Issues in the Philosophy of Cosmology
arXiv:astro-ph/0602280v2 29 Mar 2006
George F R Ellis.
Mathematics Department and Applied Mathematics,
University of Cape Town,
Rondebosch, Cape Town 8001, South Africa.
email: [email protected]
February 5, 2008
After a survey of the present state of cosmological theory and observations, this article discusses a series of
major themes underlying the relation of philosophy to cosmology. These are:
A: The uniqueness of the universe;
B: The large scale of the universe in space and time;
C: The unbound energies in the early universe;
D: Explaining the universe — the question of origins;
E: The universe as the background for existence;
F: The explicit philosophical basis;
G: The Anthropic question: fine tuning for life;
H: The possible existence of multiverses;
I: The natures of existence.
Each of these themes is explored and related to a series of Theses that set out the major issues confronting
cosmology in relation to philosophy.
Cosmology is the study of the large-scale structure of the Universe, where ‘the Universe’ means all that exists in
a physical sense [104]. This is to be distinguished from the Observable Universe, namely that part of the Universe
containing matter accessible to our astronomical observations, which is a subset of the Universe proper. Thus
cosmology considers the vast domain of galaxies, clusters of galaxies, quasi-stellar objects, etc., observable in the
sky by use of telescopes of all kinds, examining their nature, distribution, origins, and relation to their larger
environment. Observational cosmology [114, 128, 99, 189, 18] aims to determine the large-scale geometry of the
observable universe and the distribution of matter in it from observations of radiation emitted by distant objects,
while physical cosmology [163, 193, 233, 202, 171, 38] is the study of interactions during the expansion of the
universe in its early hot big bang phase, and astrophysical cosmology [193, 165, 160, 174, 38] studies the resulting
later development of large-scale structures such as galaxies and clusters of galaxies. Various forms of quantum
cosmology (see e.g. [109, 92, 26]) and studies of particle physics aspects of cosmology [126, 162, 1, 171, 38] attempt
to characterize the epochs before the hot big bang phase. These studies function in a mainly symbiotic way, each
informing and supplementing the others to create an overall cosmological theory of the origin and evolution of the
physical universe [14, 104, 201].
A unique role of the universe is in creating the environment in which galaxies, stars, and planets develop,
thus providing a setting in which local physics and chemistry can function in a way that enables the evolution of
life on planets such as the Earth. If the cosmological environment were substantially different, local conditions
would be different and in most cases we would not be here [21, 29, 7, 214, 175] — indeed no biological evolution at
all would have taken place. Thus cosmology is of substantial interest to the whole of the scientific endeavor, for it
sets the framework for the rest of science, and indeed for the very existence of observers and scientists. It is unique
as the ultimate historical/geographical science.
Cosmology as a serious scientific study began with the discovery of Einstein’s static universe in 1917,
followed by the key observational discovery of the linear redshift-distance relation by Hubble in 1929, indicating
the expansion of the universe, and the development of theoretical understanding of the geometry and dynamics
of the non-static Friedmann-Lemaı̂tre models with their Robertson-Walker geometry [159, 9, 204, 56, 127]. It has
been transformed in the past decades into a mainstream branch of physics [5, 158] by the linking of nuclear and
particle physics theories to observable features of the cosmos [233, 126, 162, 1, 38], and into an important part
of astronomy because of the massive flow of new astronomical data becoming available [99, 106, 18], particularly
through new ground-based telescopes such as Keck and through balloon and satellite observatories such as the
Hubble Space telescope (optical and ultra-violet), IRAS (infra-red), ROSAT (x-ray), and COBE and WMAP
(microwave). Thus the subject has progressed from a mainly mathematical and even philosophical exercise to
an important part of mainstream science, with a well-established standard model confirmed by various strands of
evidence [233, 166, 201, 162, 38]. Nevertheless because of its nature, it is different from any other branch of the
natural sciences, its unique features playing themselves out in the ongoing interaction between speculation, theory,
and observation.
Cosmology’s major difference from the other sciences is the uniqueness of its object of study — the Universe
as a whole [147, 148, 156] — together with its role as the background for all the rest of physics and science, the
resulting problems being accentuated by the vast scale of the universe and by the extreme energies occurring in the
very early universe. We are unable to manipulate in any way its originating conditions, and there are limitations
on our ability to observe both to very distant regions and to very early times. Additionally, there are limits to
our ability to test the physics relevant at the earliest epochs. Consequently it is inevitable that (as is also the
case for the other historical sciences) philosophical choices will to some degree shape the nature of cosmological
theory, particularly when it moves beyond the purely descriptive to an explanatory role [146] — that move being
central to its impressive progress in recent decades. These philosophical choices will strongly influence the resulting
understanding, and even more so if we pursue a theory with more ambitious explanatory aims.
After a substantial outline of present day cosmology in Section 2, these issues will be explored in the subsequent sections, based on a series of thirty-four Theses clustered around nine key aspects of the nature of cosmology,
broadly speaking relating to geometry, physics, and philosophy, that frame the context of the philosophical issues
facing cosmology and its relation to local physics. I believe this formulation helps focus on specific issues of
importance in this relation. To those who believe cosmology is simply about determining a number of physical
parameters, this will seem a vastly over-complicated approach; but then a major purpose of this paper is precisely
to counter such simplistic visions of the nature of cosmology. For other reports on the philosophy of cosmology,
see [149, 156, 58, 136, 137].
Outline of cosmology
A series of basic features of present day cosmology are now well established. Decades of painstaking work has
established the distances of galaxies and hence the huge scale of the universe, as well as the basic feature that the
universe is expanding and evolving; the old dream of a static universe is unviable [57]. Cosmology proceeds by
assuming the laws of physics are the same everywhere, and underlie the evolution of the universe. The dominant
role of gravity, despite its weakness, then arises from the fact that it is the only known force acting effectively
on astronomical scales (the other known long-range force is electromagnetism, but in this case negative charges
balance out positive charges, leaving no resultant large-scale effect). Consequently, cosmological theory describing
all but the very earliest times is based on the classical relativistic theory of gravitation, namely Einstein’s General
Theory of Relativity [145], with the matter present determining space-time curvature and hence the evolution of
the universe. The way this works out in any particular situation depends on the nature of the matter/fields present,
described by their effective equations of state and interaction potentials.
The survey of cosmology in this section looks successively at the basic models of cosmology; the hot big bang;
cosmological observations, including the Cosmic Background Radiation anisotropy spectrum; causal and visual
horizons, and their implications; recent theoretical developments (including inflation); the very early universe; and
the present concordance model, which includes both dark matter and dark energy.
Basic Theory
Cosmology starts by assuming that the large-scale evolution of spacetime can be determined by applying Einstein’s
field equations of Gravitation (‘EFE’) everywhere: global evolution will follow from local physics. The standard
models of cosmology [179, 45, 233, 110] are based on the assumption that once one has averaged over a large enough
physical scale, isotropy is observed by all fundamental observers (the preferred family of observers associated with
the average motion of matter in the universe). When this isotropy is exact, the universe is spatially homogeneous
as well as isotropic [231, 45, 50]. The matter motion is then along irrotational and shearfree geodesic curves with
tangent vector ua , implying the existence of a canonical time-variable t obeying ua = −t,a . The Robertson-Walker
(‘RW’) geometries used to describe the large-scale structure of the universe [180, 230] embody these symmetries
exactly. Consequently they are conformally flat, that is, the Weyl tensor is zero:
Cijkl := Rijkl + 12 (Rik gjl + Rjl gik − Ril gjk − Rjk gil ) − 16 R(gik gjl − gil gjk ) = 0;
this tensor represents the free gravitational field, enabling non-local effects such as tidal forces and gravitational
waves which do not occur in the exact RW geometries.
Comoving coordinates can be chosen so that the metric takes the form:
ds2 = −dt2 + S 2 (t) dσ 2 , ua = δ a 0 (a = 0, 1, 2, 3)
where S(t) is the time-dependent scale factor, and the worldlines with tangent vector ua = dxa /dt represent the
histories of fundamental observers. The space sections {t = const} are surfaces of homogeneity and have maximal
symmetry: they are 3-spaces of constant curvature K = k/S 2 (t) where k is the sign of K. The normalized metric
dσ 2 characterizes a 3-space of normalized constant curvature k; coordinates (r, θ, φ) can be chosen such that
dσ 2 = dr2 + f 2 (r) dθ2 + sin2 θdφ2
where f (r) = {sin r, r, sinh r} if k = {+1, 0, −1} respectively. The rate of expansion at any time t is characterised
by the Hubble parameter H(t) = Ṡ/S.
To determine the metric’s evolution in time, one applies the Einstein Field Equations (‘EFE’), showing the
effect of matter on space-time curvature, to the metric (2,3). Because of local isotropy, the matter tensor Tab
necessarily takes a perfect fluid form relative to the preferred worldlines with tangent vector ua :
Tab = (µ + p/c2 )ua ub + (p/c2 )gab
(c is the speed of light). The energy density µ(t) and pressure term p(t)/c2 are the timelike and spacelike eigenvalues
of Tab . The integrability conditions for the EFE are the energy-density conservation equation
T ab;b = 0 ⇔ µ̇ + (µ + p/c2 )3Ṡ/S = 0 .
This becomes determinate when a suitable equation of state function w := pc2 /µ relates the pressure p to the
energy density µ and temperature T : p = w(µ, T )µ/c2 (w may or may not be constant). Baryons have {pbar = 0
}, which by (5) imply
⇔ w = 0} and radiation has {pradc2 = µrad /3 ⇔ w = 1/3, µrad = aTrad
µbar ∝ S −3 , µrad ∝ S −4 , Trad ∝ S −1 .
The scale factor S(t) obeys the Raychaudhuri equation
3S̈/S = − κ(µ + 3p/c2 ) + Λ,
where κ is the gravitational constant and Λ the cosmological constant.1 This shows that the active gravitational
mass density of the matter and fields present is µgrav := µ + 3p/c2 . For ordinary matter this will be positive:
µ + 3p/c2 > 0 ⇔ w > −1/3
(the ‘Strong Energy Condition’), so ordinary matter will tend to cause the universe to decelerate (S̈ < 0). It
is also apparent that a positive cosmological constant on its own will cause an accelerating expansion (S̈ > 0).
When matter and a cosmological constant are both present, either result may occur depending on which effect is
dominant. The first integral of equations (5, 7) when Ṡ 6= 0 is the Friedmann equation
Ṡ 2
κµ Λ
+ − 2.
This is just the Gauss equation relating the 3-space curvature to the 4-space curvature, showing how matter directly
causes a curvature of 3-spaces [45, 50]. Because of the spacetime symmetries, the ten EFE are equivalent to the
two equations (7, 9). Models of this kind, that is with a Robertson-Walker (‘RW’) geometry with metric (2, 3) and
dynamics governed by equations (5, 7, 9), are called Friedmann-Lemaı̂tre universes (‘FL’ for short). The Friedmann
equation (9) controls the expansion of the universe, and the conservation equation (5) controls the density of matter
as the universe expands; when Ṡ 6= 0 , equation (7) will necessarily hold if (5, 9) are both satisfied.
Given a determinate matter description (specifying the equation of state w = w(µ, T ) explicitly or implicitly)
for each matter component, the existence and uniqueness of solutions follows both for a single matter component
and for a combination of different kinds of matter, for example µ = µbar + µrad + µcdm + µν where we include cold
dark matter (cdm) and neutrinos (ν). Initial data for such solutions at an arbitrary time t0 (eg. today) consists of,
• The Hubble constant H0 := (Ṡ/S)0 = 100h km/sec/Mpc;
• A dimensionless density parameter Ωi0 := κµi0 /3H02 for each type of matter present (labelled by i);
• If Λ 6= 0, either ΩΛ0 := Λ/3H02 , or the dimensionless deceleration parameter q0 := −(S̈/S)0 H0−2 .
1 A cosmological constant can also be regarded as a fluid with pressure p related to the energy density µ by {p = −µc2 ⇔ w = −1}.
For the history of the cosmological constant, see [42, 43].
Given the equations of state for the matter, this data then determines a unique solution {S(t), µ(t)}, i.e. a unique
corresponding universe history. The total matter density is the sum of the terms Ωi0 for each type of matter
present, for example
Ωm0 = Ωbar0 + Ωrad0 + Ωcdm0 + Ων0 ,
and the total density parameter Ω0 is the sum of that for matter and for the cosmological constant:
Ω0 = Ωm0 + ΩΛ0 .
Evaluating the Raychaudhuri equation (7) at the present time gives an important relation between these parameters:
when the pressure term p/c2 can be ignored relative to the matter term µ (as is plausible at the present time),2
q0 =
Ωm0 − ΩΛ0 .
This shows that a cosmological constant Λ can cause an acceleration (negative q0 ); if it vanishes, the expression
simplifies: Λ = 0 ⇒ q = 21 Ωm0 , showing how matter causes a deceleration of the universe. Evaluating the
Friedmann equation (9) at the time t0 , the spatial curvature is
K0 := k/S02 = H02 (Ω0 − 1).
The value Ω0 = 1 corresponds to spatially flat universes (K0 = 0), separating models with positive spatial curvature
(Ω0 > 1 ⇔ K0 > 0) from those with negative spatial curvature (Ω0 < 1 ⇔ K0 < 0).
The FL models are the standard models of modern cosmology, surprisingly effective in view of their extreme
geometrical simplicity. One of their great strengths is their explanatory role in terms of making explicit the way
the local gravitational effect of matter and radiation determines the evolution of the universe as a whole, this in
turn forming the dynamic background for local physics (including the evolution of the matter and radiation).
The basic solutions
For baryons (pressure-free matter) and non-interacting radiation, the Friedmann equation (9) takes the form
3Ṡ 2
Λ 3k
= 3+ 4 + − 2
where A := κµbar0 S03 and B := κµrad0 S04 . The behaviour depends on the cosmological constant Λ [179, 178].
When Λ = 0, the universe starts off at a very dense initial state — according to the classical theory, an
initial singularity where the density and curvature go infinite (see Sec.2.1.2). Its future fate depends on the value of
the spatial curvature, or equivalently the density parameter Ω0 . The universe expands forever if {k = 0 ⇔ Ω0 = 1}
or {k < 0 ⇔ Ω0 < 1}, but collapses to a future singularity if {k > 0 ⇔ Ω0 > 1}. Thus Ω0 = 1 corresponds to the
critical density µcrit separating Λ = 0 FL models that recollapse in the future from those that expand forever, and
Ω0 is just the ratio of the matter density to this critical density:
{Ωcrit = 1 ⇔ κµcrit = 3H02 } ⇒ Ω0 := κµ0 /3H02 = µ0 /µcrit .
When Λ < 0, all solutions start at a singularity and recollapse.
When Λ > 0, if k = 0 or k = −1 all solutions start at a singularity and expand forever. If k = +1 there can
again be models with a singular start, either expanding forever or collapsing to a future singularity. However in
this case a static solution (the Einstein static universe) is also possible, as well as models asymptotic to this static
state in either the future or the past. Furthermore models with k = +1 can bounce (collapsing from infinity to a
minimum radius and re-expanding).
The dynamical behaviour of these models has been investigated in depth: first for dust plus a cosmological
constant [179, 178], followed by perfect fluids, fluids with bulk viscosity, kinetic theory solutions, and scalar field
solutions. Current models employ a realistic mixture of matter components (baryons, radiation, neutrinos, cold
dark matter, a scalar field, and perhaps a cosmological constant). Informative phase planes show clearly the way
higher symmetry (self-similar) models act as attractors and saddle points for the other models [144, 47].
The simplest expanding solutions are the following:
1. The Einstein-de Sitter model, for which {p = 0, Λ = 0, k = 0} ⇒ Ω0 = 1. This is the simplest expanding
non-empty solution:
S(t) = C t2/3
2 Assuming
we represent ‘dark energy’ (Sec.2.3.6) as a cosmological constant.
starting from a singular state at time t = 0 (C is an arbitrary constant). Its age (the proper time since the start
. This is a good model of the expansion
of the universe) when the Hubble constant takes the value H0 is τ0 = 3H
of the universe since radiation domination ended until the recent times when a cosmological constant started to
dominate the expansion. It is also a good model of the far future universe if k = 0 and Λ = 0.
2. The Milne model, for which {µ = p = 0, Λ = 0, k = −1} ⇒ Ω0 = 0, giving a linearly expanding empty
S(t) = C t.
This is just flat space-time as seen by a uniformly expanding set of observers ([178], pp. 360-363), singular at t = 0.
Its age is τ0 = H10 . It is a good model of the far future universe if k < 0 and Λ = 0.
3. The de Sitter universe, for which {µ = p = 0, Λ 6= 0, k = 0} ⇒ Ω0 = 0, giving the steady state expanding
empty solution:3
S(t) = C exp (Ht),
where C and H are constants. As the expansion rate is constant forever, there is no start and its age is infinite.4
It is a good model of the far future universe for those cases which expand forever with Λ > 0. It can alternatively
be understood as a solution with Λ = 0 and containing matter with the exceptional equation of state µ + p/c2 = 0.
There are other RW forms of the de Sitter Universe: a geodesically complete form with k = +1, S(t) = S0 cosh Ht
(a regular bounce), and another geodesically incomplete form with k = −1 , S(t) = S0 sinh Ht (a singular start).
