E W L E T T E RS 19 AUGUST 2002
081302-2 081302-2
í
~
) ÷ í
~

in
) +|
Z
0
«
Ji
_
(i)
^
SS(i)í
~

in
) +O(H
2
). (11)
If we assume (i) ÷
0
+exp¦Hi¦, it is reasonable to
take the Minkowski vacuum í0
Min
) with
^
HH
Min
í0
Min
) ÷ 0
as the initial condition í
~

in
) ÷ í0
Min
). The remaining time
integration yields the (retarded) inverse of the Minkowski
Hamiltonian
^
HH
1
Min
and hence we arrive at
í0
FRW
) ÷ í0
Min
) +H
^
HH
1
Min
^
SSí0
Min
) +O(H
2
). (12)
Therefore the adiabatic QCDvacuumí0
FRW
) of an expand-
ing universe is not the instantaneous ground state í0
Min
) of
^
HH
FRW
(i) or
^
HH
Min
—it acquires corrections already in the
first order of the adiabatic expansion. Instead it is an
(approximate) eigenstate of the corrected Hamiltonian
^
HH
Min

_

^
SS (see, e.g., [3] for free fields).
The remaining question is, of course, whether the first-
order correction H
^
HH
1
Min
^
SSí0
Min
) to the vacuum state entails
a first-order correction to the expectation value of
^
TT
µ:
. In
order to illustrate this point let us consider the simple
example of a time-dependent harmonic oscillator
^
HH(i) ÷
w
2(i)
|
^
PP
2
+
2
(i)
^
QQ
2
|. (13)
where w corresponds to
QCD
. In examining the question
of whether the dynamical scale symmetry breakdown in
QCD can be modeled by such a simple quadratic potential
one might consider the CP(N 1) or the O(N) a-models
[6]. These strongly interacting theories reproduce several
features of QCD, such as dynamical scale symmetry break-
down. They can be solved in the large N limit and in the
leading order they effectively behave as massive free fields.
After a normal mode decomposition one therefore indeed
obtains terms like the one above. For the Hamiltonian in
Eq. (13) a change of the scale factor (i) as in Eq. (9) can
simply be generated by the squeezing operator
^
SS ÷
1
4
¦
^
PP.
^
QQ¦ ÷
|
4
|( ^ oo
|
)
2
( ^ oo)
2
|. (14)
Consequently, the expectation values of operators such as
^
QQ
2
or
^
PP
2
do not acquire a first-order correction. This result
can be transferred directly to free quantum fields: pitching
on a particular normal mode with the wavelength k the
conformal charge
^
SS again acts like a squeezing operator
^
SS - |( ^ oo
|
k
^ oo
|
k
^ oo
k
^ oo
k
). As a result there is no first-order
correction to the expectation value of
^
TT
µ:
for free (linear)
fields.
However, if we leave the free-field sector and take
interactions into account the situation may change: let us
consider the interaction Hamiltonian
^
HH
int
(i) ÷ g
Z
i
«
Ji
/
G
ret
(i i
/

