You are on page 1of 4

VOLUME 59, NUMBER 22 PHYSICAL REVIEW LETTERS 30 NOVEMBER 1987

Anthropic Bound on the Cosmological Constant


Steven Weinberg
Theory Group, Department of Physics, University of Texas, AustinT, exas 78712
(Received 5 August 1987)

In recent cosmological models, there is an "anthropic" upper bound on the cosmological constant A.
It is argued here that that do not recollapse, the only such bound on A is that it should not
in universes
be so large as to prevent the formation of gravitationally bound states. It turns out that the bound is
quite large. A cosmological constant that is within 1 or 2 orders of magnitude of its upper bound would
help with the missing-mass and age problems, but may be ruled out by galaxy number counts. If so, we
may conclude that anthropic considerations do not explain the smallness of the cosmological constant.

PACS numbers: 98.80. Dr, 04. 20. Cv

Our knowledge of the present expansion rate of the in cosmological constants between the inside and outside
Universe indicates that the effective value A of the of each membrane caused by the diA'erences in the values
cosmological constant is vastly less than what would be of the four-form field strength. In models of the type
produced by quantum fluctuations' in any known realis- discussed in Refs. 4 and 5 it may not be strictly neces-
tic theory of elementary particles. In view of the contin- sary to invoke the anthropic principle because gravita-
ued failure to find a microscopic explanation of the tional effects can stop the process of bubble formation
smallness of the cosmological constant, it seems when the vacuum energy is about to become negative.
worthwhile to look for a solution in other, "anthropic,
" However, it takes an enormously long time to reach this
directions. Perhaps A must be small enough to allow final stage, and anthropic arguments may be needed to
the Universe to evolve to its present nearly empty and explain why we are not still in an earlier stage, with
flat state, because otherwise there would be no scientists large effective cosmological constant.
to worry about it. Without having a definite framework (c) Fluctuations in scalar fields can trigger cosmic
for such reasoning, one can at least point to four lines of inflation in regions of the Universe where the fields hap-
current cosmological speculation, in which anthropic pen to be large. Except near the edges, the inflationary
considerations could set bounds on the value we observe region would appear to its inhabitants as a separate
for the effective cosmological constant: subuniverse. In this region further fluctuations can pro-
(a) The effective cosmological constant may evolve duce new inflations, and so on. This has been studied by
very slowly, perhaps because of slow changes in the value Linde, who remarks that the physical constants of the
of some scalar field, as in the model of Banks. In this subuniverse in which we live may be in part constrained
case, it would be natural to expect that for some very by the requirement that life could arise in such a
long epoch the cosmological constant would remain near subuniverse.
zero. The question then is, why do we find ourselves in (d) Quantum Auctuations in the very early Universe
such an epoch? As remarked by Banks, the answer may may cause incoherence between different terms in the
be anthropic: Perhaps only in such epochs is life possi- state vector of the Universe; each term would then in
ble. effect represent a separate universe. Such a picture has
(b) The Universe may evolve through a very large been considered by Hawking. Our own Universe could
number of first-order phase transitions, in which bubbles correspond to any one of the terms in the state vector,
form within bubbles within bubbles. . . , each bubble subject only to the anthropic condition, that it be a
having within it a smaller value of the vacuum energy, universe in which life could develop.
and hence of the effective cosmological constant. Ef the Without committing ourselves to any one of these
steps in vacuum energy are very small, then it would be cosmological models, it seems appropriate at least to ask,
natural to expect that there would be some phase in just what limit does the anthropic principle place on the
which the eff'ective cosmological constant is correspond- effective cosmological constant A?
ingly small. Abbott has suggested a scalar-field theory Fortunately, at least for A & 0, the anthropic principle
with a potential that has an infinite number of closely provides a rather sharp upper bound on A. This is be-
spaced local minima; bubbles form within bubbles as the cause in a continually expanding universe, the cosmologi-
scalar-field value jumps from one minimum to the next. cal constant (unlike charges, masses, etc. ) can affect the
Recently Brown and Teitelboim have proposed a model evolution of life in only one way. Without undue anthro-
in which a similar sequence of phase transitions occurs, pocentrism, it seems safe to assume that in order for any
but in which the bubble walls are elementary membranes sort of life to arise in an initially homogeneous and iso-
coupled to a three-form gauge field, with the difference tropic universe, it is necessary for sufficiently large gravi-

2607
VOLUME 59, NUMBER 22 PHYSICAL REVIEW LETTERS 30 NOVEMBER 1987

tationally bound systems to form first. (By "sufficiently


large" is meant large enough to form stars, and large
enough also to contain the heavy elements produced by
early generations of stars so that planets can form
around subsequent generations of stars. Galaxies and
Treating
:
ized by the parameter
Plim [[&p(t)] '/p'(t)]
t=o
d, k and Ap to first order for t 0 in Eq.
(4)