This lack of uniqueness is possible because this is a spacetime of constant curvature, with no preferred timelike
directions or space sections [192, 110, 178].5
An initial singularity?
The above are specific models: what can one say generically? When the inequality (8) is satisfied, one obtains
directly from the Raychaudhuri equation (7) the
Friedmann-Lemaı̂tre Universe Singularity Theorem [45, 50]: In a FL universe with Λ ≤ 0
and µ + 3p/c2 > 0 at all times, at any instant t0 when H0 ≡ (Ṡ/S)0 > 0 there is a finite time t∗ :
t0 − (1/H0 ) < t∗ < t0 , such that S(t) → 0 as t → t∗ ; the universe starts at a space-time singularity
there, with µ → ∞ and T → ∞ if µ + p/c2 > 0.
This is not merely a start to matter — it is a start to space, to time, to physics itself. It is the most dramatic
event in the history of the universe: it is the start of existence of everything. The underlying physical feature is
the non-linear nature of the EFE: going back into the past, the more the universe contracts, the higher the active
gravitational density, causing it to contract even more. The pressure p that one might have hoped would help stave
off the collapse makes it even worse because (consequent on the form of the EFE) p enters algebraically into the
Raychaudhuri equation (7) with the same sign as the energy density µ. Note that the Hubble constant gives an
estimate of the age of the universe: the time τ0 = t0 − t∗ since the start of the universe is less than 1/H0 .
This conclusion can in principle be avoided by a cosmological constant, but in practice this cannot work
because we know the universe has expanded by at least a ratio of 11, as we have seen objects at a redshift6 of
10; from (14), the cosmological constant would have to have an effective magnitude at least 113 = 1331 times the
present matter density to dominate and cause a turn-around then or at any earlier time, and so would be much
bigger than its observed present upper limit (of the same order as the present matter density). Accordingly, no
turn around is possible while classical physics holds [47]. However energy-violating matter components such as a
scalar field (Sec.2.6) can avoid this conclusion, if they dominate at early enough times; but this can only be when
quantum fields are significant, when the universe was at least 1012 smaller than at present.
Because Trad ∝ S −1 (eqn.(6)), a major conclusion is that a Hot Big Bang must have occurred; densities and
temperatures must have risen at least to high enough energies that quantum fields were significant, at something
like the GUT energy. The universe must have reached those extreme temperatures and energies at which classical
theory breaks down.
The hot big bang
The matter and radiation in the universe gets hotter and hotter as we go back in time towards the initial quantum
state, because it was compressed into a smaller volume. In this Hot Big Bang epoch in the early universe, we can
3 The
Steady State universe of Bondi, Hold and Hoyle [14] utilised this metric, but was non-empty as they abandoned the EFE.
is however singular in that it is geodesically incomplete; this metric covers only half the de Sitter hyperboloid [192, 110].
5 There is also a static (non-RW) form of the metric — the first form of the metric discovered.
6 The redshift z for light emitted at t and observed at t is related to the expansion by 1 + z = S(t )/S(t ), see Sec.2.3.3.
4 It
use standard physical laws to examine the processes going on in the expanding mixture of matter and radiation
[233, 171]. A key feature is that about 300,000 years after the start of the Hot Big Bang epoch, nuclei and
electrons combined to form atoms. At earlier times when the temperature was higher, atoms could not exist,
as the radiation then had so much energy it disrupted any atoms that tried to form into their constituent parts
(nuclei and electrons). Thus at earlier times matter was ionized, consisting of negatively charged electrons moving
independently of positively charged atomic nuclei. Under these conditions, the free electrons interact strongly
with radiation by Thomson scattering. Consequently matter and radiation were tightly coupled in equilibrium at
those times, and the Universe was opaque to radiation. When the temperature dropped through the ionization
temperature of about 4000K, atoms formed from the nuclei and electrons, and this scattering ceased: the Universe
became very transparent (today we are able to see galaxies at enormous distances from us). The time when this
transition took place is known as the time of decoupling — it was the time when matter and radiation ceased
to be tightly coupled to each other, at a redshift zdec ≃ 1100 [38]. By (6), the universe was radiation dominated
(µrad ≫ µmat ) at early times and matter dominated (µrad ≪ µmat ) at late times;7 matter-radiation density equality
occurred significantly before decoupling (the temperature Teq when this equality occurred was Teq ≃ 104 K; at that
time the scale factor was Seq ≃ 104 S0 , where S0 is the present-day value). The dynamics of both the background
model and of perturbations about that model differ significantly before and after Seq [38].
Cosmic Blackbody Radiation
Radiation was emitted by matter at the time of decoupling, thereafter travelling freely to us through the intervening
space. When it was emitted, it had the form of blackbody radiation, because this is a consequence of matter and
radiation being in thermodynamic equilibrium at earlier times. Thus the matter at z = zdec forms the Last
Scattering Surface (LSS) in the early universe, emitting Cosmic Blackbody Background Radiation8 (‘CBR’) at
4000K, that since then has travelled freely with its temperature T scaling inversely with the scale function of the
universe.9 As the radiation travelled towards us, the universe expanded by a factor of about 1100; consequently
by the time it reaches us, it has cooled to 2.75 K (that is, about 3 degrees above absolute zero, with a spectrum
peaking in the microwave region), and so is extremely hard to observe. It was however detected in 1965, and its
spectrum has since been intensively investigated, its blackbody nature being confirmed to high accuracy [161]. Its
existence is now taken as solid proof both that the Universe has indeed expanded from a hot early phase, and that
standard physics applied unchanged at that era in the early universe.
Particle interactions and element formation
The thermal capacity of the radiation is hugely greater than that of the matter. At very early times before
decoupling, the temperatures of the matter and radiation were the same (because they were in equilibrium with
each other), scaling as 1/S(t) (eqn.(6)). The early universe exceeded any temperature that can ever be attained
on Earth or even in the centre of the Sun; as it dropped towards its present value of 3 K, successive physical
reactions took place that determined the nature of the matter we see around us today. At very early times and
high temperatures, only elementary particles can survive and even neutrinos had a very small mean free path; as
the universe cooled down, neutrinos decoupled from the matter and streamed freely through space. At these times
the expansion of the universe was radiation dominated, and we can approximate the universe then by models with
{k = 0, w = 1/3, Λ = 0}, the resulting simple solution of (14) uniquely relating time to temperature:
S(t) = S0 t
, t = 1.92 sec
1010 K
(There are no free constants in the latter equation).
At very early times, even neutrinos were tightly coupled and in equilibrium with the radiation; they decoupled
at about 1010 K ([38], pp. 44-46), resulting in a relic neutrino background density in the universe today of about
Ων0 ≃ 10−5 if they are massless (but it could be higher depending on their masses). Key events in the early universe
are associated with out of equilibrium phenomena ([38], p. 58). An important event was the era of nucleosynthesis,
the time when the light elements were formed. Above about 109 K, nuclei could not exist because the radiation was
so energetic that as fast as they formed, they were disrupted into their constituent parts (protons and neutrons).
However below this temperature, if particles collided with each other with sufficient energy for nuclear reactions to
take place, the resultant nuclei remained intact (the radiation being less energetic than their binding energy and
hence unable to disrupt them). Thus the nuclei of the light elements — deuterium, tritium, helium, and lithium
— were created by neutron capture. This process ceased when the temperature dropped below about 108 K (the
nuclear reaction threshold). In this way, the proportions of these light elements at the end of nucleosynthesis were
7 The
dynamically dominant Cold Dark Matter (Sec.2.3.6) obeys the same density law (6) as baryons.
is often called “Cosmic Microwave Background”, or CMB for short. However it is only microwave at the present epoch.
9 This scaling for freely propagating radiation follows from the discussion in Sec.2.3.3.
8 This
determined; they have remained virtually unchanged since. The rate of reaction was extremely high; all this took
place within the first three minutes of the expansion of the Universe. One of the major triumphs of Big Bang theory
is that theory and observation are in excellent agreement provided the density of baryons is low: Ωbar0 ≃ 0.044.
Then the predicted abundances of these elements (25% Helium by weight, 75% Hydrogen, the others being less than
1%) agrees very closely with the observed abundances. Thus the standard model explains the origin of the light
elements in terms of known nuclear reactions taking place in the early Universe [191]. However heavier elements
cannot form in the time available (about 3 minutes).
In a similar way, physical processes in the very early Universe (before nucleosynthesis) can be invoked to
explain the ratio of matter to anti-matter in the present-day Universe: a small excess of matter over anti-matter
must be created then in the process of baryosynthesis, without which we could not exist today (if there were no such
excess, matter and antimatter would have all annihilated to give just radiation [203]). However other quantities
(such as electric charge) are believed to have been conserved even in the extreme conditions of the early Universe,
so their present values result from given initial conditions at the origin of the Universe, rather than from physical
processes taking place as it evolved. In the case of electric charge, the total conserved quantity appears to be zero:
after quarks form protons and neutrons at the time of baryosynthesis, there are equal numbers of positively charged
protons and negatively charged electrons, so that at the time of decoupling there were just enough electrons to
combine with the nuclei and form uncharged atoms (it seems there is no net electrical charge on astronomical bodies
such as our galaxy; were this not true, electromagnetic forces would dominate cosmology, rather than gravity).
After decoupling, matter formed large scale structures through gravitational instability ([18], pp. 183-222)
which eventually led to the formation of the first generation of stars [203] and is probably associated with the
re-ionization of matter ([38], p. 73). However at that time planets could not form for a very important reason:
there were no heavy elements present in the Universe. The first stars aggregated matter together by gravitational
attraction, the matter heating up as it became more and more concentrated, until its temperature exceeded the
thermonuclear ignition point and nuclear reactions started burning hydrogen to form helium. Eventually more
complex nuclear reactions started in concentric spheres around the centre, leading to a build-up of heavy elements
(carbon, nitrogen, oxygen for example), up to iron. These elements can form in stars because there is a long time
available (millions of years) for the reactions to take place. Massive stars burn relatively rapidly, and eventually run
out of nuclear fuel. The star becomes unstable, and its core rapidly collapses because of gravitational attraction.
The consequent rise in temperature blows it apart in a giant explosion, during which time new reactions take place
that generate elements heavier than iron; this explosion is seen by us as a Supernova (“New Star”) suddenly blazing
in the sky, where previously there was just an ordinary star. Such explosions blow into space the heavy elements
that had been accumulating in the star’s interior, forming vast filaments of dust around the remnant of the star. It
is this material that can later be accumulated, during the formation of second generation stars, to form planetary
systems around those stars. Thus the elements of which we are made (the carbon, nitrogen, oxygen and iron nuclei
for example) were created in the extreme heat of stellar interiors, and made available for our use by supernova
explosions. Without these explosions, we could not exist.
Cosmological Observations
Cosmological models only become meaningful when related to astronomical observations [114, 188, 50, 233]. These
are of two kinds: astronomical observations of distant matter tells us what was happening far away in the universe
and (because of the finite speed of light) a long time ago. On the other hand observations of nearby objects (matter
on Earth, the solar system, nearby stars for example) when related to theories of origins tell us what was happening
very near our past world line a very long time ago. The first set of observations may be characterized as “null
cone” observations, the second as “geological” observations, one of the most important being the determination of
local element abundances, which are then related to nucleosynthesis calculations (Sec.2.2.2).
Observations are totally dependent on telescope and detector technology [106, 18]. After the initial establishment of distance scales and the basic evidence of cosmic homogeneity and expansion in the 1920s and 1930s,
progress was slow until the 1960s when observations were extended from the optical to the entire electromagnetic
spectrum. In recent decades cosmology has changed from a data-poor to a data-rich subject. Massive new data
sets are now available because of the extraordinary improvement of telescope, detector, and computer technology
in recent decades, particularly the advent of new detectors such as Charge Coupled Devices (CCD’s) and fibre
optics (enabling simultaneous measurement of hundreds of redshifts). We now have not only optical, ultraviolet,
and infrared observations of galaxies, determining luminosities and spectra with unprecedented sensitivity, but also
radio, X-ray, and gamma-ray sky surveys. Galaxies have been detected up to a redshift of 6 and we have identified
many quasi-stellar objects and gamma-ray bursters as well as multiple images of very distant gravitationally-lensed
galaxies [106]. Large-scale structures (clusters of galaxies, superclusters, walls, and voids) have been identified,
with associated large-scale velocity flows ([18], pp. 85-137).
In addition to large-scale number-count and redshift surveys, we have measured the background radiation
spectrum and anisotropies at all wavelengths. We identify the radiation as ‘background’ precisely when it is
constant on very small angular scales (as opposed to discrete sources, which appear as isolated objects). There is
a complex relation of this radiation to the intergalactic matter density and thermal history. The most important
component of the background radiation is the Cosmic Blackbody Radiation (‘CBR’) mentioned above (Sec.2.2);
detailed observations have mapped its temperature over the whole sky at a sensitivity of better than one part in
105 . However other components of the background radiation (X-ray and radio in particular) convey important
information on the temperature and density of intergalactic matter, and hence strongly restrict its possible thermal
history. For example hot matter emits X-rays, so the X-ray background measurement restricts the amount of hot
intergalactic matter allowed; while neutral hydrogen strongly absorbs Ultra-Violet radiation to give the Lyman
alpha spectral absorption line, so absence of such absorption gives strong limits on the amount of neutral hydrogen
and hence on the temperature of intergalactic matter.
The first important point about cosmological observations is that when averaged on a large enough physical scale
(clusters of galaxies and above) they are statistically isotropic about us; there is no direction apparently pointing to
the centre of the universe. The high degree of isotropy of the CBR strongly supports this conclusion: its temperature
is the same in all directions about us to better than one part in 10,000 after we have allowed for the motion of the
Earth relative to the cosmos (about 250 km/sec), which creates a temperature dipole at one part in a thousand.10
Any inhomogeneities or anisotropies in the matter distribution lead to anisotropies in this radiation, as recently
measured at only one part in 105 by the extremely sensitive detectors of the COBE and WMAP satellites. This
high degree of isotropy is the major reason we believe the Universe itself is spatially homogeneous and isotropic to
a good approximation (see Sec.4.2.2), providing good observational support for use of the FL universe models as
accurate models of the observed region of the universe.
Distance scale and ages
The underlying problem in all astronomy is determining the distances of observed objects. This is done by a
‘cosmic distance ladder’ ([18], pp. 25-83) whereby nearest objects have their distance determined by parallax (i.e.
essentially by local trigonometry); and more distant ones by a series of consecutive distance indicators (Cepheid
variables, RR Lyrae variables, brightest red supergiants) until at a cosmological distance, redshift z is a primary
distance indicator, but is contaminated by local velocities of matter relative to the rest-frame of the universe.
Other distance indicators (for example the Tully-Fisher method, the luminosity function of planetary nebulae, the
globular cluster luminosity function, surface brightness fluctuations) serve to refine the estimates [18].
Closely associated with the distance scale is determination of the Hubble constant H0 (the present rate of
expansion of the universe), because estimates of the size of the observable region of the universe scale with the
Hubble constant. But the Hubble constant also determines the age of the universe, so its determination underlies a
crucial consistency condition for cosmology: the age of objects in the universe (rocks, planets, stars, star clusters,
galaxies) must be less than the age of the universe. This condition has been a cause of concern ever since we have
had good estimates of ages and of the Hubble constant.11 It seems not to be violated by current observations of
low redshift objects given the current estimates of H0 ≃ 70 km/sec/Mpc, giving an age of the universe of about
15 billion years whereas the oldest star clusters seem to be about 14 billion years old. However it is very tight,
perhaps even problematic, for very distant (and so much younger) objects [120].
Observational relations
Light travels on null geodesics xa (λ) in spacetime (the tangent vector k a := dxa /dλ is such that k a;b k b = 0,
k a ka = 0). In a RW geometry, it suffices to consider only radial null geodesics (by the symmetries of the model,
these are equivalent to generic geodesics). Then from (2) we find that for light emitted at time te and received at
time t0 , the comoving radial distance u(t0 , t1 ) := r0 − r1 between comoving emitters and receivers is given by
{ds = 0, dθ = 0 = dφ} ⇒ u(t0 , t1 ) =
with Ṡ given by the Friedmann equation (9). The key quantities related to cosmological observations are redshift,
area distance (or apparent size), and local volume corresponding to some increment in distance (determining number
10 The CBR dipole that could be interpreted as due to a major cosmological inhomogeneity is rather interpreted as being due to our
motion (‘peculiar velocity’) relative to a spatially homogeneous universe.
11 Indeed Hubble himself never fully accepted the expanding universe theory because of age difficulties, preferring to refer to a
redshift-distance relation rather than a velocity distance relation [115]. However the problem has been eased by a series of revisions of
the value of the Hubble constant since then, due to a better understanding of the primary distance indicators.
counts) [188, 50, 233]. The redshift z measured for an object emitting light of wavelength λe that is observed with
wavelength λ0 is given by
= (1 + zc )(1 + zv ),
1 + z :=
where zv is the redshift caused by the local peculiar motion of the object observed (zv = 0 for comoving objects),
and zc is the cosmological redshift given by
S(t0 )
1 + zc =
S(te )
From eqn.(21), the same ratio of observed to emitted light holds for all wavelengths: a key identifying property of
redshift. The problem in using redshifts of objects as a distance indicator is to separate out the cosmological from
the Doppler components, which lead to redshift-space distortions ([38], pp. 275-282); this can reasonably be done
for a cluster of galaxies by appropriate averaging over cluster members (hzv i = 0 for a comoving cluster). The area
distance r0 of an object at redshift zc and of linear size l which subtends angular size α is given by12
r0 (zc ) :=
= f (u)S(te ).