^
QQ
2
(i).
^
QQ
2
(i
/
)¦. (15)
where
^
QQ(i) ÷
^
QQcos(wi) +
^
PP sin(wi) denotes the unper-
turbed time-dependent operator in the interaction picture.
The retarded propagator G
ret
(i i
/
) encodes the dynamics
of an intermediate (interaction) degree of freedom which
has been integrated out. Again such a term can be moti-
vated by the CP(N 1) models: in the large N limit these
strongly interacting massless theories effectively transform
into massive fields obeying weak (i.e., next-to-leading
order in 1¡N) long-range four-point interactions; cf. [6].
For general Green functions G
ret
(i i
/
), the operator in
Eq. (15) does entail a first-order correction. This can be
most easily verified by assuming g < 1 which allows for a
perturbative treatment.
In view of these considerations one might expect a first-
order contribution to (
^
TT
µ:
)
ren
to be possible in the case of
QCD: since the classical as well as the free quantum field
in Eq. (4) are (to first order in H and for m ÷ 0) confor-
mally invariant, their solutions would simply be scaled
during the expansion of the Universe—like the redshift
of the photon field. (This would actually happen if the
Universe were to expand very rapidly H >
QCD
.)
However, the strong self-interaction on the quantum level
breaks the conformal invariance and introduces a fixed
scale
QCD
leading to a positive pressure given by
Eq. (5). Within an expanding universe the balance of these
two tendencies, i.e., following the expansion on the one
hand and retaining the scale on the other hand, leads to a
displaced vacuum state (12).
Let us assume that a part of the positive vacuum pressure
in Eq. (5) can be explained by relatively localized (non-
perturbative) vacuum fluctuations (e.g., instantons [7] or
oscillons [8]) which repel each other (at least in average;
cf. [7]). Let us further assume that the dynamical break-
down of the scale symmetry is basically encoded by these
(nonperturbative) vacuum fluctuations, whereas their (re-
pulsive) interactions are adequately described by the free
(perturbative) and thus conformally invariant field equa-
tions. In this case their solutions would simply be scaled
during the cosmic expansion in contrast to the nonpertur-
bative fluctuations which retain their scale and hence are
not affected. Within an expanding universe, then, every
vacuum fluctuation ‘‘sees’’ all other vacuum fluctuations
‘‘redshifted,’’ i.e., their repulsion acquires a correction
proportional to HR, where R denotes their (mean) distance
(cf. [7]). Accordingly, the positive vacuum pressure in
Eq. (5) gets diminished by an amount of first order in H.
Based on this intuitive picture it appears plausible to
admit a correction to the expectation value of the energy-
momentum tensor within the FRW vacuum in Eq. (12)
which is linear in H. After the renormalization described
in Eq. (6), i.e., the subtraction of the Minkowski contribu-
tion, we therefore obtain
(
^
TT
µ:
)
FRW
ren
÷ H(
^
TT
µ:
^
HH
1
Min
^
SS)
Min
ren
+H:c: +O(H
2
)
÷ O(H
3
QCD
). (16)
Let us estimate the associated order of magnitude:
although
QCD
depends on the renormalization scheme
VOLUME 89, NUMBER 8 P HYS I CAL RE VI E W L E T T E RS 19 AUGUST 2002
081302-3 081302-3
we may fix it approximately via
QCD
÷ O(10
8
eV) ÷
O(10
14
m
1
). The masses of the light quarks which
dominantly couple to the gluonic field are roughly of a
similar order of magnitude. The Hubble expansion parame-
ter H is about 10
26
m
1
. Inserting the above values we
finally arrive at (
^
TT
µ:
)
FRW
ren
÷ O(10
16
m
4
) or (
^
TT
µ:
)
FRW
ren
÷
O(10
29
g cm
3
). By inspection one finds that the deduced
order of magnitude nicely fits the empirical data n
crit
-
10
29
g cm
3
. In view of the huge difference of the in-
volved scales [H¡
QCD
÷ O(10
40
)] this remarkable co-
incidence seems to be almost too good to be just an
accident. At least it indicates the potential relevance of
the effect described in the present Letter with regard to the
interpretation of the astrophysical data [1,2].
It should be mentioned here that a pressure induced by
the expansion of our Universe with µ · H generates a
cosmic evolution which differs from that with a true cos-
mological constant µ ÷ const: by inserting the FRW met-
ric (1) into the Einstein equations (3) one obtains the
Friedmann equation 3H
2
÷ 8rG
N
n. Furthermore the
Einstein equations imply \
µ
(
^
TT
µ:
)
ren
÷ 0; i.e., Jn¡Ji ÷
3(n +µ)H. Combining these two equalities and specify-
ing the pressure µ one may determine the time evolution of
our Universe. Unfortunately the presently available data
(such as the Hubble parameter or the age of the Universe)
are not precise enough to distinguish the two cases (µ · H
and µ ÷ const).
Of course one may ask whether the remaining electro-
weak sector of the standard model generates similar con-
tributions: typically (see, e.g., [5]) nonperturbative effects
(such as (
^
GG
o
µ:
^
GG
µ:
o
)
ren
and (