(1),
we find for the perturbed curvature constant
probably also the larger globular clusters are suSciently
large in this sense. ' ) However, once a sufficiently large Ak =", zGa (p+Ap) p (5)
gravitationally bound system has formed, a cosmological
constant would have no further eAect on its dynamics, or Equation (1) shows that the perturbed scale factor
on the eventual evolution of life. In particular, it makes a(t) will increase to a maximum and then collapse back
no diff'erence if the e-folding time of the cosmic expan- to a =0, provided that there is a value of a(t) where the
sion is much shorter than the time required for the evolu- right-hand side of Eq. (I) is equal to Ak. The right-
tion of intelligent life. '' (Note that I am here not re- hand side of Eq. (1) reaches a minimum at a value of
quiring that the cosmological constant have a value con- a(t) such that p+Ap=2pv, and so the condition for a
sistent with the astronomical knowledge, but only that it given perturbation to undergo gravitational condensation
have a value consistent with the appearance of beings is that at this minimum, the right-hand side of Eq. (1)
that could measure it. ) Thus, irrespective of what we should be less than h, k. This condition can be written
think are the possible forms of intelligent life, the neces- gtrGpv" [ ,' (p+Wp)]"'a' & ak. (6)
sary and sufhcient anthropic condition on the cosmologi-
cal constant is that it should not be so large as to prevent With use of Eq. (5) to express the perturbed curvature
the appearance of gravitationally bound states. constant in terms of the parameter p, this becomes
I evaluate this bound here in the context of conven- 500
tional big-bang cosmologies, described by a Robertson- pv&9p. (7)
Walker metric with initially small perturbations. For If there is an upper bound on the parameter p, and if this
definiteness I will concentrate here on the case of positive upper bound is independent of A (because it refers to
cosmological constant and zero spatial curvature, k=0, very early times, when pv is negligible), then Eq. (7)
and with the energy density of the Universe dominated provides our anthropic bound on the vacuum energy den-
since recombination by nonrelativistic matter. However, sity.
it would not be hard to adapt the arguments to any other [It is instructive to compare this with the result of
case in which the Universe does not recollapse. a linear analysis, in which h, k and hp are treated
Let us consider then the fate of a density perturbation throughout as first-order perturbations. The solution of
in a Universe with A&0. Such a perturbation can be the second-order differential equation' for Ap/p takes
modeled' as a sphere within which there is a uniform the form
excess density hp(t), and a gravitational field described
by a Robertson-Walker (RW) metric with positive cur- ~ (sinhr ) 't'Q (coth r),
vature constant Ak ) 0, and with RW scale factor a(t).
The evolution of the perturbation is governed by a Fried-
where
Ap/p lttg

r=(6trGpv) ' t,
mann equation
and Q,"(z) is the associated Legendre function of the
(da/dt) +6k =
', ~Ga (p+Ap+pv),
second kind. ' This behaves for z((1 like and then z,
rises monotonically' to a constant limit as z By ~.

:
where p(t) is the unperturbed cosmic mass density, pv is
comparing the asymptotic behavior of Ap/p for r 0
the constant vacuum energy density
and z ~, we can see that if we normalize so that
pvA/gtrG, Ap/p er for r 0, then for r
(2)
", )r(
~p/p- (2/J~)r( -', )~ = 1.437~.
and Ap(t) is the perturbation, satisfying the equation of
mass conservation With this normalization, the parameter p is just e pt, so
that hp/p 1.437(p/pv) 't . One might guess that the
a '(p+ Ap) = const. necessary and sufhcient condition for gravitational con-
densation is that the linear analysis should give an
I am not assuming here that hp or hk is small, but there asymptotic value of Ap/p at least of order unity. If this
is a branch of the solutions
t 0, hp tx t while p ~ t,
of Eq. (1) for which, as
so that in this limit
p~&&hp&&p. I assume that all perturbations are on this
were correct, then the upper bound on p~ for gravitation-
al condensation to occur would be

branch, so that the universe looks smooth for t 0. The pv & (1 437)P =2.97p.
strength of such a perturbation can then be character- This is quite different from the result (7), showing the

2608
VOLUME 59, NUMBER 22 PHYSICAL REVIEW LETTERS 30 NOVEMBER 1987

inadequacy here of linear methods. ] Now, if the intrinsic distribution of py values is