Thus measures of apparent sizes will determine the area distance if the source physical size is known. The flux of
radiation F measured from a point source of luminosity L emitting radiation isotropically is given by the fraction
of radiant energy emitted by the source in a unit of time that is received by the telescope:
F =
4π f 2 (u)S 2 (t0 )(1 + z)2
(the two redshift factors account firstly for time dilation observed between observer and source, and secondly
for loss of energy due to redshifting of photons). The source’s apparent magnitude m is defined from the flux:
m = −2.5log10F + const. On using (22, 23), equation (24) becomes
F =
4π r0 (1 + z)4
showing that measures of magnitudes will determine the area distance if the source’s intrinsic luminosity is known.
On using (23) it follows from (25) that the point-wise surface brightness of extended objects (the flux received
per unit solid angle) depends only on redshift [128, 50] — a key feature in determining detection probabilities
and in gravitational lensing observations. It further follows from this result that a blackbody spectrum emitted
at temperature Te when observed with a redshift z remains a blackbody spectrum but with observed temperature
T0 = Te /(1 + z) — a crucial feature in analyzing the CBR observations.
Using the Friedmann equation and the relevant equation of state for matter, the area distance can be determined as a function of redshift zc in terms of the Hubble constant H0 , deceleration parameter q0 , and cosmological
constant Λ. In the case of pressure-free matter with vanishing cosmological constant, one obtains from (20), (9),
(22), and (23)13 the Mattig relation [188]
r0 (zc ) =
1 (q0 − 1)(1 + 2q0 zc )1/2 + (q0 (zc − 1) + 1)
H0 q02
(1 + zc )2
Consequently measures of either apparent size of sources of known physical size, or of radiant flux from sources
of known intrinsic luminosity, will determine the deceleration parameter q0 . Generalizations of this relation hold
if a cosmological constant or radiation is present. An interesting aspect is that there is a minimum apparent
size for objects of fixed physical size at some redshift zc = z∗ depending on the density parameter and the
cosmological constant. The past light cone of the observer attains a maximum area at z∗ ; the entire universe
acts as a gravitational lens for further off objects, magnifying their apparent size so that very distant objects
can appear to have the same angular size as nearby ones [114, 188, 50]. For the Einstein-de Sitter universe, the
minimum angular diameter is at z∗ = 1.25; in low density universes, it occurs at higher redshifts.
The number of objects seen in a solid angle dΩ for a distance increment du (characterized by increments dz,
dm in the observable variables z and m) is given by
dN = W (u) ρ(te )S 3 (te )f (u)dudΩ
where the detection probability or ‘selection function’ is W (u) ([38], p. 263) and ρ(te ) is the number density of
objects at the time of emission (spatial homogeneity is expressed in the fact that this is independent of the spatial
coordinates). The observed total number N of objects measured in a survey is given by integrating from the
observer to the survey limit: in terms of the radial coordinate re of the source (which can be related to redshifts
12 This
13 Or,
depends only on zc because apparent shapes and sizes are independent of the motion of the source.
more elegantly, from the geodesic deviation equation (see [84]).
or magnitudes), N = r0e dN . If the number of objects is conserved (e.g. observing galaxies in an epoch when they
are neither created nor destroyed), ρ(te ) = ρ(t0 )(1 + z)3 and we find from (27) that in the idealized case when W
is independent of distance (a reasonable assumption for relatively nearby objects),
Z re
f (u)du.
N = W ρ(t0 )dΩ
The simple integral has to be separately done for the cases k = +1, 0, −1 [188].
The above equations enable one to determine observational relations between observable variables, for example (m, z), (α, z) or (N, m) relations for objects with known intrinsic properties (known size or luminosity, for
example), and so to observationally determine q0 . These relations have to be modified if there is absorption by
an intergalactic medium, gravitational lensing, or anisotropic emission of radiation; and detailed comparisons with
observations have to take into account the spectrum of the source as well as source detection and identification
probabilities [106]. Here we encounter the contrast between image and reality : there can be many objects out there
that we either do not detect, or do not recognize for what they are [37]. An “observational map” relating source
properties to the nature of their images gives a useful view of how this occurs [75].
One important feature here is that a specific object will look completely different at different wavelengths
(optical, radio, X-ray for example); indeed it may be detectable at one wavelength and not at another. This shows
very clearly how our images of reality are dependent on the detectors we use. To get a full picture of what is
out there, we need to use multiple modes of investigation — imaging at all wavelengths together with intensity,
spectral, and polarization measurements [106], as well as watching for time variations. A second important feature
is observational selection effects such as the Malmquist bias — if we have a population of objects with different
luminosities, at large distances we will only see the more luminous objects (the fainter ones will not be detected);
hence the average luminosity will appear to increase with distance, but this is just an observational effect rather
than the real state of affairs. Using different detection thresholds controls this effect to some degree.
Number Counts and the visible matter density
Number counts of galaxies as a function of redshift or luminosity show approximate spatial homogeneity of the
universe [115]. However counts of radio sources and quasi-stellar objects (qso’s) show that the universe has not
been in a steady state as proposed by Bondi, Gold, and Hoyle [14]. Indeed number counts are only compatible with
a RW geometry if there has been evolution of source numbers and/or luminosities [193].
Number counts also give estimates of the density of visible (luminous) matter in the universe: Ωvm0 ≃ 0.015.
This is very low relative to the critical density (Ω0 = 1) and is also considerably less than the amount of baryons
determined by nucleosynthesis studies (Ωbar0 ≃ 0.044, see Sec.2.2.2). Thus much of the baryonic matter in the
universe is in some hidden (non-luminous) form ([18], pp. 223-272)), e.g. burnt out stars [112].
Apparent Luminosities and sizes: Dark Energy
Apparent sizes or luminosities as a function of redshift can be used to determine the deceleration parameter q0
(Sec.2.1) if the intrinsic source sizes or luminosities are known. The problem is that until recently there were
not known enough galaxies or other objects of standard size or luminosity to use to determine q0 , and scatter in
their properties leads to biassing of observations by the Malmquist effect (Sec.2.3.3). However this dramatically
changed with recent observations of the decay curves of the luminosity of supernovae in distant galaxies. It turns
out that the peak luminosity of type Ia supernovae is closely correlated with their light curve decay time, for
the first time giving reliable ‘standard candles’ for galaxies at large distances [172]. The conclusion from these
observations is that, rather than slowing down as expected, the rate of expansion of the universe is speeding up at
a rate corresponding to a cosmological constant with ΩΛ0 = 0.7. This evidence is concordant with that from CBR
observations (Sec.2.3.7) and number counts [38, 203].
The nature of the field or matter causing this acceleration is unclear. Its equation of state w := pc2 /µ, is
unknown, and many physical and unphysical proposals are being made in this regard. From (7), it has to violate the
strong energy condition (8) and so must have a large negative pressure. It could indeed be due to a cosmological
constant (w = −1), which would have dominated the expansion of the universe since a redshift z ≃ 0.33, and
would have been negligible earlier (and is also negligible on small scales — it does not affect local astrophysics).
However it could also be some other form of matter or field with effective negative pressure so that w < −1/3,
as can happen in the case of a scalar field (see eqn.(33) below). In that case it is called ‘quintessence’. There
are many speculations as to what this might be, but there is no clarity on the matter. One should note here
that alternative explanations of the observations are possible, for they can be exactly reproduced by a spherically
symmetric inhomogeneous universe model where we are near the centre [157], or could at least partly be due to the
back-reaction of inhomogeneities on the cosmic expansion [67] or the effect of inhomogeneities on the effective area
distance [122, 123]. These alternatives are being investigated, but the most probable cause remains some unknown
kind of matter or field with effective negative energies.
In summary, the standard gravitational equations together with the supernovae observations imply presence
of a cosmological constant or some equivalent form of ‘dark energy’ with a large effective negative energy density
µgrav (due to negative pressure) dominating the present expansion of the universe; its physical nature is unknown.
There is no known physics reason why this force should exist at this level where it is just detectable — quantum
field theory relates the cosmological constant to the zero-point energy of the vacuum, and suggests it should be
enormously larger than observed [234, 235, 186, 243, 212]. It is a major mystery why it exists at the small (just
detectable) level that observations indicate [197]. A key aspect of present day cosmology is trying on the one
hand to observationally determine the effective equation of state of this ‘dark energy’ (running the field equations
backwards to obtain w(z) from the observations [187], and in particular determining whether w is constant or
varying over time), and on the other attempting to give a plausible theoretical explanation for its physical origin.
Matter Distribution and Motion: Dark Matter
Detailed studies have been made of the distribution of galaxies and their motions. They occur in clusters, in
turn making up superclusters imbedded in vast walls surrounding relatively sparsely populated intergalactic voids.
The galaxy luminosity function characterizes the numbers of galaxies occurring within each luminosity class; the
covariance function characterizes their spatial clustering [165, 38]. Large scale motions occur for galaxies in clusters,
and for the clusters themselves. It is easy to conceive of matter that is hard to detect (for example, small rocks
distributed through space); studies of galactic rotation curves and of motions of galaxies in clusters ([18], pp.
139-181) imply the existence of huge amounts of unseen dark matter, dominating the dynamics of the Universe:
its density is Ωdm0 ≃ 0.3, much greater than both visible matter (Ωvm0 = 0.015) and baryons (Ωbar0 = 0.044), but
significantly less than the critical density Ω0 = 1. Thus the dark matter is non-baryonic, meaning it has some kind
of exotic nature rather than being the protons and neutrons that are the substance of ordinary matter [197]. In
contrast to the ‘dark energy’ discussed in the previous section, dark matter is dynamically effective on astrophysical
scales as well as on cosmological scales. Many attempts have been made to identify its nature, for example it might
be axions, supersymmetric partners of known particles, quark nuggets, or massive neutrinos [96, 171], but what it
is composed of is still unknown. Laboratory searches are under way to try to detect this matter, so far without
success. A key question is whether its apparent existence is due to our using the wrong theory of gravity on
these scales. This is under investigation, with various proposals for modified forms of the gravitational equations
that might explain the observations without the presence of large quantities of dark matter. This is a theoretical
possibility, but the current consensus is that this dark matter does indeed exist.
An important distinction is whether dark matter consists of
(i) weakly interacting massive particles that cooled down quickly, thereafter forming cold dark matter
(‘CDM’) moving slowly at the time of structure formation (and resulting in a bottom-up process with large scale
structure forming from smaller scale structures), or
(ii) particles that have a low mass and cooled slowly, therefore for a long time forming hot dark matter,
moving very fast at the time of structure formation (and resulting in a top-down galaxy formation scenario).
Structure formation studies currently support the CDM hypothesis, with hierarchical formation of gravitationally
bound objects taking place in a complex bottom up process involving interactions of CDM, baryons, and radiation,
with dwarf galaxies forming initially [203, 154] and then aggregating to form larger structures. These studies are
based on massive numerical simulations, with initial conditions provided by the inflationary scenario discussed
below, see Sec.2.6. Unlike ‘dark energy’, CDM has an ordinary baryonic equation of state (it is a perfect fluid (4)
with pcdm = 0 ⇔ wcdm = 0).
Another way of detecting dark matter in clusters is by its gravitational lensing effects [190]. The bending
of light by massive objects was one of the classic predictions of General Relativity theory. Rich clusters of galaxies
or galaxy cores can cause strong lensing of more distant objects, where multiple images of distance galaxies and
qso’s occur, sometimes forming rings or arcs; and weaker lensing by closer masses results in characteristic patterns
of distortions of images of distant objects. Analysis of multiple images can be used to reconstruct the lensing
mass distributions, and statistical analysis of weak lensing patterns of image distortions are now giving us detailed
information on the matter distribution in distant galaxies and clusters. These studies show that to get enough
lenses in an almost flat cosmology (Ω0 ≃ 1) requires the presence of a cosmological constant — there cannot be a
critical density of dark matter present [38, 203].
A key feature of present-day cosmology is attempts to identify the nature of this dark matter, and if possible
to detect it in a laboratory situation. While observations favour the CDM scenario, some residual problems as
regards the emergence of fine-scale structure still need resolution [203].
The CBR Power spectrum
The CBR angular anisotropies are characterized by an angular power spectrum showing the amount of power in
perturbations at each physical scale on the LSS [8, 197, 38]. In the time from the the end of inflation to the LSS,
modes of different wavelengths can complete a different number of oscillations. This translates the characteristic
periods of the waves into characteristic lengths on the LSS, leading to a series of maxima (‘acoustic peaks’) and
minima in the inhomogeneities on the LSS and consequently in the CBR angular anisotropy power spectrum
[118, 162, 171]. These inhomogeneities then form the seeds for structure formation and so are related to the power
spectrum of physical scales for structures that form later. They are characterised by a (3-dimensional) spatial
power spectrum on the LSS; because we receive the observed CBR radiation from the 2-sphere S2:LSS where our
past light cone intersects the LSS, this is seen by us as a 2-dimensional power spectrum of anisotropies on the sky
(characterised by the unit sphere S2 of all direction vectors ea : ea ea = 1, ea ua = 0).
The apparent angular size of the largest CBR peak (about 1o ) allows estimates of the area distance to
the LSS and hence of the density of matter in the universe for various values of the cosmological constant, and
determines the major cosmological parameters [206]:
“By combining WMAP data with other astronomical data sets, we constrain the geometry of the universe: Ωtot = 1.02 ± 0.02, the equation of state of the dark energy, w < −0.78 (95% confidence limit),
and the energy density in neutrinos, Ων h2 < 0.0076 (95% confidence limit). For 3 degenerate neutrino
species, this limit implies that their mass is less than 0.23 eV (95% confidence limit). The WMAP
detection of early reionization rules out warm dark matter.”
There is however a problem here: while the agreement of theory and observations for all small angular scales is
remarkable, there is a divergence at the largest angular scales: the observations show less power than expected.
Specifically, the quadrupole and octopole are much lower than theory predicts. Also the axes of the quadrupole
and octopole are very precisely aligned with each other, and there are other angular anomalies [207]. These effects
might be due to (i) observational contamination by the galaxy (which gets in the way of our view of the LSS),
(ii) a contingent (‘chance’) event (it represents ‘cosmic variance’, discussed below, see Sec.3), (iii) our living in a
well-proportioned ‘small universe’ which is spatially closed so that there is a maximum size to possible fluctuations
[232], or (iv) some unexpected new physical effect or deeper problem with our understanding of the early universe.
The jury is out as to which the case is; this could turn out to be a crisis for the CBR analysis, but on the other
hand one can always just resort to saying it is a statistical fluke (the underlying problem here being the uniqueness
of the universe, as discussed in Sec.3).
There are similar expected peaks in the polarization spectrum of this radiation, and polarization maps
should have a mode associated with gravitational waves predicted by inflation to exist in the very early universe
(Sec.2.6); detection of such modes will be a crucial test of inflation [38, 199]. Studies of polarization indicate that
reionisation of the universe took place as early as a redshift of 17, contrary to what is deduced from qso studies.
More detailed studies of anisotropies involve the Sunyaev-Zel’dovich effect (changes in the observed temperature
due to scattering by hot matter in galaxy clusters) and gravitational lensing.
There is a huge amount of information in the CBR maps, and their more accurate measurement and
interpretation is a central feature of current cosmology [208, 162, 38, 171].
Causal and visual horizons
A fundamental feature affecting the formation of structure and our observational situation is the limits arising
because causal influences cannot propagate at speeds greater than the speed of light. Thus the region that can
causally influence us is bounded by our past null cone. Combined with the finite age of the universe, this leads to
the existence of particle horizons limiting the part of the universe with which we can have had causal connection.14
A particle horizon is by definition comprised by the limiting worldlines of the furthest matter that ever
intersects our past null cone [177, 178]. This is the limit of matter that we can have had any kind of causal contact
with since the start of the universe, characterized by the comoving radial coordinate value
Z t0
uph =
The present physical distance to the matter comprising the horizon is
dph = S(t0 )uph .
The key question is whether the integral (29) converges or diverges as we go to the limit of the initial singularity
where S → 0. Horizons will exist in standard FL cosmologies for all ordinary matter and radiation, for uph will
14 There
are also event horizons and apparent horizons in some cosmological models ([177]; [219]; [178], pp. 376-383).
be finite in those cases; for example in the Einstein-de Sitter universe (see Sec.2.1.1), uph = 3t0 , dph = 3t0 . We
will then have seen only a fraction of what exists, unless we live in a universe with spatially compact sections so
small that light has indeed had time to traverse the whole universe since its start; this will not be true for universes
with the standard simply-connected topology. Penrose’s powerful use of conformal methods (see [110, 219]) gives a
very clear geometrical picture of the nature of horizons [81]. They may not exist in non-FL universes, for example
Bianchi (anisotropic) models [152]. In universes with closed spatial sections, a supplementary question arises: Is
the scale of closure smaller than the horizon scale? There may be a finite time when causal connectivity is attained,
and particle horizons cease to exist. In standard k = +1 FL models, this occurs just as the universe reaches the
final singularity; if however there is a positive cosmological constant or other effective positive energy density field,
it will occur earlier. The horizon always grows, because (29) shows that uph is a monotonically increasing function
of t0 . Despite many contrary statements in the literature, it is not possible that matter leave the horizon once it has
entered. In a (perturbed) FL model, once causal contact has taken place, it remains until the end of the universe.