!!
^
!!)
ren
) display a dependence
on the coupling of exp¦8r
2
¡g
2
¦ ÷ exp¦2r¡n¦. The
scale of the dynamical symmetry breaking
QCD
obeys
a similar nonanalytical dependence on the coupling g.
Inserting n
QED
- 1¡137 into the above expression one
obtains a suppression by an order of magnitude of
10
370
. Hence the contributions arising from the dynami-
cal breaking of scale invariance can safely be neglected in
this case.
The remaining explicit breaking of the scale symmetry
induced by the Higgs field of course also generates con-
tributions to
^
TT
µ:
. However, the general structure of all
these terms is given by m
2
(
^

| ^
)
ren
and according to the
arguments after Eq. (14) they do not contribute to the first
order in H. Although the mixture of these terms caused by
interactions remains subject to further considerations, a
contribution of the electroweak sector in analogy to QCD
is not obvious.
In summary, the present Letter motivates a deeper ex-
amination of the vacuum of strongly interacting fields in
the gravitational background of our expanding Universe—
for the present epoch as well as for earlier stages; cf. [9].
These investigations might perhaps lead to a better under-
standing of some of the problems in cosmology without
necessarily invoking yet unknown low-energy fields, for
example, quintessence (see, e.g., [10]).
The author is indebted to B. Unruh and E. Zhitnitsky for
fruitful discussions and acknowledges valuable comments
and criticism from Y. Khriplovich, F. Krauss, R. Kuhn,
J. Matthews, M. Meyer-Hermann, D. Page, G. Plunien,
G. Soff, and R. Woodard. This work was supported by
the Alexander von Humboldt foundation and by the
NSERC.
*Electronic address: schuetz@physics.ubc.ca
[1] C. Pryke et al., Astrophys. J. 568, 46 (2002); A. H. Jaffe
et al., Phys. Rev. Lett. 86, 3475 (2001); A. E. Lange et al.,
Phys. Rev. D 63, 042001 (2001); C. J. Miller et al.,
Science 292, 2302 (2001); S. Padin et al., Astrophys. J.
549, L1 (2001); P. Mauskopf et al., ibid. 536, L59 (2000);
A. Melchiorri et al., ibid. 536, L63 (2000);
P. de Bernardis et al., Nature (London) 404, 955 (2000).
[2] S. Perlmutter et al., Nature (London) 391, 51 (1998);
Astrophys. J. 517, 565 (1999); A. G. Riess et al., Astron.
J. 116, 1009 (1998); P. M. Garnavich et al., Astrophys. J.
509, 74 (1998); 493, L53 (1998); B. P. Schmidt et al., ibid.
507, 46 (1998).
[3] N. D. Birrell and P. C. W. Davies, Quantum Fields in
Curved Space (Cambridge University Press, Cambridge,
England, 1982).
[4] R. Crewther, Phys. Rev. Lett. 28, 1421 (1972);
M. Chanowitz and J. Ellis, Phys. Lett. 40B, 397 (1972);
Phys. Rev. D 7, 2490 (1973); J. C. Collins, A. Duncan, and
S. D. Joglekar, Phys. Rev. D 16, 438 (1977).
[5] M. A. Shifman, Vacuum Structure and QCD Sum Rules
(Elsevier, Amsterdam, 1992).
[6] V. A. Novikov et al., Phys. Rep. 116, 103 (1984); Sov. J.
Part. Nucl. 17, 204 (1986); Fiz. Elem. Chastits At. Yadra
17, 472 (1986); S. R. Coleman, Aspects of Symmetry
(Cambridge University Press, Cambridge, England,
1985); W. A. Bardeen, B. W. Lee, and R. E. Shrock,
Phys. Rev. D 14, 985 (1976).
[7] E. V. Shuryak, Nucl. Phys. B203, 93 (1982); M. A.
Shifman, A. I. Vainshtein, and V. I. Zakharov, Nucl.
Phys. B165, 45 (1980); T. Scha¨fer and E. V. Shuryak,
Rev. Mod. Phys. 70, 323 (1998).
[8] I. L. Bogolyubskiı ˘ and V. G. Makhan’kov, Pis’ma Zh.
Eksp. Teor. Fiz. 24, 15 (1976); [JETP Lett. 24, 12
(1976)]; Pis’ma Zh. Eksp. Teor. Fiz. 25, 120 (1977);
[JETP Lett. 25, 107 (1977)]; E. J. Copeland, M. Gleiser
and H. R. Mu¨ller, Phys. Rev. D 52, 1920 (1995).
[9] S. W. Hawking, T. Hertog, and H. S. Reall, Phys. Rev. D
63, 083504 (2001); H. Salehi and Y. Bisabr, Int. J. Theor.
Phys. 39, 1241 (2000); Y. V. Novozhilov and D. V.
Vassilevich, Phys. Lett. B 220, 36 (1989); A. A.
Starobinsky, Phys. Lett. 91B, 99 (1980); M. V. Fischetti,
J. B. Hartle, and B. L. Hu, Phys. Rev. D 20, 1757 (1979).
[10] L. M. Wang et al., Astrophys. J. 530, 17 (2000); P. Brax
and J. Martin, Phys. Lett. B 468, 40 (1999); T. Chiba,
Phys. Rev. D 60, 083508 (1999); P. Binetruy, ibid. 60,
063502 (1999); I. Zlatev, L. Wang, and P. J. Steinhardt,
Phys. Rev. Lett. 82, 896 (1999); S. M. Carroll, ibid. 81,
3067 (1998); R. R. Caldwell, R. Dave, and P. J. Steinhardt,
ibid. 80, 1582 (1998).
VOLUME 89, NUMBER 8 P HYS I CAL RE VI E W L E T T E RS 19 AUGUST 2002
081302-4 081302-4