Now, what is the distribution of p values for actual smooth and featureless below the anthropic bound (as
perturbations? Eventually, the theory of the very early would be expected in the models of Refs. 3-5, 8, and 9 if
Universe may develop to the point that the distribution the natural scale for pz is set by the Planck mass), then
of p will be known in terms of fundamental constants. it seems likely that pv would be within 1 or 2 orders of
Alternatively, it may some day become possible to read magnitude of its upper bound. (This can be made more
oA the distribution of values of p from observations of precise by calculations like those of Carter in Ref. 2. )
small-angle fluctuations in the cosmic microwave back- With a lower bound of 550pp on the anthropic upper
ground temperature. bound on py, we would then conclude that p1 must be
For the present, it seems that the best we can do is to much greater than the present mass density pp. Is this
use the existence of quasars at high red shifts to set an plausible? The answer unfortunately depends on data
empirical lower bound on any maximum value for p, and whose interpretation is far from settled.
hence a lower bound on the anthropic upper bound (6) On one hand, a vacuum energy density much larger
on pz. This is not as pointless as it may seem. If it turns than pp would resolve a problem that has been posed by
out that the empirical lower bound on the maximum measurements of pp for those who believe, on grounds of
value of the right-hand side of Eq. (6) is much larger either inflation or aesthetics, that (as assumed here) the
than empirically allowed values for pz, then we would Universe has a vanishing unperturbed curvature constant
have to conclude that the anthropic principle does not k. This implies that the total energy density p&+pp is
explain why the cosmological constant is as small as it is. equal to its critical value 3HO/8rrG; that is Qz+ Oo =1,
In the model we have been considering, the time re- where A y =8+Gpy/3HD and Qo =8+Gpo/3HO. The deu-
quired for the scale factor a(r) of a perturbation to terium abundance indicates' that the contribution of
reach its maximum and then collapse back to a =0 is an baryons to Qo is no greater than about 0.03 (for Ho
increasing function of pz, so that it is bounded below by =100 km/sec Mpc), while observations of galaxies sug-
its value when py =0: gest' that they contribute about 0.02 to Ap. Even al-
r, )
', x(250+Gp) (8)
lowing for nonbaryonic extra-galactic matter, the dy-
namics of clusters of galaxies lead to estimated values'
Actually, we do not measure the age of early gravitation- of Ap only about 0. 1 to 0.2. It would be dificult to satis-
al condensations, but we do observe their red shifts. The fy the condition Op+01 =1 if At were much smaller
time corresponding to a red shift z, is bounded above by than Ap. However, if for instance we suppose that pv is
its value for py =0: at least 1% of its anthropic upper bound (10), then
Ay/Ao must be at least 5. 5, so that we would not need
(9) Ap to be greater than 0. 15.
The assumption of a vacuum energy density roughly
where pp is the present cosmic mass density. Thus, put-
comparable with the anthropic upper bound would also
ting together (8) and (9), we see that the observation of
help' with the problem of cosmic ages that arises if the
gravitational condensations (e.g. , quasars) of red shift z,
Hubble constant Ho is as large as 100 km/sec Mpc. In
or greater sets an empirical lower bound on the max-
this case, if A =0 and k =0, then the age of the Universe
imum value of the parameter p: '
is about , Ho ' =6.5x10 yr, less than the (1020)
500
729 pmax ) 3 + 2 po(1 +zc) 3 (10) x 10 yr usually given ' for the globular clusters. How-
ever, if at present p1 is much greater than pp, then over
For instance, we know that quasars exist with red shifts most of the history of the Universe R (t ) behaves as
up to about z =4.4, so taking z, =4.5 in (10) gives a exp(Hr ) rather than t i,
and the age of the Universe is
lower bound of 550pp on the anthropic upper bound on greater than & Hp '. Specifically, the present age of a
pv. gravitational condensation which forms at a red shift z,
1s

to
r, = ' (1+po/p~)
,
'i Ho '
finvsinh(py/po) 'i invsinh[(p~/po) '
(1+z, ) ~
]I.
If for instance z, =4 and py/po=9 (i.e. , 00=0. 1), then
the age (11) is 1. 1HO ', which even for Ho close to 100 thropic upper bound (7).
km/sec Mpc leaves adequate time for the evolution of A similar conclusion would be reached if it turns out
globular clusters. that gravitational condensations occur at red shifts larger
On the other hand, counts of galaxies as a function than 4. 5. For instance, if gravitational condensations
of red shift indicate (for k =0) that A/3HD =0. 1+0'4 occur at a red shift z, =30, then according to Eq. (10),
or in other words py/p0=0. 1 ~0.3. According to (10), the anthropic upper bound on py is at least 10 pp. But
this is at least 3 orders of magnitude less than the an- Qo cannot be less than about 0.01, indicating (for k =0)