The importance of horizons is two-fold: they underlie causal limitations relevant in the origin of structure
and uniformity [152, 100], and they represent absolute limits on what is testable in the universe [52, 54].
Causal limitations
As to causal limitations, horizons are important in regard both to the smoothness of the universe on large scales,
and the lumpiness of the universe on small scales. The issue of smoothness is encapsulated in the horizon problem:
if we measure the temperature of the CBR arriving here from opposite directions in the sky in a standard FL
model, it came from regions of the surface of last scattering that can have had no causal contact of any kind
with each other since the start of the universe. In a radiation-dominated early universe with scale factor (19),
the size of the particle horizon at the time of last scattering appears as an angular scale of about 1o in the sky
today, and corresponds to a comoving physical length of about 400,000 light years when evaluated today. Why
then are conditions so similar in these widely separated regions? [151, 100, 11, 126]. Note that this question is of
a philosophical rather than physical nature, i.e. there is no contradiction here with any experiment, but rather an
unease with an apparent fine tuning in initial conditions. This problem is claimed to be solved by the inflationary
universe scenario mentioned below, see Sec.2.6.
Associated with the existence of horizons is the prediction that physical fields in different regions in the
universe should be uncorrelated after symmetry breaking takes place, because they cannot have interacted causally.
Consequently, if grand unified theories are correct, topological defects such as monopoles and cosmic strings may
be expected as relics of the expansion of the very early universe [126]. In a standard cosmology, far too many
monopoles are predicted. This is also solved by inflation.
As to the lumpiness, the issue here is that if we believe there was a state of the universe that was very
smooth — as indicated at the time of decoupling, by the low degree of anisotropy of the CBR, and represented by
the RW geometry of the FL models — then there are limits to the sizes of structures that can have grown since then
by causal physical processes, and to the relative velocities of motion that can have been caused by gravitational
attraction in the available time (for example, the peculiar motion of our own galaxy relative to the CBR rest
frame caused by the huge overdensity called the ‘Great Attractor’). If there are larger scale structures or higher
velocities, these must have been imprinted in the perturbations at the time of last scattering, for they cannot have
been generated in a causal way since that time. They are set into the initial conditions, rather than having arisen
by physical causation from a more uniform situation. This is a key factor in the theory of growth of perturbations
in the early universe where the expansion damps their growth. The quantity determining the relevant physical
scales for local causal influences in an expanding universe is the comoving Hubble radius λH := (SH)−1 ; the way
perturbations of wavelength λ develop depends on whether λ > λH or λ < λH ([38], pp. 146-150).
Actually the domain of causal influence is even more tightly constricted than indicated by the past light
cone: the limits coming from the horizon size are limits on what can be influenced by particles and forces acting
at the speed of light. However only freely travelling photons, massless neutrinos, and gravitons can move at that
speed; and such particles coming from cosmological distances have very little influence on our galaxy or the solar
system (indeed we need very delicate experiments to detect them). Any massive particles, or massless particles
that are interacting with matter, will travel slower (for example before decoupling, light has a very small mean
free path and information will travel only by sound waves and diffusion in the tightly coupled matter-radiation
fluid). The characteristics for pressure-free scalar and vector perturbations are timelike curves, moving at zero
velocity relative to the matter; while density perturbations with pressure can move at the speed of sound, only
tensor perturbations can travel at the speed of light. Thus the true domain that influences us significantly is much
less than indicated by the particle horizon. It is the small region round our past world line characterised after
decoupling by the comoving scale from which matter coalesced into our galaxy: a present distance of about 1 to
1.95 Mpc,15 corresponding to an observed angle of about 0.64 arcminutes on the LSS. Before decoupling it would
have been limited by the sound horizon ([38], p. 257) rather than the particle horizon.
15 Dodelson
[38], p. 283; W Stoeger, private communication.
Observational limitations
Clearly we cannot obtain any observational data on what is happening beyond the particle horizon; indeed we
cannot even see that far because the universe was opaque before decoupling. Our view of the universe is limited by
the visual horizon, comprised of the worldlines of furthest matter we can observe — namely, the matter that emitted
the CBR at the time of last scattering [79, 183]. This occurred at the time of decoupling t = tdec (zdec ≃ 1100),
and so the visual horizon is characterized by r = uvh where
Z t0
uvh =
< uph .
Indeed the LSS delineates our visual horizon in two ways: we are unable to see to earlier times than its occurrence
(because the early universe was opaque for t < tdec ), and we are unable to detect matter at larger distances than
that we see on the LSS (we cannot receive radiation from matter at co-moving coordinate values r > uvh ). The
picture we obtain of the LSS by measuring the CBR from satellites such as COBE and WMAP is just a view of
the matter comprising the visual horizon, viewed by us at the time in the far distant past when it decoupled from
radiation. The position of the visual horizon is determined by the geometry since decoupling. Visual horizons do
indeed exist, unless we live in a small universe, spatially closed with the closure scale so small that we can have seen
right around the universe since decoupling. This is a possibility that will be discussed below (Sec.4.3.1). There is no
change in these visual horizons if there was an early inflationary period, for inflation does not affect the expansion
or null geodesics during this later period. The major consequence of the existence of visual horizons is that many
present-day speculations about the super-horizon structure of the universe — e.g. the chaotic inflationary theory
(Sec.2.6) — are not observationally testable, because one can obtain no definite information whatever about what
lies beyond the visual horizon [52, 54]. This is one of the major limits to be taken into account in our attempts to
test the veracity of cosmological models (Sec.4.3).
Theoretical Developments
The cosmological application of Einstein’s Theory of Gravitation has also progressed greatly in past decades, as
regards exact solutions and generic properties of the field equations; as regards approximate solutions; and in terms
of understanding the relationship between them.
Exact solutions and generic properties
Theory initially predicted there must have been a start to the universe, but it was not clear for a long time if
this was simply due to the very special exactly isotropic and spatially homogeneous geometry of the standard
Friedmann-Lemaı̂tre models. It was possible that more realistic models with rotation and acceleration might show
the prediction was a mathematical artefact resulting from the idealized models used. The singularity theorems
developed by Penrose and Hawking [110, 219, 41] showed this was not the case: even for realistic geometries,
classical gravitational theory predicts a beginning to the universe at a space-time singularity, provided the usual
energy conditions were satisfied. This study has led inter alia to a greatly increased understanding of causality and
topology of generic universe models [219], including the fact that singularities may have a quite different nature
than those in the Robertson-Walker models, for example being anisotropic [219] or of a non-scalar nature [68].
Various classes of exact cosmological solutions are known (Kantowski-Sachs and Bianchi spatially homogeneous but anisotropic models, Tolman-Bondi spherically symmetric inhomogeneous models, and ‘Swiss-Cheese’
non-analytic models) enabling understanding of dynamical and observational behaviour of more general classes of
models than just the FL models [83]. Dynamical systems studies [226, 224] relate the behaviour of whole classes of
anisotropic models in suitable state spaces, enabling identification of generic patterns of behaviour (fixed points,
saddle points, attractors, etc.) and hence the relationship between dynamics of higher symmetry and lower symmetry universes. These studies help understanding to what degree the FL models are generic within the families
of possible cosmological models, and which models might give observations similar to those in the FL models. In
particular they are relevant in considering the possible geometry of the universe at very early or very late times.
Perturbation theory, the gauge issue, and back reaction
Sophisticated perturbation theory has been developed to underlie the theory of structure formation in the expanding
universe, examining the dynamics of perturbed FL models. The fluid flow in these models can have shear, vorticity,
and acceleration, and the Weyl tensor Cijkl (see (1)) is not zero, so that density variations, tidal forces, peculiar
velocities, and gravitational waves can be present. Detailed studies use the kinetic theory approximation for
matter (electrons, protons, dark matter) and radiation (photons, neutrinos), with their dynamics described by the
Boltzmann equation ([38], Ch.4; [223]), interacting with space-time inhomogeneities characterised by a perturbed
FL metric. A key issue here is the gauge problem — how to choose the background model in the perturbed
spacetime [78]. If this is not properly handled then one may attain apparent perturbation solutions that are pure
gauge (they are mathematical rather than physical), so that one can alter the apparent growth rate simply by
changing coordinates. The key to handling this is either to keep careful track at all stages of remaining gauge
freedom and possible changes of gauge, or (preferably, in my view) to use gauge invariant variables (see [4, 66, 24]).
Most of the literature on perturbation theory deals with the linear case, but some studies tackle the nonlinear regime (e.g. [132]), and some consider questions such as the origin of magnetic fields and the causes of
galactic rotation. A key problem here is properly relating relativistic analyses of astrophysical dynamics to the
Newtonian approaches most often used by astrophysicists (e.g. [18], pp. 183-222); this is not straightforward.16 A
further unresolved issue is the nature of gravitational entropy [168, 61, 169]. Many statements about the nature
of entropy in physics textbooks are wrong when gravity is dominant, leading to the spontaneous formation of
structures such as stars and galaxies. There is as yet no agreed definition of gravitational entropy that is generally
applicable; until there is, cosmological arguments relying on entropy concepts are ill-founded.
The existence of inhomogeneities in the universe raises the issue of fitting and back-reaction. To what degree
does the nature of the exactly smooth FL models reflect the geometrical and dynamical nature of more realistic
‘lumpy’ universe models? [78]. Inhomogeneities lead to extra terms appearing in the evolution equations for the
idealized background models, representing the back-reaction of the perturbations on their dynamics [55]. These
could possibly be dynamically significant [67], but this is a matter of dispute.
Particle physics processes dominated the very early eras, when exotic processes took place such as the condensation
of a quark-gluon plasma to produce baryons. Quantum field theory effects were significant then, and this leads to an
important possibility: scalar fields producing repulsive gravitational effects could have dominated the dynamics of
the universe at those times. This leads to the theory of the inflationary universe, proposed by Alan Guth [100, 101]:
if µgrav = µ + 3p/c2 < 0, which can happen if a scalar field dominates the dynamics of the early universe, an
extremely short period of accelerating expansion will precede the hot big bang era [11]. This produces a very cold
and smooth vacuum-dominated state, and ends in ‘reheating’: conversion of the scalar field to radiation, initiating
the hot big bang epoch. This inflationary process is claimed to explain the puzzles mentioned above (Sec.2.4.1):
why the universe is so special (with spatially homogeneous and isotropic geometry and a very uniform distribution
of matter), and also why the space sections are so close to being flat at present (we still do not know the sign of the
spatial curvature), which requires very fine tuning of initial conditions at very early times. Inflationary expansion
explains these features because particle horizons in inflationary FL models will be much larger than in the standard
models with ordinary matter, allowing causal connection of matter on scales larger than the visual horizon, and
inflation also will sweep topological defects outside the visible domain.
In more detail: in the case of a single scalar field φ with spacelike surfaces of constant density, on choosing
ua orthogonal to these surfaces, the stress tensor has a perfect fluid form with
1 2
φ̇ + V (φ), p/c2 = φ̇2 − V (φ),
and so
µ + 3p/c2 = 2φ̇2 − 2V (φ).
The slow-rolling case is φ̇ ≪ V (φ), leading to µ + p/c = 2φ̇ ≃ 0 ⇒ µ + 3p/c ≃ −2µ < 0. This then enables a
resolution of the horizon problem in inflationary FL models: if sufficient inflation took place in the early universe,
then all the regions from which we receive CBR were causally connected; indeed if the universe began in an
inflationary state, or was inflationary with compact spatial sections, there may be no causal horizons at all. The
inflationary models also cause initial perturbations to die away, including velocity perturbations, hence explaining
the observed smoothness of the universe on large scales. This process is expected to create a universe with very
flat spatial sections at late times:
Ω0 = Ωdm 0 + ΩΛ 0 ≃ 1 ⇔ Ωk ≃ 0.
This theory led to a major bonus: a proposal that initial tiny quantum fluctuations were expanded to such a
large scale by inflation that they provided seeds initiating growth by gravitational attraction of large scale structures
such as clusters of galaxies. This theory makes clear observational predictions for the spectrum of CBR anisotropies,
which have since been spectacularly verified by observations from balloons and satellites, such as WMAP [206].
Thus inflation has provided us with our first coherent theory of structure formation. Inhomogeneities started
16 Some exact General Relativity results, which must necessarily apply in the Newtonian limit of General Relativity, have no Newtonian
analogue; an example is the shear-free theorem applying to pressure-free matter [49]. The underlying issue is that there are 10 field
equations to be satisfied in General Relativity, with 20 integrability conditions (the Bianchi identities), but only one field equation to
be satisfied in Newtonian theory (Poisson’s equation) together with 4 conservation equations.
as quantum fluctuations in the inflationary epoch which are then amplified in physical scale by the inflationary
expansion but remain constant in amplitude when larger than the contemporary Hubble scale, leading to Gaussian
scale-free perturbations at the start of the Hot Big Bang era. Starting from these fluctuations, Cold Dark Matter
(‘CDM’) creates potential wells for baryons to fall into, but the radiation (tightly coupled to the electrons and
baryons) resists collapse. Gravity wins if the wavelength λ is greater than the Jean’s length λJ (which is proportional
to the speed of sound [174, 83]). There are acoustic oscillations (sound waves) when λ < λJ ; these oscillations
ceased at decoupling, which led to a dramatic decrease in λJ and the growth of structure by gravitational instability
in a ‘bottom up’ way (Sec.2.3.6).17
A popular version of inflation is chaotic inflation [141, 102, 212] where inflation ends at different times in
different places, so that one ends up with numerous ‘pocket universe’ (expanding universe domains like the one we
see around us, or perhaps very different) all imbedded in a still-inflating universe region and starting at different
times, the whole forming a fractal-like structure. It is argued this is an inevitable consequence of the nature of
plausible scalar field potentials.
Inflation is not an inevitable conclusion, for there are some alternatives proposed [117, 124], and the WMAP
results can be reproduced by any scenario where Gaussian scale-free perturbations of suitable amplitude occur
at the start of the Hot Big Bang era. However inflation is regarded by most cosmologists as the best proposal
available for the era prior to the Hot Big Bang epoch, leading to the presence of such perturbations. Nevertheless
one should note it is a generic proposal for what happened, rather than a specific physical theory. While a great
many possibilities have been proposed (it could for example be an effective field due to higher-order gravity effects,
or it could involve multiple scalar fields), at the present time the identity of the proposed inflationary field (‘the
inflaton’) has not been established or linked to any known particle or field. The hoped-for link between early
universe dynamics and particle physics is potential rather than real [44]. Detailed studies of the CBR anisotropies
and structure formation in conjunction with the observations hope to distinguish between the various possibilities,
for example testing whether the spectral index n takes the scale-free value: n = 1, or whether rather there is a
tilted power spectrum (n 6= 1). A unique spectrum of gravitational waves will also be produced at very early times
in an inflationary universe, and detection of these waves either directly by proposed gravitational wave detectors
or indirectly by measuring the associated curl mode in the CBR polarization will be an important test of inflation,
for example determining the ratio r of scalar to tensor perturbations in the early universe [38].
The very early universe
Quantum gravity processes are presumed to have dominated the very earliest times, preceding inflation. There
are many theories of the quantum origin of the universe, but none has attained dominance. The problem is that
we do not have a good theory of quantum gravity [185], so all these attempts are essentially different proposals
for extrapolating known physics into the unknown. A key issue is whether quantum effects can remove the initial
singularity and make possible universes without a beginning. Preliminary results suggest this may be so [12, 184,
Is there a quantum gravity epoch?
A preliminary issue is, can there be a non-singular start to the inflationary era, thus avoiding the need to contemplate
a preceding quantum gravity epoch? In the inflationary epoch the existence of an effective scalar field leads to a
violation of the strong energy condition (8), therefore at first sight it seems that a bounce may be possible preceding
the start of the expanding inflationary era and avoiding the inevitability of a quantum gravity epoch.
However a series of theorems suggest that inflationary models cannot bounce: they are stated to be future
infinite but not past infinite [102]. This is an important issue, so it is worth looking at it further. There are two
major requirements to get a bounce. The Friedmann equation (9) relates the scale factor S(t), curvature constant
k, and the effective total energy density µ(t), which is defined by this equation whatever dynamics may be involved
(multiple scalar fields, higher order gravity, higher dimensional theories leading to effective 4-dimensional theories,
etc.).18 The Raychaudhuri equation (7) includes the effective total pressure p(t), which again is defined by this
equation. In this section, a cosmological constant Λ is represented as perfect fluid with µΛ + pΛ /c2 = 0. To get a
bounce, first one needs the curve S(t) of the scale factor as a function of time to bend up: that is,
≥ 0 ⇔ µ + 3p/c2 < 0,
which is just a violation of the strong energy condition (8). This is the case if µ + p/c2 = 0 (a vacuum); and indeed
by eqn.(33) it is possible for example for any slow-rolling scalar field. Second, one also needs a time when the scale
17 This
18 See
is a highly simplified account; for more detailed versions, see e.g. [38, 203].