[3]. its trace corresponds to the change of A under the conformal transformations g …x† ! 2 …x†g …x† by virtue of Euler’s  law: T ˆ ÿ …ÿg†ÿ1=2 A= . e. 2 (3) the aforementioned observations and supernova [2] data consistently yield the following conclusions: the 00 component of the right-hand side (r. of the Einstein equations (3) in the following. i. Since T can be derived via the variation of the action A with respect to the metric g . The temporal variation of the scale factor inducing the cosmological redshift is represented by the Hubble parameter Hˆ 1 d  10ÿ10 yrÿ1 : dt (2) Inserting the FRW metric in Eq.) of Eq. Its dynamics are governed by the well-known Lagrangian density 081302-1 0031-9007=02=89(8)=081302(4)$20. Let us focus on the contribution of the SU(3)-color gauge field theory of QCD to the r. respectively.e.h. T ˆ 2…ÿg†ÿ1=2 A=g .00 The Callan-Symanzik or Gell-Mann–Low .g.s..…g† ^ a ^  hG  G a iren ‡ …1 ‡ m † m h  ^ iren : (5) 2g with t denoting the proper (comoving wristwatch) time and  the conformal time.s. As a result the Universe is presently undergoing an accelerated expansion as approximately described by the de Sitter metric …t† ˆ expfHtg.. (3) equals (at least approximately) the critical density %  %crit and the spatial ii components — associated with the pressure p — are negative: p=%crit  ÿ2=3.h. A negative pressure together with a positive energy density necessarily implies a nonvanishing trace of the energy-momentum tensor T . see. (1) into the Einstein equations (with the cosmological constant ) 1 ^ R ÿ g R ˆ …8 GN hT  i ÿ g †ren .

function describes the scale dependence .

ˆ  @g=@ of the renormalized coupling g…† and reflects the dynamical  2002 The American Physical Society 081302-1 .

e. Both are of great experimental relevance and their values have been confirmed within several contexts.. the . e. NUMBER 8 PHYSICAL REVIEW LETTERS 19 AUGUST 2002 breaking of the conformal invariance of the classical theory (dimensional transmutation).g. Similarly. As and h QCD is well known. [5]. These inherently nonperturbative quantities again reflect the dynamical breaking of the classical scale invariance. [5].VOLUME 89. Since the symmetry breaking scale QCD is (for m ˆ 0) the only ^  ^ a scale in the theory (4). see.g. it yields hG a G  iren ˆ O…4 † QCD  ^ iren ˆ O…3 † — at least for m  QCD . (5) represent the so-called gluonic and quark condensates.. respectively. see. ^  ^ a The expectation values hG a G  iren and h  ^ iren occurring in Eq. the m function corresponds to the running of the renormalized mass.