2609
VOLUME 59, NUMBER 22 PHYSICAL REVIEW LETTERS 30 NOVEMBER 1987

a vacuum energy density pz no greater than 100po, ing out of the gas from the clusters during these passages. I f
which is again at least 3 orders of magnitude less than globular clusters formed first, and then did not associate with
the anthropic upper bound. galaxies, there could still be heavy elements formed by the
Thus if the interpretation of galaxy number counts in evolving stars in a cluster, condensing to the plane of rotation
Ref. 20 holds up, or if gravitational condensations are of the cluster, perhaps reaching densities high enough to allow
the formation of second-generation stars with planets. This
found at red shifts z 4, we will be able to conclude that
could work for large enough globular clusters which thus
the cosmological constant is so small that even the an- eA'ectively become small galaxies. A small globular cluster
thropic principle could not explain its smallness. would probably not have sufhcient gravitational attraction to
I am grateful for conversations with T. Banks, hold the gas against the blast effects of supernovae stripping it
G. Field, F. Wilczek, E. Witten, A. Zee, and with many from the cluster.
colleagues at the University of Texas, and especially to iiln this respect, I differ from Barrow and Tipler (Ref. 2)
P. Shapiro for helpful discussions of the data and refer- and consequently obtain a different anthropic bound on A.
They require that A 't (roughly the vacuum Hubble time)
ences in observational cosmology. This work is support- ~ i

ed in part by the Robert A. Welch Foundation and Na- should be at least as large as the main-sequence stellar lifetime
tional Science Foundation Grant No. 8605978. This seems to me correct for A & 0 [where the Universe
recollapses in a time rt(3 A ) ' ], but not for A) 0, the case
~ ~

under consideration here. It is true that for A) 0, gravitation-


al condensation must occur within a time of order A ', be-
cause after that time the matter density drops below the vacu-
'See, e.g. , Ya. B. Zeldovich, Pis'ma Zh. Eksp. Teor. Fiz. 6, um density, and gravitational condensation becomes impossi-
883 (1967) [JETP Lett. 6, 316 (1967)]. ble. However, once gravitationally bound systems form, their
2For a comprehensive survey of the anthropic principle, see subsequent evolution is unaA'ected by the cosmological con-
J. D. Barrow and F. J. Tipler, The Anthropic Cosmological stant, and it makes no difI'erence how long it takes for stars to
Principle (Clarendon, Oxford, 1986). Also see P. C. W. evolve on the main sequence.
Davies, The Accidental Universe (Cambridge Univ. Press, '2P. J. E. Peebles, Astrophys. J. 147, 859 (1967).
Cambridge, 1982), Chap. 5; B. Carter, Philos. Trans. Roy. Soc. ' See, e. g. S. Weinberg, Gravitation and Cosmology (Wiley,
,
London A 310 347 (1983). New York, 1972), Eq. (15.10.57).
3T. Banks, Nucl. Phys. B249, 332 (1985). i41 use the notation of the Handbook of Mathematical Func
4L. F. Abbott, Phys. Lett. 150B, 427 (1985). tions, edited by M. Abramowitz and I. A. Stegun (Dover, New
~J. D. Brown and C. Teitelboim, to be published. York, 1965), Sec. 8.
The introduction of a three-form gauge field has the effect i5I am very grateful to P. Candelas for a numerical integra-
of making the effective cosmological constant a constant of in- tion of the diff'erential equation for Ap/p.
tegration, as shown by E. Witten, in Shelter Island II: A. M. Boesgaard and G. Steigman, Annu. Rev. Astron. As-
Proceedings of the 1983 Shelter Island Conference on Quan- trophys. 23, 310 (1985).
turn Field Theory and the Fundamental Problems of Physics, Dark Matter in the Universe, edited by G. R. Knapp and
Shelter Island, 1Vew York, 1983, edited by R. Jackiw et al. J. Kormendy, Proceedings of the International Astronomical
(M. I.T. Press, Cambridge, MA, 1985); M. Henneaux and Union Symposium No. 117 (Reidel, Dordrecht, 1987).
C. Teitelboim, Phys. Lett. 143B, 415 (1984). ' G. de Vaucouleurs has long advocated the introduction of a
S. Coleman and F. De Luccia, Phys. Rev. D 21, 3305 cosmological constant to reconcile the apparent inconsistency
(1980). of his measurement of the Hubble constant with globular clus-
8A. D. Linde, to be published. ter ages; see Astrophys. J. 268, 468 (1983), Appendix B, and
S. Hawking, as quoted by M. Gell-Mann, Phys. Scr. T15, Nature (London) 299, 303 (1982), etc. Also see P. J. E. Pee-
202 (1987). bles, Astrophys. J. 284, 439 (1984).
' This remark about globular clusters is due to H. Smith. He i9A. Renzini, in Galaxy Distances and Deviations prom
points out that the observed low gas abundance in globular Universal Expansion, edited by B. F. Madore and R. B. Tully
clusters presumably results from their inevitable frequent pas- (Reidel, Dordrecht, 1986), pp. 177-184.
sages through the plane of our galaxy with consequent sweep- ~OE. D. Loh, Phys. Rev. Lett. 57, 2865 (1986).

2610

You might also like