[26] for the ways various quantum gravity theories result in modified Friedmann equations.
factor is a minimum. Thus there must be a time t∗ such that Ṡ(t∗ ) = 0. From the Friedmann equation (9),
Ṡ 2 (t∗ ) = 0 ⇔
κµ(t∗ )
= 2
S (t∗ )
With k ≤ 0 this is possible only if µ(t∗ ) < 0. Even with a scalar field (see eqn.(32)) this can only be achieved by
having negative potential energies, which appears to be an unphysical requirement. With k = +1 this is possible
with µ(t∗ ) > 0 [179], which is compatible with ordinary matter.
Thus if you want a bounce in an inflationary universe, it is sensible to look to k = +1 inflationary models,
which indeed will turn around if a vacuum domain occurs for long enough (curvature will eventually always win
over a vacuum as we go back into the past [80, 73]). The theorems mentioned above do not include this case (see
[102]); they only consider inflationary universes with k = 0 and k = −1. And one should note here that although
the scale-free k = 0 exponential case clearly is the model underlying the way many people approach the problem,
it is highly exceptional — it is of zero measure within the space of all inflationary FL models.
Explicit non-singular models can be constructed, the simplest being the de Sitter universe in the k = +1
frame (Sec.2.1.1), which is an exact eternal solution that bounces at a minimum radius S0 . This model has the
problem that it does not exit inflation (it corresponds to an exactly constant potential), but variants exist where
exit is possible; there are also viable non-singular models that start off asymptotic to the Einstein Static universe
in the distant past and avoid the need for a quantum gravity epoch [70]. These models start off in a very special
state, precisely because they are asymptotic to the Einstein static universe in the distant past. This is a possible
situation. It seems likely that the options for the start of inflation are (i) avoiding the quantum gravity era, but
at the cost of having special (‘fine tuned’) initial conditions, or (ii) having a quantum gravity epoch preceding the
inflationary era. Thus a key issue is whether the start of the universe was very special or generic.
Quantum gravity effects: The origin of the universe
Contemporary efforts to explain the beginning of the universe, and the particular initial conditions that have
shaped its evolution, usually adopt some approach or other to applying quantum theory to the creation of the
universe [134]. Many innovative attempts have been made here; as this article focuses on General Relativity and its
application to cosmology, and it would be impossible to do justice to the various approaches to quantum cosmology
[185] without a very much longer article. I will just make a few comments on these approaches.
The attempt to develop a fully adequate quantum gravity approach to cosmology is of course hampered by
the lack of a fully adequate theory of quantum gravity, as well as by the problems at the foundation of quantum
theory (the measurement problem, collapse of the wave function, etc. — see [119, 35, 131]) which can be ignored
in many laboratory situations but have to be faced in the cosmological context [170]. The various attempts at
quantum cosmology each develop in depth some specific aspect of quantum theory that may be expected to emerge
from a successful theory of quantum gravity applied to the universe as a whole, being for example based on either
(i) the Wheeler-deWitt equation and the idea of the wave function of the universe, or (ii) on some version of
embedding in higher dimensional space time (inspired by string theory), or (iii) an appropriate application of loop
quantum gravity. In effect they attempt either
(a) to give a true theory of creation ex nihilo [225]; such efforts however cannot truly “solve” the issue of
creation, for they rely on some structures or other (e.g. the elaborate framework of quantum field theory and much
of the standard model of particle physics) pre-existing the origin of the universe, and hence themselves requiring
explanation; or
(b) to describe a self-sustaining or self-referential universe which by-passes the issue of creation, either by
(b1) originating from an eternally pre-existing state, via the recurring idea of a Phoenix universe [34] (as in
Veneziano’s ‘pre-big bang theory’ based on analogues of the dualities of string theory, or self-repeating universes
such as the chaotic inflationary models of Linde); creation from fluctuations in some quite different pre-existing
structure (e.g. emergence from de Sitter space time; or the ‘ekpyrotic universe’ initiated by a collision between
pre-existing ‘branes’ in a higher dimensional spacetime); or emerging from an eternal static initial state; or
(b2) starting from a state with different properties of time than usual (or with an an emergent notion of time):
as in the Hartle–Hawking no-boundary proposal [108, 109], and the Gott causal violation proposal [95] where the
universe ‘creates itself’ and starts normal expansion in the domain without closed timelike lines.
Any of these may be combined with one or other proposals for
(c) an effective ensemble of universes [216], realized either
(c1) in space-time regions that are part of either a larger entangled quantum entity, or are part of a single classical
space-time, but are effectively disconnected from each other, or
(c2) in truly disconnected form.
All of these proposals however are strongly speculative, none being based solidly in well-founded and tested physics,
and none being in any serious sense supported by observational evidence. They are all vast extrapolations from
the known to the unknown. They may or may not be true. One thing is certain: they can’t all be true!
The concordance model
Observational support for the idea of expansion from a Hot Big Bang epoch is very strong, the linear magnituderedshift relation for galaxies demonstrating the expansion,19 with source number counts and the existence of
the blackbody CBR being strong evidence that there was indeed evolution from a hot early stage. Agreement
between measured light element abundances and the theory of nucleosynthesis in the early universe confirms this
interpretation. This basic theory is robust to critical probing. Much present activity attempts to link particle
physics interactions during very early stages of the expansion of the universe to the creation of structures by
gravitational instability much later, traces of the early seed fluctuations being accessible to us through present day
CBR anisotropy patterns. Thus the present dominant cosmological paradigm is a quantum gravity era of some kind
followed by inflation; a hot big bang epoch; decoupling of matter and radiation; and then gravitational instability
leading to formation of clusters of galaxies from the seed density perturbations that occur on the LSS.
Together with supernova data, analysis of the CBR angular anisotropies and in particular their peaks gives
a concordance model of this kind [8, 215, 217, 38, 195] that is then confirmed by the statistics of matter clustering
[48] together with observations of gravitational lensing and large-scale motions of matter [203]. This model is
characterized by specific values for a set of cosmological parameters [139], in particular
Ωcdm0 ≃ 0.3, ΩΛ0 ≃ 0.7, Tcbr0 = 2.75K, H0 ≃ 65km/sec/mpc, t0 ≃ 1.4 × 1010 years.
Also Ωbar0 ≃ 0.044 is the density of baryons, Ωvis0 ≃ 0.015 that of luminous matter, and Ων0 ≃ 10−5 that of massless
neutrinos, implying Ω0 ≃ 0.3+0.7 ≃ 1 in agreement with the inflationary prediction (34). The sign of k is uncertain,
but if the combined evidence of all current observations is taken at face value it is positive, with Ω0 = 1.02 ± 0.02
[206]. As noted above, there are some concerns firstly over age issues (see Sec.2.3.2); secondly concerning the large
angle CBR anisotropies (see Sec.2.3.7); and thirdly regarding details of CDM structure formation at small scales
(see Sec.2.3.6); but none of these issues seems to be crucial at present.
Some misunderstandings
Despite its simplicity, there are some common misconceptions about the standard universe models (cf. [142]) that
can lead to philosophical misunderstandings.
Misconception 1: The universe is expanding into something. It is not, as it is all there is. It is just getting
bigger, while always remaining all that is. One should note here that a RW universe can be represented as a
4-dimensional curved spacetime expanding in a 5-dimensional flat embedding space time [179]; however there is no
necessity to view the 5-dimensional spacetime in this representation as physically real. Furthermore this embedding
is no longer possible when we take perturbations into account; a 10 dimensional flat spacetime is needed for locally
embedding a realistic (perturbed) universe model (and to do so globally requires many more dimensions, in general).
Misconception 2: The universe expands from a specific point, which is the centre of the expansion. All spatial points are equivalent in these universes, and the universe expands equally about all of them. Every fundamental
observer sees exactly the same thing in an exact RW geometry. There is no centre to a FL universe.
Misconception 3: Matter cannot recede from us faster than light. It can, at an instant; two distantly
separated fundamental observers in a surface {t = const} can have a relative velocity greater than c if their spatial
separation is large enough [183, 31]. No violation of special relativity is implied, as this is not a local velocity
difference, and no information is transferred between distant galaxies moving apart at these speeds. For example,
there is presently a sphere around us of matter receding from us at the speed of light;20 matter beyond this sphere
is moving away from us at a speed greater than the speed of light. The matter that emitted the CBR was moving
away from us at a speed of about 61c when it did so [183].
Misconception 4: The existence of a preferred FR frame (that in which the universe appears isotropic)
contradicts relativity theory, which says all reference frames are equally good. But this equivalence of frames is true
for the equations rather than their solutions. Almost all particular solutions will have preferred world lines and
surfaces; this is just a particular example of a broken symmetry — the occurrence of solutions of equations with
less symmetries than the equations display. This feature is a key theme in modern physics [19, 107].
19 The alternative interpretation as gravitational redshifts in a static universe does not work because of the linearity of the observed
redshift-distance relation [71].
20 This sphere is not the same as the particle horizon, as is sometimes claimed (see [183]).
Misconception 5: The space sections are necessarily infinite if k = 0 or −1. This is only true if they have
their ‘natural’ simply connected topology. If their topology is more complex (e.g. a 3-torus) they can be spatially
finite [50, 130]. There are many ways this can happen; indeed if k = −1 there is an infinite number of possibilities.
Misconception 6: Inflation implies spatial flatness (k = 0 ⇔ Ωk = 1) exactly. There is nothing in
inflationary theory which determines the sign of the spatial curvature. Inflationary universes are very nearly flat
at late times; this is very different from being exactly flat (a condition which requires infinite fine tuning of initial
conditions; if say the two millionth digit in the value of Ωk is non-zero at any time, then the universe is not
spatially flat). Inflationary theory does not have the theoretical teeth required to imply that the universe has
exactly flat spatial sections; hence a key issue for cosmology is observationally determining the sign of the spatial
curvature, which is potentially dynamically important in both the very early universe [80, 73] and the late universe
(it determines if recollapse is possible, should the dark energy decay away).
Cosmology has changed from a speculative enterprize into a data-driven science that is part of standard physical
theory [5]; a wealth of observations supports this dominant theory [166, 201, 171]. Nevertheless some theoretical
proposals are being made for the very early stages that have no observational support; and sometimes it may be
impossible to ever obtain such support, both as regards the proposed physics and the geometry. Thus in some
respects it remains a principle driven enterprise, with observation subordinate to theory.
We now explore the relation between cosmology and philosophy in terms of a series of Theses clustered
around a set of major Issues. One can obtain a synoptic overview of the overall argument by simply considering
the full set of Issues and Theses. They are summarized in the Table at the end.
Issue A: The uniqueness of the universe.
The first and most fundamental issue is that there is only one Universe [156, 148, 58]. This essential uniqueness
of its object of study sets cosmology apart from all other sciences. In particular, the unique initial conditions that
led to the particular state of the universe we see were somehow “set” by the time that physical laws as we know
them started governing the evolution of both the universe and its contents, whenever that time may be. We cannot
alter these unique initial conditions in any way — they are given to us as absolute and unchangeable, even though
they are understood as contingent rather than necessary; that is, they could have been different while still being
consistent with all known physical laws. The implications are that
Thesis A1: The universe itself cannot be subjected to physical experimentation. We cannot
re-run the universe with the same or altered conditions to see what would happen if they were different, so we
cannot carry out scientific experiments on the universe itself. Furthermore,
Thesis A2: The universe cannot be observationally compared with other universes. We cannot
compare the universe with any similar object, nor can we test our hypotheses about it by observations determining
statistical properties of a known class of physically existing universes.
Where this all becomes of observational relevance is in the idea of cosmic variance ([38], pp. 241, 343). The
theory of structure formation in the early universe makes statistical predictions only (it cannot attempt to predict
the specific structures that will actually be formed). Testing the theory compares our universe to a theoretical
ensemble of universes, and declares a variance between what is measured in the actual universe and the expected
properties based on the ensemble of models. If this variance is small enough, a deviation from expected values is
pronounced as a statistical deviation, i.e. of no physical significance — we do not need to explain it any further; if
it is large, it needs explanation. This is a key issue for example in the analysis of the CBR anisotropy observations
[237, 121]. The power spectrum of the CBR as measured by WMAP is less than expected at large angular scales
(Sec.2.3.7). One school of thought claims this is just a statistical fluctuation; another that it needs explanation,
and might for example be evidence of a small universe [143]. This debate arises because there is just one universe,
and on large angular scales there are just a few measurements that can possibly be made (on small angular scales
we can make many measurements and so this uncertainty becomes very small).
Consequent on A1 and A2,
Thesis A3: The concept of ‘Laws of Physics’ that apply to only one object is questionable. We
cannot scientifically establish ‘laws of the universe’ that might apply to the class of all such objects, for we cannot
test any such proposed law except in terms of being consistent with one object (the observed universe).
This is insufficient: one observational point cannot establish the nature of a causal relation. Indeed the concept of
a ‘law’ becomes doubtful when there is only one given object to which it applies [156]. The basic idea of a physical
law is that it applies to a set of objects all of which have the same invariant underlying behaviour (as defined by
that law), despite the apparent variation in properties in specific instances, this variation resulting from varying
initial conditions for the systems on which the law acts. This understanding is tested by physical experiments in
which initial conditions for evolution of a set of similar systems are varied, and observations by which the statistical
nature of a set of objects of the same broad kind is investigated. Neither is possible in the case of cosmology.
The laws of physics apply locally to the objects in the cosmos, and determine the evolution of the cosmos
as a whole when locally applied everywhere with suitable initial/boundary conditions imposed (in the case of the
FL models, via the Friedmann equation for example). Apart from this, we cannot establish higher-level effective
laws that apply to all universes and determine their structure, as we can at all other levels of the hierarchy of
complexity. All that we can do at this level of structure is observe and analyze the one unique object that exists.
This is expressed by McCrea as follows: “When we speak of the other solutions of the equations of stellar structure,
besides the one we are interested in at the moment, as representing systems that could exist, we mean that they
could exist in the universe as we know it. Clearly no such attitude is possible towards the universe itself” [147].
Since the restriction of a global solution to a local neighborhood is also a solution, we have zillions of
“mini-universe” on which to test the laws that control the local nature of the universe. But a mini-universe is
not the universe itself; it is a small part of the whole. By examining these “mini-universes” and seeing if they
are essentially the same everywhere, we can to some degree check firstly that the laws of physics are the same
everywhere in the universe (a key feature of all cosmological analysis, cf. Sec.7.1), and secondly that the universe
is spatially homogeneous (this is discussed in depth below, see Sec.4.2.2). But the latter feature is what has to be
explained by a ‘law of the universe’; verifying homogeneity does not explain why it is the case; this comes about
because of specific initial conditions, which some suggest are due to hypothesized ‘laws of the universe’, applicable
to the whole rather than to its parts. Finally,
Thesis A4: The concept of probability is problematic in the context of existence of only one
object. Problems arise in applying the idea of probability to cosmology as a whole — it is not clear that this makes
much sense in the context of the existence of a single object which cannot be compared with any other existing
But a concept of probability underlies much of modern argumentation in cosmology. Talk of ‘fine tuning’ for
example is based on the use of probability (it is a way of saying something is improbable). This assumes both
that things could have been different, and that we can assign probabilities to the set of unrealized possibilities
in an invariant way. The issue here is to explain in what sense they could have been different with well-defined
probabilities assigned to the different theoretical possibilities, if there is indeed only one universe with one set of
initial conditions fixed somehow before physics came into being, or perhaps as physics came into being. We cannot
scientifically establish laws of creation of the universe that might determine such initial conditions and resulting
probabilities. If we use a Bayesian interpretation, which some suggest can be meaningfully applied to only one
object [89], the results depend on our ‘prior knowledge’, which in this case can be varied by changing our initial
pre-physics assumptions. Related issues arise concerning the meaning of ‘the wave function of the universe’, at the
heart of quantum cosmology. This wave function gives no unique prediction for any specific single universe.
Two comments on the above. First, it is useful to distinguish between the experimental sciences — physics,
chemistry, microbiology for example — on the one hand, and the historical and geographical sciences — astronomy,
geology, evolutionary theory for example, on the other. It is the former that are usually in mind in discussions of
the scientific method. The understanding in these cases is that we observe and experiment on a class of identical
or almost identical objects and establish their common behaviour. The problem then resides in just how identical
those objects are. Quarks, protons, electrons, are all exactly identical to each other, and so have exactly the same
behaviour (indeed this feature underlies well-tested quantum statistics). All DNA molecules, frogs, human beings,
and ecosystems are somewhat different from each other, but are similar enough nevertheless that the same broad
descriptions and laws apply to them; if this were not so, then we would be wrong in claiming they belonged to the
same class of objects in the first place. Water molecules, gases, solids, liquids are in an intermediate category —
almost identical, certainly describable reliably by specific physical and chemical laws.