e. here the associ081302-2 _ ^ ^ Treating H 1 ˆ ÿ S ˆ O…H† as a perturbation and ^ switching to the interaction representation S …t† ˆ ^ ^ Min tgS expfÿiH Min tg we may solve the above ^ expf‡iH equation in linear response. In summary.. Here we give just a preliminary estimate of the expected order of magnitude of the effect. i..2]. [5]) that the QCD trace anomaly gives rise to a negative energy density (since ^  ^ a hG a G  iren > 0 and h  ^ iren < 0) of the QCD vacuum in the Minkowski space-time. a nontrivial geometry of the spacetime may also induce a nonvanishing trace—even for free ^ fields (second limiting case g ˆ 0).g. one has to fix the bare cosmological constant by demanding that the r. However. first-order adiabatic expansion 081302-2 .e. of the Einstein equations (3) vanishes for the Minkowski vacuum ^ …8 GN hT  i ÿ g †Minkowski vacuum ˆ 0: ren (6) ated energy density is far too small to explain the observations [1.s. one finds ( ˆ c ˆ 1) h ^ hT  iren 281 ˆ H4 : 1202 (7) ^ with S being the generator for the spatial conformal trans¨ formations gij ! 2 gij in the Schrodinger picture. (1). [3]. In order to calculate the renormalized expectation value of the energy-momentum tensor it is essential to specify the correct vacuum state associated with our expanding ¨ Universe. dt (8) ^ where H FRW …t† denotes the Hamilton operator. e. it is now commonly accepted (cf. for example. The cosmic evolution— governed by H— is extremely slow compared to the typical fluctuations of the quantum field as determined by QCD . In view of its potentially space-time dependent character such a procedure would be rather strange. (5) can be calculated within the framework of perturbation theory.. with respect to the coordinates …t.e. of the Einstein equations (3) associated with a flat space-time vanishes. But this is just what one might expect. To this end we employ an adiabatic approximation (cf. Consequently the adiabatic approximation is an expansion in the small parameter H=QCD ˆ O…10ÿ40 †. and it turns out to be negative. since realistic investigations have to involve both contributions simultaneously. as well as for free fields in curved spacetimes (7) on the other hand. In the same manner as one adjusts the mass counterterm in the selfenergy renormalization of the electron. (9) On the other hand. QCD This drastic and global violation of the (weak and dominant) energy conditions (see. (5). for example. So the zeroth-order term is the pure Minkowski (flat space-time) contribution.. the generator of the time evolution. Within the adiabatic approxima^ tion. However.h. such a huge amount of negative energy density of order O…4 † blatantly contravenes our observations. [3]) by exploiting the huge difference of the involved time scales. the Schrodinger equation reads ¨ expf‡i …t†S d ~ _ ^ ~ ^ j i ˆ ÿi…H Min ÿ S †j i: dt (10) In contrast to the contribution in Eq. However. In ~ terms of a dynamically scaled state defined via j i ˆ ^ gj i.h. Consequently. cf. For the free QCD field (with g ˆ m ˆ 0) within the de Sitter space-time …t† ˆ expfHtg.2]. i. [3]) in the Minkowski space-time goes along with a fundamental contradiction if one includes gravity since the r. whereas the first-order term represents the lowest correction induced by the cosmic expansion. r† in Eq. regarding the Einstein equations (3). cf. there is no reason to absorb this term by renormalization of . In this case hT  iren is given by the sum of a bilinear form of the curvature tensor (such as R R or R2 ) and second derivatives of it ( R). We adopt the Schrodinger picture d ^ j i ˆ ÿiH FRW …t†j i. the (renormalized) expectation value of the energy-momentum tensor acquires an anomalous trace for self-interacting quantum fields in flat space-times (5) on the one hand. one is led to absorb the aforementioned energy density by renormalizing the cosmological constant . [3]. A rigorous derivation of the renormalized expectation value of the energy-momentum tensor for this scenario appears to be rather involved and is not the aim of the present Letter.s. in complete analogy with the case of the zero-point energy (which determines the divergent part of  only). both effects taken alone are not capable of explaining the negative pressure as suggested by the observations [1. the (nonperturbative) effects of self-interacting fields in curved space-times. i. Consequently. function occurring in Eq. the explicitly time-dependent Hamiltonian H FRW …t† of an expanding universe can be related to the (time^ independent) Minkowski Hamiltonian H Min via ^ ^ ^ ^ H FRW …t† ˆ expfÿi …t†S gH Min expf‡i …t†S g.

then. They can be solved in the large N limit and in the leading order they effectively behave as massive free fields. whether the first^ Min ^ order correction H H ÿ1 S j0Min i to the vacuum state entails ^ a first-order correction to the expectation value of T  . [3] for free fields). we therefore obtain ^ ^ ^ Min ^ ren hT  iFRW ˆ HhT  H ÿ1 S iMin ‡ H:c: ‡ O…H2 † ren ˆ O…H3 †: QCD (16) ^ ^ ^ where Q …t† ˆ Q cos…!t† ‡ P sin…!t† denotes the unperturbed time-dependent operator in the interaction picture. (13) a change of the scale factor …t† as in Eq. In view of these considerations one might expect a first^ order contribution to hT  iren to be possible in the case of QCD: since the classical as well as the free quantum field in Eq. Again such a term can be motivated by the CP…N ÿ 1† models: in the large N limit these strongly interacting massless theories effectively transform into massive fields obeying weak (i. For the Hamiltonian in Eq. Within an expanding universe. (5) gets diminished by an amount of first order in H. [7]). In examining the question of whether the dynamical scale symmetry breakdown in QCD can be modeled by such a simple quadratic potential one might consider the CP…N ÿ 1† or the O…N† -models [6]. In order to illustrate this point let us consider the simple example of a time-dependent harmonic oscillator ! ^2 ^ ^ ‰P ‡ 2 …t†Q 2 Š. if we leave the free-field sector and take interactions into account the situation may change: let us consider the interaction Hamiltonian Zt ^ ^ ^ H int …t† ˆ g dt0 Gret …t ÿ t0 †fQ 2 …t†. Q 2 …t0 †g. For general Green functions Gret …t ÿ t0 †. the strong self-interaction on the quantum level breaks the conformal invariance and introduces a fixed scale QCD leading to a positive pressure given by Eq. These strongly interacting theories reproduce several features of QCD. NUMBER 8 Z0 ÿ1 PHYSICAL REVIEW LETTERS 19 AUGUST 2002 ~ ~ j i ˆ j in i ‡ i _ ~ ^ dt …t†S …t†j in i ‡ O…H2 †: (11) If we assume …t† ˆ 0 ‡ expfHtg. where R denotes their (mean) distance (cf. (5) can be explained by relatively localized (nonperturbative) vacuum fluctuations (e. Let us further assume that the dynamical breakdown of the scale symmetry is basically encoded by these (nonperturbative) vacuum fluctuations.. (4) are (to first order in H and for m ˆ 0) conformally invariant.. cf. This can be most easily verified by assuming g  1 which allows for a perturbative treatment. every vacuum fluctuation ‘‘sees’’ all other vacuum fluctuations ‘‘redshifted. the expectation values of operators such as ^ ^ Q 2 or P 2 do not acquire a first-order correction. As a result there is no first-order ^ k ^ ÿk ^ ^ S ^ correction to the expectation value of T  for free (linear) fields. After the renormalization described in Eq. of course. it is reasonable to ^ take the Minkowski vacuum j0Min i with H Min j0Min i ˆ 0 ~ in i ˆ j0Min i. Instead it is an (approximate) eigenstate of the corrected Hamiltonian _ ^ ^ H Min ÿ S (see. whereas their (repulsive) interactions are adequately described by the free (perturbative) and thus conformally invariant field equations. the operator in Eq. H …t† ˆ (13) 2 …t† where ! corresponds to QCD . i.. (6). such as dynamical scale symmetry breakdown. Based on this intuitive picture it appears plausible to admit a correction to the expectation value of the energymomentum tensor within the FRW vacuum in Eq. instantons [7] or oscillons [8]) which repel each other (at least in average. their repulsion acquires a correction proportional to HR. Let us assume that a part of the positive vacuum pressure in Eq. This result can be transferred directly to free quantum fields: pitching on a particular normal mode with the wavelength k the ^ conformal charge S again acts like a squeezing operator ^ ! i…a y a y ÿ a k a ÿk †.g. In this case their solutions would simply be scaled during the cosmic expansion in contrast to the nonperturbative fluctuations which retain their scale and hence are not affected.. Q g ˆ ‰…a y †2 ÿ …a †2 Š: 4 4 (14) Consequently. [7]). (5). the subtraction of the Minkowski contribution.e. However. The remaining question is. After a normal mode decomposition one therefore indeed obtains terms like the one above. the positive vacuum pressure in Eq. e. (15) does entail a first-order correction. (12) which is linear in H.g.’’ i. leads to a displaced vacuum state (12).VOLUME 89. 081302-3 Let us estimate the associated order of magnitude: although QCD depends on the renormalization scheme 081302-3 . Accordingly. Within an expanding universe the balance of these two tendencies. cf. their solutions would simply be scaled during the expansion of the Universe — like the redshift of the photon field. (This would actually happen if the Universe were to expand very rapidly H  QCD . The remaining time as the initial condition j integration yields the (retarded) inverse of the Minkowski ^ Min Hamiltonian H ÿ1 and hence we arrive at ^ Min ^ j0FRW i ˆ j0Min i ‡ HH ÿ1 S j0Min i ‡ O…H2 †: (12) Therefore the adiabatic QCD vacuum j0FRW i of an expanding universe is not the instantaneous ground state j0Min i of ^ ^ H FRW …t† or H Min — it acquires corrections already in the first order of the adiabatic expansion.. i.e.e.) However. (9) can simply be generated by the squeezing operator 1 ^ ^ i ^ ^ ^ S ˆ fP . [6].. following the expansion on the one hand and retaining the scale on the other hand. (15) ÿ1 The retarded propagator Gret …t ÿ t0 † encodes the dynamics of an intermediate (interaction) degree of freedom which has been integrated out. next-to-leading order in 1=N) long-range four-point interactions.e.