As regards the geographical and historical sciences, here one explicitly studies objects that are unique (the
Rio Grande, the continent of Antarctica, the Solar System, the Andromeda galaxy, etc.) or events that have
occurred only once (the origin of the Solar System, the evolution of life on Earth, the explosion of SN1987a,
etc.). Because of this uniqueness, comment A1 above applies in these cases also: we can only observe rather than
experiment; the initial conditions that led to these unique objects or events cannot be altered or experimented
with. However comment A2 does not apply: at least in principle, there is a class of similar objects out there (other
rivers, continents, planetary systems, galaxies, etc.) or similar events (the origin of other galaxies, the evolution
of other planetary systems, the explosion of other supernovae, etc.) which we can observe and compare with our
specific exemplar, also carrying out statistical analyses on many such cases to determine underlying patterns of
regularity; and in this respect these topics differ from cosmology.
If we truly cannot carry out such analyses — that is, if A2 applies as well in some particular case — then
that subject partakes in this respect of the nature of cosmology. One may claim that the dividing line here is
that if we convince ourselves that some large-scale physical phenomenon essentially occurs only once in the entire
universe, then it should be regarded as part of cosmology proper ; whereas if we are convinced it occurs in many
places or times, even if we cannot observationally access them (e.g. we believe that planets evolved around many
stars in other galaxies), then study of that class of objects or events can be distinguished from cosmology proper
precisely because there is a class of them to study.
The second comment is that some workers have tried to get around this set of problems by essentially
denying the uniqueness of the universe. This is done by proposing the physical existence of ‘many universes’ to
which concepts of probability can be properly applied (cf. Sec.2.7.2), envisaged either as widely separated regions
of a larger universe with very different properties in each region (as in chaotic inflation for example), as multiple
realizations of quantum outcomes, or as an ensemble of completely disconnected universes — there is no physical
connection whatever between them — in which all possibilities are realized. We return to this in Sec.9.2.
Issue B: The large scale of the universe in space and time.
The problems arising from the uniqueness of the universe are compounded by its vast scale in both space and time,
which poses major problems for observational cosmology. We therefore need to adduce various Principles in addition
to the observations, in order to attain unique models: theory comes in as basis for interpreting observations.
Observations in a large scale universe
The distance to the nearest galaxy is about 106 light years, that is about 1024 cm., while the size of the earth is
about 109 cm. The present size of the visible universe is about 1010 light years, that is about 1028 cm. This huge
size relative to our own physical scale (about 102 cm) places major constraints on our ability to observe distant
regions (and certainly prevents us experimenting with them). The uniqueness of cosmology in this respect is that
it deals with this scale: the largest with which we can have causal or observational contact.
Thesis B1: Astronomical observations are confined to the past null cone, fading with distance.
We can effectively only observe the universe, considered on a cosmological scale, from one space-time event. Visual
observations are possible only on our past light cone, so we are inevitably looking back into the past as we observe
to greater distances. Uncertainty grows with distance and time.
The vast scale of the universe implies we can effectively only view it from one spacetime event (‘here and now’)
[50, 52]. If we were to move away from this spatial position at almost the speed of light for say 10, 000 years, we
would not succeed in leaving our own galaxy, much less in reaching another one; and if we were to start a long-term
astronomical experiment that would store data for say 20, 000 years and then analyze it, the time at which we
observe the universe would be essentially unchanged (because its age is of the order of 1010 years: the extra time
would make a negligible difference). This is quite unlike other geographic sciences: we can travel everywhere on
earth and see what is there. The situation would be quite different if the universe were much smaller. Given its
actual scale, such that we are now seeing galaxies whose present distance from us is about 109 light years, the effect
is as if we were only able to observe the earth from the top of one mountain, and had to deduce its nature from
those observations alone [52].
Because we can only observe by means of particles — photons, massless neutrinos, gravitons — travelling
to us at the speed of light, astronomical observations of distant sources and background radiation by telescopes
operating at all wavelengths (optical, infrared, ultraviolet, radio, X-ray) are constrained to rays lying in our past
light cone. These allow detailed observations (including visual pictures, spectral information, and polarization
measurements) of matter as it intersects our past light cone. In observing distant regions, we can also aspire to use
neutrino and gravitational wave telescopes, and perhaps cosmic rays, also representing information coming to us
at the speed of light or less. However all our detailed data about distant regions is gathered along our past light
As a consequence, three interrelated problems occur in interpreting the astronomical observations. The first
is that (because we can only view the universe from one point) we only obtain a 2-dimensional projection on the sky
of the 3-dimensional distribution of matter in the universe. To reconstruct the real distribution, we need reliable
distance measurements to the objects we see. However because of variation in the properties of sources, most are
not reliable standard candles or standard size objects to use in calibrating distances, and in these cases we have to
study statistical properties of classes of sources to estimate distances.
Second, we necessarily see distant galaxies and other objects at earlier times in their history (where their
world lines intersect this past light cone).21 Thus cosmology is both a geographic and a historical science combined
into one: we see distant sources at an earlier epoch, when their properties may have been different. As we are
21 For
example we see the Andromeda galaxy as it was two million years ago, long before humans existed on Earth [203].
looking back in the past, source evolution must be taken into account; their properties at the time they emitted
the light may be quite different from their properties now. We can only determine the distances of objects if we
understand this evolution; but in practice it is one of the unknowns we have to try to determine (cf. Sec.4.2.3).
Third, distant sources appear very small and very faint, both because of their physical distance, and because
their light is highly redshifted (due to the expansion of the universe). Simply detecting them, let alone determining
their characteristics, becomes rapidly more difficult with distance. Furthermore absorption by intervening matter
can interfere with light from distant objects. The further back we look, the worse these problems become; thus our
reliable knowledge of the universe decreases rapidly with distance [52].
The situation is however improved by the availability of geological-type data [114]; that is, the present-day
status of rocks, planets, star clusters, galaxies, and so on, which contains much information on the past history
of the matter comprising those objects. Thus we can obtain detailed information on conditions near our past
world-line in spacetime [50, 52] at very early times if we can interpret this data reliably, for example by relating
theories of structure formation to statistical studies of source properties.
Thesis B2: ‘Geological’ type observations can probe the region near our past world line in the
very distant past. Physical and astrophysical observations tell us about conditions near matter world-lines in the
far distant past. They can be used also to investigate the far distant past of more distant objects.
This involves us in physical cosmology: namely the study of the evolution of structures in the universe, tested by
comparison with astronomical observation. Particularly useful are measurements of the abundances of elements
which resulted from nucleosynthesis in the Hot Big Bang, giving us data about conditions long before decoupling
(Sec.2.2.2). If we can obtain adequate quality data of this kind for objects at high redshifts, we can use this to
probe conditions very early on in their histories at some distance from our past worldline. Encouraging in this
regard is the possibility of determination of element abundances at high redshift ([38], pp. 11-12; [173]).
Determining Spacetime Geometry: Observational Limits.
The unique core business of observational cosmology is determining the large-scale geometry of everything there
is, or at least of everything we can observe.
Direct determination versus theory based approaches
One can go about this in a direct manner: trying to determine the geometry of the universe directly from observations (assuming one has some understanding of the sources observed). The way this can be done (curiously known
as the ‘inverse approach’) has been fully characterized [128, 74]; indeed there is an interesting result here, namely
Observational Cosmology Theorem: The data in principle available on our past null cone from astronomical observations is just necessary and sufficient to determine the space-time geometry on that null cone [74].
From this data one can in principle determine the space time in the past of the null cone and, if a no-interference
conditions is assumed, to its future.
However this is difficult to carry out both because of the problem of estimating distances for all observed sources,
requiring a knowledge of the nature of the sources (Sec.4.2.3),22 and because of the serious difficulty in obtaining
some of the needed data (which include apparent distortions of all distant objects, and the transverse velocities of
all observed matter). The further we observe down the past light cone, the larger the uncertainty becomes. This
direct observational approach, where no prior model is assumed for the space-time geometry, has been pursued to
some degree (and in essence underlies for example the observational studies that discovered large-scale structure
such as the great walls and voids). Nevertheless it is not widely adopted as an overall approach to cosmology, both
because of these observational difficulties, but also because it has little explanatory value; it just tells us what the
geometry and matter distribution is, but not why it is of that nature.
The usual option in cosmology proper is rather to use a theory-based approach: we a priori assume a
model based on a space-time geometry with high symmetry (usually a FL model, see Sec.2.1), and then determine
its essential free parameters from comparison of theoretical relations with astronomical observations (Sec.2.3.3).
Detailed observations of the matter distribution and large-scale velocities as well as CBR anisotropies then help us
determine deviations from the exact model, both statistically (an astrophysical description [38]) and in detail (an
astronomical description [78]).
22 The
link between observations and models always requires some theory, and is never direct.
Indirect determination: justifying a Friedmann-Lemaı̂tre geometry
The standard models of cosmology are the Friedmann-Lemaı̂tre (FL) family of universe models that are exactly
spatially homogeneous and isotropic everywhere (Sec.2.1). They are easy to understand, and have tremendous
explanatory power; furthermore their major physical predictions (the existence of blackbody CBR and specific
light element production in the early universe) seem confirmed. The issue is, to what degree does observational
data uniquely indicate these universe models for the expanding universe geometry? Here one is assuming a large
enough averaging scale for spatial homogeneity to be valid; this scale should be explicitly indicated [55] (it is
about 100 Mpc at present [38]).23 These are the background models for cosmology; perturbed FL models then
characterize the nature of deviations from the exact FL geometry that are expected on smaller scales (Sec.2.5.2).
The key feature here is the observed isotropy about our location (Sec.2.3.1). Considered on a large enough
angular scale, astronomical observations are very nearly isotropic about us, both as regards source observations
and background radiation; indeed the latter is spectacularly isotropic, better than one part in 104 after a dipole
anisotropy, understood as resulting from our motion relative to the rest frame of the universe, has been removed
[161]. Because this applies to all observations (in particular, there are not major observed matter concentrations
in some other universe region), this establishes that in the observable region of the universe, to high accuracy
both the space-time structure and the matter distribution are isotropic about us. We can easily construct exact
spherically symmetric universe models [13, 83], as indicated by these observations. In general they will be spatially
inhomogeneous, with our Galaxy located at or near the centre; this is currently a philosophically unpopular
proposal, but is certainly possible. The question is whether we can give convincing observational evidence for
spatial homogeneity in addition to the spherical symmetry. Various arguments are used for this purpose.
(a) The cosmological principle [14, 233]: Just assume spatial homogeneity because it is the simplest case and
you don’t need anything more complex on the basis of current data. We simply adopt a philosophical principle as
the basis of argument. This is essentially an a priori prescription for initial conditions for the universe (a universe
that initially has a RW geometry will have that geometry at later times, because symmetries of the initial data are
preserved by the Einstein equations [110]); but it is not usually expressed that way.
(b) FL observational relations: If we could show that the source observational relations had the unique FL
form (26, 28) as a function of distance, this would establish spatial homogeneity in addition to the isotropy, and
hence a RW geometry [74]. This is essentially what is done for example in using number counts towards establishing
spatial homogeneity [115]. However because of Thesis B1 above, the observational problems mentioned earlier —
specifically, unknown source evolution — prevent us from carrying this through: we cannot measure distances
reliably enough. Astrophysical cosmology could resolve this in principle, but is unable to do so in practice. Indeed
the actual situation is the inverse: taking radio-source number-count data at its face value, without allowing for
source evolution, contradicts a RW geometry.
In the face of this, the usual procedure is to assume that spatial homogeneity is known in some other way,
and deduce the source evolution required to make the observations compatible with this geometric assumption
[52]. It is always possible to find a source evolution that will achieve this [157]. Thus attempts to observationally
prove spatial homogeneity this way fail; indeed an alternative interpretation would be that this data is evidence of
spatial inhomogeneity, i.e. that we live in a spherically symmetric inhomogeneous universe where we are situated
somewhere near the centre, with the cosmological redshift being partly gravitational, cf. [71] (and conceivably with
a contribution to the CBR dipole from this inhomogeneity if we are a bit off-centre). Similarly the supernova data
usually understood as implying the existence of a cosmological constant (Sec.2.3.5) could also be interpreted in
this way as evidence of inhomogeneity, without the need for ‘dark energy’. Most people regard such proposals as
very unappealing — but that does not prove they are incorrect.
(c) Physical arguments: One can claim that physical processes such as inflation (Sec.2.6) make the existence
of almost-RW regions highly likely, indeed much more probable than spherically symmetric inhomogeneous regions.
This is a viable argument, but we must be clear what is happening here — we are replacing an observational test by
a theoretical argument based on a physical process that may or may not have happened (for there is no definitive
observational proof that inflation indeed took place). It is strongly bolstered because predictions for the detailed
pattern of CBR anisotropy on small scales [118], based on the inflationary universe theory, have been confirmed
[171]; but that argument will only become rigorous if it is shown that spherically symmetric inhomogeneous models
(with or without inflation) cannot produce similar patterns of anisotropy. But they probably can, because the
acoustic oscillations that lead to the characteristic predicted anisotropy patterns in fact take place after inflation,
and can equally happen if suitable initial conditions occur without a previous inflationary phase.
What about alternative observational routes? Another proposal is,
(d) Uniform thermal histories: the idea is to use the uniformity in the nature of the objects we see in
the sky (we see the same types of galaxy at large distances, for example) to deduce they must have all undergone
23 There exist hierarchical models where neither the fluid approximation nor homogeneity is ever attained at any scale because of
their fractal nature [33]. The regularity of the observed galactic motions, as evidenced by the (m, z) relations, speaks against these
models, as do large-scale observations of the matter distribution [164].
essentially the same thermal history, and then to prove from this homogeneity of thermal histories that the universe
must be spatially homogeneous. For example, observations showing that element abundances at high redshift in
many directions are the same as locally, are very useful in constraining inhomogeneity by showing that conditions
in the very early universe at the time of nucleosynthesis must have been the same at distant locations in these
directions [82]. However turning this idea into a proper test of homogeneity has not succeeded so far: indeed it is
not clear if this can be done, because some (rather special) counter-examples to this conjecture have been found
[15]. Nevertheless the approach could be used to give evidence against spatial homogeneity: for example, if element
abundances were measured to be different at high redshifts in any direction [173, 200], or if ages of distant objects
were incompatible with local age estimates [120].
Finally the argument for spatial homogenity that is most generally accepted:
(e) Isotropy everywhere: If all observers see an isotropic universe, then spatial homogeneity follows [231, 45,
50]; indeed homogeneity follows if only three spatially separated observers see isotropy. Now we cannot observe
the universe from any other point, so we cannot observationally establish that far distant observers see an isotropic
universe. Hence the standard argument is to assume a Copernican Principle: that we are not privileged observers.
This is plausible in that all observable regions of the universe look alike: we see no major changes in conditions
anywhere we look. Combined with the isotropy we see about ourselves, this implies that all observers see an
isotropic universe, and this establishes a RW geometry [231, 50, 110]. This result holds if we assume isotropy of all
observations; a powerful enhancement was proved by Ehlers, Geren, and Sachs [46, 110], who showed that if one
simply assumes isotropy of freely-propagating radiation about each observer in an expanding universe domain,24
the result follows from the Einstein and Liouville equations; that is,
EGS Theorem: Exact isotropy of the CBR for every geodesically moving fundamental observer at each
point in an expanding universe domain U implies an exact RW geometry in U.
Thus we may establish spatial homogeneity by assuming a weak Copernican principle: we are not in a privileged
position where the CBR just happens to be highly isotropic by chance; hence all comoving observers may be assumed
to measure highly isotropic CBR, and the result follows. This is currently the most persuasive observationally-based
argument we have for spatial homogeneity.
A problem is that it is an exact result, assuming exact isotropy of the CBR; is the result stable? Indeed it is:
almost-isotropy of freely-propagating CBR for an expanding family of geodesically-moving fundamental observers
everywhere in some region proves the universe geometry is almost-RW in that region [210]. Thus the result applies
to the real universe — provided we make the Copernican assumption that all other observers, like us, see almost
isotropic CBR. And that is the best we can do at present. Weak tests of the isotropy of the CBR at other spacetime
points come from the Sunyaev-Zel’dovich effect [93] and from CBR polarization measurements [121], giving broad
support to this line of argument but not enough to give good limits on spatial inhomogeneity.
The observational situation is clear:
Thesis B3: Establishing a Robertson-Walker geometry for the universe relies on plausible
philosophical assumptions. The deduction of spatial homogeneity follows not directly from astronomical data,
but because we add to the observations a philosophical principle that is plausible but untestable.
The purpose of the above analysis is not to seriously support the view that the universe is spherically symmetric and
inhomogeneous, as is allowed by the observations, but rather to show clearly the nature of the best observationallybased argument by which we can (quite reasonably) justify the assumption of spatial homogeneity.