Nucl. [10] L. J. C. J. A. Sov. By inspection one finds that the deduced order of magnitude nicely fits the empirical data %crit  10ÿ29 g cmÿ3 . Inserting the above values we ^ ^ finally arrive at hT  iFRW ˆ O…1016 mÿ4 † or hT  iFRW ˆ ren ren O…10ÿ29 g cmÿ3 †. 17. M. However. Bogolyubskiı and V. Muller. Astron. ibid. 60. S. M. 985 (1976). Riess et al. P. 74 (1998). R. Combining these two equalities and specifying the pressure p one may determine the time evolution of our Universe. J.. H. M. Phys. NUMBER 8 PHYSICAL REVIEW LETTERS 19 AUGUST 2002 we may fix it approximately via QCD ˆ O…108 eV† ˆ O…1014 mÿ1 †. Shifman.ca [1] C. Pis’ma Zh. 28. Chanowitz and J. Cambridge. [3] N. Phys. The masses of the light quarks which dominantly couple to the gluonic field are roughly of a similar order of magnitude. Shuryak. Phys. Cambridge. L1 (2001). 82. Phys. Furthermore the ^ Einstein equations imply r hT  iren ˆ 0. 81. Garnavich et al. Plunien. [5]) nonperturbative effects ^  ^ a (such as hG a G  iren and h  ^ iren ) display a dependence on the coupling of expfÿ82 =g2 g ˆ expfÿ2= g. Astrophys. Phys. [2] S. Phys. Shuryak. R.. Vainshtein. e. *Electronic address: schuetz@physics. The Hubble expansion parameter H is about 10ÿ26 mÿ1 . Pis’ma Zh. D. Caldwell. Vassilevich. 24. A. D. D 16. [4] R. 17 (2000). A. Lett. 80. 25. Lange et al. Coleman. E.. 2490 (1973). Khriplovich. Joglekar. Makhan’kov. S. Kuhn. Teor. Matthews. J. Inserting QED  1=137 into the above expression one obtains a suppression by an order of magnitude of 10ÿ370 . 472 (1986). 99 (1980). This work was supported by the Alexander von Humboldt foundation and by the NSERC. Elem. Rev. G. A. 39. Soff. Astrophys. A. Wang et al. G. e. 1985). D 14. Krauss. Rev. L53 (1998). 530. Nucl. B 220. J. 565 (1999). Hartle. A. G. A. 509.VOLUME 89. Davies. A. V. Crewther. Meyer-Hermann. 1992). Rev. Lett. Zakharov. B203. J. Salehi and Y. Y. Zhitnitsky for fruitful discussions and acknowledges valuable comments and criticism from Y. T. G. 120 (1977). Rev. 51 (1998). 36 (1989). Lett. C. H. D 63. Lett. V. de Bernardis et al. A. and H. R. Bisabr. Rev. and P. ibid. and V. Steinhardt. Unfortunately the presently available data (such as the Hubble parameter or the age of the Universe) are not precise enough to distinguish the two cases (p / H and p ˆ const). Vacuum Structure and QCD Sum Rules (Elsevier. V. 24. 1920 (1995). Miller et al. Phys. I. P.. 323 (1998). Binetruy.2]. Rev. 517. 93 (1982). [JETP Lett. ¨ Phys. J. J.. Mod.. Unruh and E. Fiz. ibid. Shrock. 549.. 083508 (1999). 1009 (1998). P. Zlatev. D. W. Science 292. Eksp.. Bardeen. Amsterdam. Of course one may ask whether the remaining electroweak sector of the standard model generates similar contributions: typically (see. Lee. 083504 (2001). J. The remaining explicit breaking of the scale symmetry induced by the Higgs field of course also generates con^ tributions to T  . Yadra 17.g. (14) they do not contribute to the first order in H. Phys.g. 25.. It should be