Accepting this argument, the further question is, in which spacetime regions does it establish a RW-like
geometry? The CBR we detect probes the state of the universe from the time of decoupling of matter and
radiation (at a redshift of about 1100) to the present day, within the visual horizon. The argument from CBR
isotropy can legitimately be applied for that epoch. However, it does not necessarily imply isotropy of the universe
at much earlier or much later times, because there are spatially homogeneous anisotropic perturbation modes that
are unstable in both directions of time; and they will occur in a generic situation. Indeed, if one examines the
Bianchi (spatially homogeneous but anisotropic) universes, using the powerful tools of dynamical systems theory,
one can show that intermediate isotropisation can occur [226, 227]: despite being highly anisotropic at very early and
very late times, such models can mimic a RW geometry arbitrarily closely for an arbitrarily long time, and hence
can reproduce within the errors any set of FL-like observations. We can obtain strong limits on the present-day
strengths of these anisotropic modes from CBR anisotropy measurements and from data on element abundances,
the latter being a powerful probe because (being of the ‘geological’ kind) they can test conditions at the time
of element formation, long before decoupling. But however low these observational limits, anisotropic modes can
dominate at even earlier times as well as at late times (long after the present). If inflation took place, this conclusion
is reinforced: it washes out any information about very early universe anisotropies and inhomogeneities in a very
efficient way.
24 This result does not hold in a static universe, for then the radiation temperature depends only on the potential difference between
the emitter and observer, hence the radiation is isotropic everywhere even if the universe inhomogeneous, cf. [71].
As well as this time limitation on when we can regard homogeneity as established, there are major spatial
limitations. The above argument does not apply far outside the visual horizon, for we have no reason to believe
the CBR is highly isotropic there. Indeed if chaotic inflation is correct, conditions there are not the same.
Determining the model parameters
Given that a RW geometry is a good description of the observable universe on a large scale, the further issue is
what are the best-fit parameters that characterize it, selecting the specific universe we observe from the family of
all FL models (Sec.2.1). Important observational issues are:
• Determining the Hubble parameter H0 , which sets the overall scale of the observed universe region.
• Determining the trio of the density parameter Ω0 , deceleration parameter q0 , and cosmological constant Λ
(or equivalently the density parameter ΩΛ ), which are the major defining characteristics of a specific FL
model. The CBR data, supernova observations, deep number counts, source covariance functions, velocity
measurements, and gravitational lensing observations can determine these quantities.
• Determining the sign of the curvature k, showing whether the universe has closed spatial sections and also
whether it is possible for it to recollapse in the future or not. Analyses of the observations should always
attempt to determine this sign, and not assume that k = 0 (as is often done) [240].
• Various parameters are used to characterize the nature of dark matter (Sec.2.3.6) and dark energy (Sec.2.3.5).
As their dynamics is unknown, these too have to be determined observationally.
We only obtain good estimates of these quantities by the observational relationships characterized above (Sec.2.3.3)
using statistical analysis of the classes of objects we observe. Problems arise because of our lack of adequate theories
of their historical development.
Thesis B4: Interpreting cosmological observations depends on astrophysical understanding.
Observational analysis depends on assessing a variety of auxiliary functions characterizing the sources observed
and the observations made. These introduce further parameters that have to be observationally or theoretically
determined, allowing considerable freedom in fitting specific models to the observations. Physical cosmology aims to
characterize perturbed FL models (which account for structure formation) rather than just the background exactly
smooth FL models; this introduces further parameters to be determined.
It is useful here to distinguish between methods aimed at determining the properties of the background (zeroth
order) FL model directly, and those aimed at determining properties of the perturbations of these models [215].
Methods for determining the parameters of the background model (Sec.2.1) depend on assuming properties of the
distance indicators used (galaxies, radio sources, etc.). They will have their own properties (brightness profiles,
luminosities, physical sizes, spectra, etc.) and dynamical evolution; but these are often not well understood, and
will have to be represented in a parametric way (e.g. by parameters describing luminosity evolution). In each
case we end up assuming important aspects of the astrophysics and evolutionary histories of the objects observed,
which are not part of the cosmological model proper. The statistical properties of the sources observed are also
characterized by parametrized functions (e.g. the luminosity function characterizing the numbers of galaxies in
each luminosity class) that have to be known in order to analyze the observations. This situation is an exemple of
Lakatos’ view of how scientific programmes work, with a belt of auxiliary hypotheses interposing between the core
theoretical proposal and the data used to test it [133]. This makes the analysis rather model-dependent, where the
models are only indirectly related to the background model — their explanation is the aim of astrophysics rather
than cosmology. Thus if observational results disagree with a particular cosmological model, one can always claim
it is the understanding of the auxiliary hypotheses that is at fault rather than the model being proposed [133].
By contrast, many of the methods of estimating Ω0 (and to some degree Λ) depend on studying the growth
and nature of inhomogeneities in the universe, that is they investigate perturbed FL models (Sec.2.5.2), whose
properties of course depend on the background model, but introduce a whole set of further functions and parameters
describing the perturbations [38], for example the angular correlation function for matter (or its Fourier transform,
the 2-dimensional power spectrum), the power spectrum of density fluctuations [215], red-shift space correlation
functions [164, 48], and correlation function for velocities [38]. Associated parameters include a scalar spectral
index (characterizing the spectrum of physical sizes of inhomogeneities), the bias parameter b (expressing how
galaxy formation is biassed towards density peaks in the inhomogeneities ([38], p. 280)) and the initial fluctuation
magnitudes Q (the seeds for structure formation). Determining these parameters is part of the task of cosmology
proper: to fully characterize the perturbed cosmological model, we aim to determine both the background parameters
and the quantities describing the perturbations. Model selection then depends on the parameters used to describe
them — what is assumed known, and what is to be determined [139, 195]. For example, standard inflationary
theory predicts a scale-invariant spectrum of Gaussian perturbations; do we test that assumption, or take it for
granted? This comes up in the issue of what ‘priors’ are assumed when conducting statistical tests.
Consistency tests
A key question for cosmology is what kinds of observations provide critical tests of the standard FL models. If
there were no observations that could disprove them, the subject would be of questionable scientific status. An
important such test is obtaining estimates of the age of the universe t0 , which is dependent on H0 , Ω0 , and Λ, and
comparing them with estimates of the ages of objects in the universe (determined on astrophysical grounds):
Thesis B5: A crucial observational test for cosmology is that the age of the universe must
be greater than the ages of stars. The tension between the age of the universe and ages of stars is one area
where the standard models are vulnerable to being shown to be inconsistent, hence the vital need to establish reliable
distance scales, basic to estimates of both H0 and the ages of stars, and good limits on Λ. Other consistency tests
help confirm the standard model and consolidate cosmology’s standing as an empirical science.
At present this age issue is acceptable for local objects, because of a recent revision of our distance scale estimates
[103], assisted by data that Λ is positive [172]; but continued vigilance is needed on this front, particularly as there
are indications of problems for high redshift objects [120]. If this ever became serious we might have to resort
to spherically symmetric inhomogeneous models rather than spatially homogeneous models, with the ‘bang time’
(characterizing the start of the universe) dependent on distance from us [157].
Note that this issue is crucially unlike the case of the large angle CBR anisotropies (Sec.2.3.7): the low CBR
anisotropies at large angular scales can as a last resort be dismissed as a statistical fluke; the age issue cannot. It
is to do with the internal consistency of individual cosmological models, not with probabilities. Thus it is a plus
for cosmology that the age issue exists. Other consistency tests include
• Showing that the CBR temperature Tcbr varies with redshift according to Tcbr = 2.75 (1 + z) [150];
• Confirming that helium abundances are consistent with a primordial value of 25% at large distances (high
redshifts) in all directions ([38], pp. 11-12; also [173, 200]); and
• Checking that there is a 2% number count dipole parallel to the CBR dipole for all cosmological sources [64].
The hidden universe
If we do not live in a small universe (Sec.4.3.1), the further essential point is that the region of the universe we
can observe is restricted, firstly because we cannot see to earlier times than the LSS (the universe was opaque
before then (see Sec.2.2)), and secondly because a finite time has elapsed since the universe became transparent to
radiation, and light can only have travelled a finite distance in that time. As no signal can travel to us faster than
light, we cannot receive any information from galaxies more distant than our visual horizon [79]. The most distant
matter we can observe is that which emitted the CBR (Sec.2.4.2).
Thesis B6: Observational horizons limit our ability to observationally determine the very large
scale geometry of the universe. We can only see back to the time of decoupling of matter and radiation, and
so have no direct information about earlier times; and unless we live in a ‘small universe’, most of the matter
in the universe is hidden behind the visual horizon. Conjectures as to its geometry on larger scales cannot be
observationally tested. The situation is completely different in the small universe case: then we can see everything
there is in the universe, including our own galaxy at earlier times.
The key point here is that unless we live in a small universe, the universe itself is much bigger than the observable
universe. There are many galaxies — perhaps an infinite number — at a greater distance than the horizon, that
we cannot observe by any electromagnetic radiation. Furthermore no causal influence can reach us from matter
more distant than our particle horizon — the distance light can have travelled since the creation of the universe,
so this is the furthest matter with which we can have had any causal connection [177, 110, 219]. We can hope to
obtain information on matter lying between the visual horizon and the particle horizon by neutrino or gravitational
radiation observatories; but we can obtain no reliable information whatever about what lies beyond the particle
horizon. We can in principle feel the gravitational effect of matter beyond the horizon because of the force it exerts
(for example, matter beyond the horizon may influence velocities of matter within the horizon, even though we
cannot see it). This is possible because of the constraint equations of general relativity theory, which are in effect
instantaneous equations valid on spacelike surfaces.25 However we cannot uniquely decode that signal to determine
what matter distribution outside the horizon caused it: a particular velocity field might be caused by a relatively
small mass near the horizon, or a much larger mass much further away [77]. Claims about what conditions are
like on very large scales — that is, much bigger than the Hubble scale — are unverifiable [52], for we have no
observational evidence as to what conditions are like far beyond the visual horizon. The situation is like that of
an ant surveying the world from the top of a sand dune in the Sahara desert. Her world model will be a world
25 They are valid at any late time in a solution of the EFE because they were valid initially — the initial data must satisfy constraint
equations — and once they are satisfied, the constraints are preserved by the dynamic field equations.
composed only of sand dunes — despite the existence of cities, oceans, forests, tundra, mountains, and so on beyond
her horizon.
It is commonly stated that if we live in a low-density universe and the cosmological constant vanishes, the
universe has infinite spatial sections. However this deduction only applies if firstly the RW-like nature of the
universe within the past light cone continues to be true indefinitely far outside it, and secondly the space sections
have their ‘natural’ simply-connected topology — and there is no way we can obtain observational evidence that
these conditions are both true. In contrast to this, in chaotic inflationary models (Sec.2.6), it is a definite prediction
that the universe will not be like a RW geometry on a very large scale — rather it will consist of many RW-like
domains, each with different parameter values, separated from each other by highly inhomogeneous regions outside
our visual horizon [141], the whole forming a fractal-like structure. This prediction is just as untestable as the
previously prevalent assumption (based on a Cosmological Principle) that the universe is RW-like on such scales
[14, 233]. Neither can be observationally confirmed or denied. The same issue arises in an even more extreme form
in relation to the idea of a multiverse. We return to this below, see Sec.9.2.
Small universes
There is one case where this kind of spatial observational limit does not obtain. This is when a Small Universe
occurs, that is, a universe which closes up on itself spatially for topological reasons [51], and does so on such a
small scale that we have seen right round the universe since the time of decoupling. Then we can see all the matter
that exists, with multiple images of many objects occurring [76]. This possibility is observationally testable by
examining source statistics, by observation of low power in the large angle CBR anisotropies, and by detecting
identical temperature variation on various circles in the CBR sky [130]. There are weak hints in the observed
CBR anisotropies (the lack of power on large angular scales) that this could actually be the case [143], but this is
not solidly confirmed. Checking if the universe is a small universe or not is an important task; the nature of our
observational relationship to the universe is fundamentally different if it is true [76].
The observed universe
The observable part of the universe (i.e. back to the visual horizon) is strictly limited, and we have already seen
most of it. We can only observe distant objects by electromagnetic radiation at all wavelengths, by neutrinos,
and by gravitational waves. We already have very complete broad coverage of the entire sky by electromagnetic
observations at all wavelengths right back to the surface of last scattering, which is the limit of what will ever be
observable by electromagnetic radiation. Detailed observations (such as the Hubble Deep Field) are available for
restricted domains in angle and depth. Detailed observations at suitable wavelengths are beginning to discern what
lies behind the Milky Way, which tends to obscure a substantial fraction of the sky. It is unlikely there are many
new astronomical phenomena undiscovered in this observable region, although it will be crucial determining more
detailed features of the phenomena we have already discovered (e.g. the nature of dark matter and dark energy).
Thesis B7: We have made great progress towards observational completeness. We have already
seen most of the part of the universe that is observable by electromagnetic radiation. It is plausible that not many
new astronomical phenomena remain to be discovered by us observationally; we will determine more details (so
understanding more about what we have seen) and see more objects, but not discover many new kinds of things.
Indeed Harwit [106] has used the multiplicity of discovery of specific astronomical phenomena to estimate how
many new essentially different such phenomena there are still waiting to be discovered.
Neutrinos and gravitational waves will in principle allow us to peer back to much earlier times (the time of
neutrino decoupling and the quantum gravity era respectively), but are much harder to observe at all, let alone in
useful directional detail. Nevertheless the latter has the potential to open up to us access to eras quite unobservable
in any other way. Maybe they will give us unexpected information on processes in the very early universe which
would count as new features of physical cosmology.
Issue C: The unbound energies in the early universe
The analogous problems for physical cosmology arise because energies occurring in the Hot Big Bang early universe
phase (Sec.2.2) are essentially unbounded, so the highest energies we can attain in particle accelerators cannot reach
the levels relevant to very early times. The uniqueness of cosmology in this regard is that it is the only science
contemplating spacetime regions that have experienced such high energies, and with which we are in intimate causal
contact despite the huge timescales involved — indeed events at those early times determined much of what we see
around us today. The nuclear reactions underlying nucleosynthesis are well understood, and their cross-sections
reasonably well-known; the processes of baryosynthesis and quark-gluon recombination are reasonably understood
and are on the border of being testable; but physical processes relevant at earlier times are inaccessible to testing by
laboratory or accelerator-based experiment. The Physics Horizon by definition separates those aspects of physics
we can hope to test by high-energy experiments on Earth or in the Solar System, from those where it is reasonable
to expect no such test will ever be possible:
Thesis C1: The Physics Horizon limits our knowledge of physics relevant to the very early
universe. We cannot experimentally test much of the physics that is important in the very early universe because
we cannot attain the required energies in accelerators on Earth. We have to extrapolate from known physics to the
unknown and then test the implications; to do this, we assume some specific features of known lower energy physics
are the true key to how things are at higher energies. We cannot experimentally test if we have got it right.
Note that this is independent of the issue of setting of initial conditions for the universe, considered below, see
Sec.6.2: the problem arises after the initial conditions have been set and the universe is running according to
invariable physical laws. We cannot be confident of the validity of the physics we presuppose then. Rather
than using known physics to predict the evolution of the universe, we end up testing proposals for this physics by
exploring their implications in the early universe, which is the only ‘laboratory’ where we can test some of our ideas
regarding fundamental physics at the highest energies [241]; this is particularly true in the case of quantum gravity
proposals. The problem is we cannot simultaneously do this and also carry out the aim of physical cosmology,
namely predicting the evolution of the early universe from known physical theory.
Our understanding of physics at those times has of necessity to be based on extrapolation of known physics
way beyond the circumstances in which it can be tested. The trick is to identify which features are the key
to use in that extrapolation: for example, variational principles, broken symmetries and phase changes, duality
invariance, entropy limits are candidates. If we confirm our guesses for the relevant physics by their satisfactory
implications for the early universe, tested in some suitable way, then this is impressive progress; but if this is
the only way we can test the proposed physics, the situation is problematic. If the hypothesis solves only the
specific issues it was designed to solve in the early universe and nothing else, then in fact it has little explanatory
power, rather it is just an alternative (perhaps theoretically preferable) description of the known situation. One
obtains positive observational support for a particular proposal for the relevant physics only if it predicts multiple
confirmed outcomes (rather than just one), for example predicting particles that are then confirmed to exist in
a laboratory, so that a single hypothesis simultaneously solves several different observational issues. Some of
the options may be preferred to others on various theoretical grounds; but one must distinguish this from their
having observational support. They lack physical power if they have no other testable consequences. A particular
example is the inflationary universe proposal (Sec.2.6): the supposed inflaton field underlying an inflationary era
of rapid expansion in the early universe [100, 91, 126, 101] has not been identified, much less shown to exist by
any laboratory experiment. Because this field φ is unknown, one can assign it an arbitrary potential V (φ), this
arbitrariness reflecting our inability to experimentally determine the relevant behaviour. It can be shown that
virtually any desired scale evolution S(t) of the universe can be attained by suitable choice of this potential [72];
and also almost any desired perturbation spectrum can be obtained by a (possibly different) suitable choice [140].
Indeed in each case one can run the mathematics backwards to determine the required potential V (φ) from the
desired outcome (Sec.9.3.1 below). The mathematical existence of such a theoretical potential of the desired form
for cosmological purposes does not by itself prove a particle or field exists with that effective potential.
Thesis C2: The unknown nature of the inflaton means that inflationary universe proposals
are incomplete. The promise of inflationary theory in terms of relating cosmology to particle physics has not
been realized. This will only be the case when the nature of the inflaton has been pinned down to a specific field
that experiment confirms or particle physics requires to exist.