mentioned here that a pressure induced by the expansion of our Universe with p / H generates a cosmic evolution which differs from that with a true cosmological constant p ˆ const: by inserting the FRW metric (1) into the Einstein equations (3) one obtains the Friedmann equation 3H2 ˆ 8 GN %. [10]). Astrophys. Shifman. Collins. Phys. Theor. ibid. [6] V. and P. J. M. Page. Melchiorri et al. 1757 (1979). [5] M. Schmidt et al. D 52. Teor. Phys. 1241 (2000). Chastits At. Fischetti. B. D 60. In view of the huge difference of the involved scales [H=QCD ˆ O…10ÿ40 †] this remarkable coincidence seems to be almost too good to be just an accident. B165.. 493. for example. Rev. Steinhardt. Starobinsky. ibid. 116. 3475 (2001). 1982). Fiz. M. M. a contribution of the electroweak sector in analogy to QCD is not obvious. L59 (2000). T. Padin et al. A. Phys. W. Rev. 40B. Rep. England. S. J. Rev. Carroll. A. A. 3067 (1998). J. 12 (1976)]. Phys. Novikov et al. [9] S. L. Part. R. B 468. T.. D 63. Lett. and S. Brax and J. Astrophys. Gleiser ¨ and H. 40 (1999). Lett. R. M. Woodard. J. 568. 45 (1980). The scale of the dynamical symmetry breaking QCD obeys a similar nonanalytical dependence on the coupling g. Fiz. Hawking. Rev. B. and R. L63 (2000). 103 (1984). D 20. Although the mixture of these terms caused by interactions remains subject to further considerations. E. Hence the contributions arising from the dynamical breaking of scale invariance can safely be neglected in this case. 91B. 46 (1998). B. Phys. 107 (1977)]. Nucl. 955 (2000). Int. F. 1582 (1998). d%=dt ˆ ÿ3…% ‡ p†H. England. Aspects of Symmetry (Cambridge University Press. Martin. Birrell and P.. Eksp. Hu. i. Ellis. M. Reall. 70. 397 (1972). 042001 (2001). Wang. Schafer and E. 896 (1999).e. At least it indicates the potential relevance of the effect described in the present Letter with regard to the interpretation of the astrophysical data [1. P. L.ubc. quintessence (see. Lett.. E. V. 46 (2002). Rev. Phys. V. I. C. Chiba. W. J. P. ibid. 081302-4 . Rev. Dave. 204 (1986). Phys. 2302 (2001).. Phys. [JETP Lett. W. 536. L. 116. J. These investigations might perhaps lead to a better understanding of some of the problems in cosmology without necessarily invoking yet unknown low-energy fields. Novozhilov and D. the present Letter motivates a deeper examination of the vacuum of strongly interacting fields in the gravitational background of our expanding Universe — for the present epoch as well as for earlier stages. and B. 438 (1977). 507. Jaffe et al. Phys. 86. S. D 7. [9]. Pryke et al. 1421 (1972). Phys. Mauskopf et al. I. 536. Astrophys. P. ˘ [8] I. Duncan.. Hertog. 081302-4 The author is indebted to B. 063502 (1999)... R. Phys. 15 (1976). Quantum Fields in Curved Space (Cambridge University Press. Perlmutter et al. Copeland. and R. Nature (London) 391. In summary. Nature (London) 404. the general structure of all ^ ^ these terms is given by m2 h y  iren and according to the arguments after Eq. cf. [7] E.