The very impressive achievement of inflation is that the predicted CBR anisotropy spectrum is verified and agrees
with the matter power spectrum [48]; but that prediction depends only on the physics from the era of tight coupling
of matter and radiation to the present day, given a suitable initial fluctuation spectrum in the early universe, rather
than on the specific hypothesis of an inflationary origin for that spectrum. The true clincher would be if properties
of an inflationary field were predicted from the cosmology side and then confirmed in the laboratory; indeed
that would count as one of the great feats of theoretical physics. This may not happen however because of the
experimental problems focused on here, arising because we cannot reproduce on Earth all the conditions relevant
to very early cosmology.
One key application where this issue becomes significant is in respect of the chaotic inflation theory (Sec.2.6).
As remarked above, see Sec.4.3, its geometric predictions are observationally unverifiable. It would nevertheless
be a good physical prediction if it was a more or less inevitable outcome of known and tested underlying physics.
However this is not the case: the proposed underlying physics is not experimentally tested, indeed it is not even
uniquely defined or associated with any specific known physical particle or field. The claim that it inevitably follows
from string theory [212] suffers from the problem that string theory is not a well-defined or tested part of physics.
Issue D: Explaining the universe — the question of origins.
This is the unique core business of physical cosmology: explaining both why the universe has come into existence
and evolved to the present very high-symmetry FL geometry on large scales, and how structures come into existence
on smaller scales.
Start to the universe
Did a start to the universe happen? If so, what was its nature? This has been discussed above (Sec.2.7.2), and
the issue is unresolved. The major related question is whether the process of expansion only happens once in the
life of the Universe, or occurs repeatedly. The first option is the standard model, where the entire evolution of the
Universe is a once-off affair, with all the objects we see, and indeed the Universe itself, being transient objects that
will burn out like dead fireworks after a firework display. In this case everything that ever happens occurs during
one expansion phase of the Universe (possibly followed by one collapse phase, which could occur if k = +1 and
the present ‘dark energy’ field dies away in the future). This evolution might have a singular start at a space-time
singularity; a beginning where the nature of time changes character; a non-singular bounce from a single previous
collapse phase; or a start from a non-singular static initial state [155]. An alternative is that many such phases
have occurred in the past, and many more will occur in the future; the Universe is a Phoenix Universe [34], new
expansion phases repeatedly arising from the ashes of the old. While the idea of one or more bounces is an old one
[220], actual mechanisms that might allow this bounce behaviour have not yet been elucidated in a fully satisfactory
way. A variant is the chaotic inflation idea of new expanding universe regions arising from vacuum fluctuations
in old expanding regions, leading to a universe that has a fractal-like structure at the largest scales, with many
expanding regions with different properties emerging out of each other in a universe that lasts forever (Sec.2.6).
As discussed above, see Sec.2.7.1, it is possible (if the universe has positive spatial curvature) that the
quantum gravity domain can be avoided and there was no start to the universe; however this probably requires
special initial conditions [70]. If a quantum gravity domain indeed occurred, we cannot come to a definite conclusion
about whether there was a creation event or not because we do not know the nature of quantum gravity, nor
how to reliably apply it in the cosmological context where the issue of initial conditions arises. Loop quantum
gravity suggests the universe may be singularity-free [12], with bounces or a non-singular start, but that theory is
unconfirmed. Tested physics cannot give a decisive answer; it is possible that testable physics also cannot do so.
Thesis D1: An initial singularity may or may not have occurred. A start to the universe may
have occurred a finite time ago, but a variety of alternatives are conceivable: eternal universes, or universes where
time as we know it came into existence in one or another way. We do not know which actually happened, although
quantum gravity ideas suggest a singularity might be avoided.
This is a key issue in terms of the nature of the universe: a space-time singularity is a dramatic affair, where the
universe (space, time, matter) has a beginning and all of physics breaks down and so the ability to understand
what happens on a scientific basis comes to an end. However eternal existence is also problematic, leading for
instance to the idea of Poincaré’s eternal return: everything that ever happened will recur an infinite number of
times in the future and has already occurred an infinite number of times in the past [7]. This is typical of the
problems associated with the idea of infinity (discussed further below, see Sec.9.3.2). It is not clear in the end
which is philosophically preferable: a singularity or eternal existence. That decision will depend on what criteria
of desirability one uses (such criteria are discussed below, see Sec.8.1).
The issue of initial conditions
While occurrence of an initial singularity is striking in that it is a start to physics and spacetime as well as matter,
whether it occurred or not is in a sense irrelevant to the key issue of what determined the nature of the universe:
Thesis D2: Testable physics cannot explain the initial state and hence specific nature of the
universe. A choice between different contingent possibilities has somehow occurred; the fundamental issue is
what underlies this choice. Why does the universe have one specific form rather than another, when other forms
consistent with physical laws seem perfectly possible? The reasons underlying the choice between different contingent
possibilities for the universe (why one occurred rather than another) cannot be explored scientifically. It is an issue
to be examined through philosophy or metaphysics.
Even if a literal creation does not take place, as is the case in various of the present proposals, this does not resolve
the underlying issue of what determined why the universe is the way it is, given that it could presumably have
been otherwise. If the proposal is evolution from a previous eternal state — Minkowski space for example — then
why did that come into existence, and why did the universe expansion as a bubble from that vacuum start when
it did, rather than at some previous time in the pre-existent eternity? Whenever it started, it could have started
before! Some attempts involve avoiding a true beginning by going back to some form of eternal or cyclic initial
state, for example Tolman’s series of expansion and collapse cycles [220], proposals for creation of the universe as
a bubble formed in a flat space-time [221], Linde’s eternal chaotic inflation [141], Veneziano’s re-expansion from a
previous collapse phase [90], the ekpyrotic universe proposal [124], and theories involving foundational limits on
information through a “holographic principle” [213]. These do not avoid the ultimate problem; it can be claimed
they simply postpone facing it, for one now has to ask all the same questions of origins and uniqueness about
the supposed prior state to the Hot Big Bang expansion phase. The Hartle-Hawking ‘no-boundary’ proposal [109]
avoids the initial singularity because of a change of space-time signature, and so gets round the issue of a time of
creation in an ingenious way; and Gott’s causality violation in the early universe [95] does the same kind of thing
in a different way. Such proposals cannot overcome the ultimate existential question: Why has one specific state
occurred rather than any of the other possibilities? How was it decided that this particular kind of universe would
be the one actually instantiated? This question cannot be solved by physics alone, unless one can show that only
one form of physics is self-consistent; but the variety of proposals made is evidence against that suggestion.
The explanation of initial conditions has been the aim of the family of theories one can label collectively as
‘quantum cosmology’ [109, 95, 92]; however as discussed earlier, here we inevitably reach the limits to what the
scientific study of the cosmos can ever say — if we assume that such studies must of necessity involve an ability to
observationally or experimentally check our theories. No physical experiment at all can help here because of the
uniqueness of the universe, and the feature that no spacetime exists prior to (in a causal sense) such a beginning;
so brave attempts to define a ‘physics of creation’ stretch the meaning of ‘physics’. Prior to the start (if there
was a start) physics as we know it is not applicable and our ordinary language fails us because time did not exist,
so our natural tendency to contemplate what existed or happened ‘before the beginning’ is highly misleading —
there was no ‘before’ then, indeed there was no ‘then’ then! Talking as if there was is commonplace, but quite
misleading in trying to understand a scientific concept of ‘creation’ [97]. We run full tilt into the impossibility
of testing the causal mechanisms involved, when physics did not exist. No experimental test can determine the
nature of any mechanisms that may be in operation in circumstances where even the concepts of cause and effect
are suspect. This comes particularly to the fore in proposing ‘laws of initial conditions for the universe’ — for here
we are apparently proposing a theory with only one object. Physics laws are by their nature supposed to cover
more than one event, and are untestable if they do not do so (Sec.3).
Special or general
The present state of the universe is very special. Explanation of the present large-scale isotropy and homogeneity
of the universe means determining the dynamical evolutionary trajectories relating initial to final conditions, and
then essentially either explaining initial conditions, where we run into difficulties (Sec.6.2), or showing they are
irrelevant. The issue raised is whether the universe started off in a very special geometrical state:
Thesis D3: The initial state of the universe may have been special or general. Whether there
was generality or speciality of geometrical initial conditions for the universe is a key question. It seems likely that
the initial state of the observed part of the universe was not generic.
The assumption that the universe is geometrically special was encoded in the Cosmological Principle, taken
as a founding principle in cosmology until the 1960’s, i.e. as an ‘explanation’ of special initial conditions [14, 233].
Then Misner introduced the chaotic cosmology programme [151], based on the idea of a universe with generic
initial conditions being isotropised at later times by physical processes such as viscosity, making initial conditions
irrelevant. This concept of isotropisation then became central to the inflationary family of theories (Sec.2.6), with
the underlying assumption being that ‘fine tuning’ of initial conditions is unphysical and to be avoided. Both
programmes are however only partially successful: one can explain a considerable degree of isotropisation and
homogenization of the physical universe by either process, but this will not work in all circumstances. Inflation
can get rid of much anisotropy [228] but inhomogeneity must be restricted if inflation is to succeed in producing
a universe like that we see today, and the success of inflation in solving the horizon issue for FL models — where
exact homogeneity exists to start with — will not necessarily be replicated in anisotropic models. Universes that
are initially too anisotropic may never inflate, and the horizon problem may not be solved in such models if they
do;26 and only rather special states lead to ordinary thermodynamics [167, 169, 229, 23], which is taken to be true
in inflationary physics.
Inflation can only be guaranteed to succeed if initial conditions are somewhat restricted; some degree of
geometric speciality must have occurred at the start of the observed region of the universe. This special domain
might possibly occur within the context of a much larger universe domain where conditions vary randomly, and
only isolated regions lead to inflation and eventually domains such as that we see around us; attractive as this may
be, it is an untestable hypothesis (essentially a version of the multiverse proposal, see Sec.9.2).
Special initial conditions (which inflation proposes to explain) might have just occurred that way. The
ultimate issue is that we have no proof as to whether initial conditions for the universe were special or general;
26 Most inflationary studies show only that the geometric horizon problem is solved in the very special RW geometries; but there is
no physical horizon problem in those geometries, for they are by by assumption spatially homogeneous and isotropic ab initio.
either could have occurred. If we state these conditions must have been general, we are making a philosophical
claim, for it is not a provable physical statement. Part of the problem is that we have no agreed measure on the
space of possible universes; what seems special or general depends on the choice of such a measure.
Issue E: The Universe as the Background for Existence
The universe provides the environment for all of science, by determining the initial conditions within which all
physical laws are constrained to operate, thus setting boundary conditions for all local physics. Together with
suitable equations of state for the matter or structural equations for complex systems, these determine the nature
of physical outcomes. The uniqueness of cosmology lies in that it considers the origin of such conditions.
Laws and boundary conditions
A fundamental assumption underlying physical cosmology is the idea that the laws of physics are the same everywhere in the physical universe: those we determine in a laboratory here and now will be the same as apply at very
distant places (e.g. determining the astrophysics of qso’s at redshift z = 6), at very early times (e.g. at the time of
nucleosynthesis), and at very late times. Without this assumption, explanatory theories have no solid foundation.
However because of the uniqueness of the universe discussed above (see Sec.3), unlike the rest of physics where
the distinction is clear and fundamental, in the cosmological context the distinction between laws and boundary
conditions becomes blurred.
Thesis E1: Physical laws may depend on the nature of the universe. We have an essential difficulty
in distinguishing between laws of physics and boundary conditions in the cosmological context of the origin of the
universe. Effective physical laws may depend on the boundary conditions of the universe, and may even vary in
different spatial and/or temporal locations in the cosmos.
Because we cannot vary the initial conditions in any way, as far as we are concerned they are necessary rather
than contingent — so the essential distinction between initial conditions and laws is missing. The distinction is
clear once the cosmos has come into existence — but we are concerned with ‘prior’ conditions associated with the
creation of the cosmos and the very existence of physical laws. Certainly any proposal for distinguishing between
laws of nature and boundary conditions governing solutions to those laws is untestable in this context. Given the
feature that the universe is the unique background for all physics, it is therefore not far-fetched to suggest that it
is possible the cosmos influences the nature of local physical laws, rather than just their initial conditions [77, 61].
This has been examined over many decades in three specific cases.
(a) Varying ‘constants’ : It might be that there is a time variation in physical constants of nature [6] related
to the expansion of the universe, as proposed in the case of the gravitational constant G by Dirac [36], developed
in depth by Jordan and then Brans and Dicke [20]. Such proposals must be consistently developed in relation to
the rest of physics and should be related to dimensionless constants, as otherwise they may simply be disguised
variations in the units of measurements used, rather than being a genuine physical change (some claims that the
speed of light ‘c’ may vary fall into this category [82]). This proposal has received impetus in recent times from
ideas based in quantum field theory and string theory, suggesting that many of the ‘constants of nature’ are in
fact contingent, depending on the nature of the vacuum state [211]. This kind of proposal is to some degree open
to observational test [27, 238], and in the cases where it has been investigated it seems that it does not occur in
the visible region of the universe — the constants of nature are indeed invariant, with one possible exception: the
fine structure constant, where there is claimed to be evidence of a very small change over astronomical timescales
[6]. That issue is still under investigation. Testing such invariance is fundamentally important, precisely because
cosmology usually assumes as a ground rule that physics is the same everywhere in the universe. If this were
not true, local physics would not guide us adequately as to the behaviour of matter elsewhere or at other times,
and cosmology would become an arbitrary guessing game. In order to proceed in a scientific manner when such
variation is proposed, one needs then to hypothesize the manner of such variation. Thus the old laws where G
was constant are replaced by new laws governing its time variation [20]; the principle of nature being governed
by invariant (unchanging) physical laws and associated constants remains.27 Thus in the end the proposal is to
replace simpler old laws by new more complex ones. These must then be assumed invariant, or we cannot proceed
(b) Inertia and Mach’s Principle: It might be that the local inertial properties of matter are determined by
the distant distribution of matter in the universe, so that if the universe were different, inertia would be different.
This is the complex of ideas referred to as Mach’s principle [3], which served as a major impetus for Einstein’s
cosmological ideas. The precise meaning and implications of this idea remain controversial.
27 “Despite the incessant change and dynamic of the visible world, there are aspects of the fabric of the universe which are mysterious
in their unshakeable constancy. It is these mysterious unchanging things that make our universe what it is and distinguish it from
other worlds we might imagine” (Barrow [6], p. 3).
(c) The arrow of time: The existence and direction of the macroscopic arrow of time in physics — and hence
in chemistry, biology, psychology, and society — is related to boundary conditions in the past and future of the
universe. The fundamental physical laws by themselves are time symmetric, and so unable to explain this feature
[28, 77, 242, 223]. A recent argument of this kind is Penrose’s claim that the existence of the arrow of time is
crucially based in the universe having had rather special initial conditions [168, 167, 229]. Thus what appears in
ordinary physics as an immutable law of nature (viz. the Second Law of Thermodynamics with a given arrow of
time) may well be the result of specific boundary conditions at the start and end of the universe. It might not be
true in all universes, even if the underlying fundamental physical laws are the same.
In each case proposals have been made as to the possible nature of the deeper underlying unchanging laws, and the
relations between the state of the universe and the resultant effective laws in that context. This is also proposed
in ‘landscape’ of possibilities of string theory [212]. These proposals are however intrinsically untestable, for the
reasons explained above (Sec.3): we cannot change the boundary conditions of the universe and see what happens;
but they do serve as a continuing fertile source of ideas.
Alternative physics
In any case, the important conclusion is that it is certainly appropriate for cosmology to consider what would have
happened if, not only the boundary conditions at the beginning of the universe, but also the laws of physics had
been different [212]:
Thesis E2: We cannot take the nature of the laws of physics for granted. Cosmology is interested in
investigating hypothetical universes where the laws of physics are different from those that obtain in the real universe
in which we live — for this may help us understand why the laws of physics are as they are (a fundamental feature
of the real physical universe).
One cannot take the existence and nature of the laws of physics (and hence of chemistry) as unquestionable in
cosmology — which seems to be the usual habit in biological discussions on the origin and evolution of life. This is
in stark contrast to the rest of science, where we are content to take the existence and nature of the laws describing
the fundamental behaviour of matter as given and unchangeable. Cosmological investigation is interested in the
properties of hypothetical universes with different physical behaviour. Consideration of ‘what might have been’ is a
useful cosmological speculation that may help throw light on what actually is; this is a statement of the usefulness
of ‘Gedanken experiments’ in cosmology.
Indeed if one wants to investigate issues such as why life exists in the universe, consideration of this larger
framework — in essence, a hypothetical ensemble of universes with many varied properties — is essential (this
is of course not the same as assuming an ensemble of such universes actually exists, cf. the discussion below in
Sec.9.2). However we need to be very cautious about using any claimed statistics of universes in such a hypothetical
ensemble of all possible or all conceivable universes. This is usually not well defined, and in any case is only relevant
to physical processes if either the ensemble actually exists, rather than being a hypothetical one, or if it is the
outcome of processes that produce well-defined probabilities — an untestable proposal. We can learn from such
considerations the nature of possible alternatives, but not necessarily the probability with which they might occur
(if that concept has any real meaning).
Emergence of complexity
As the universe evolves an increase of complexity takes place in local systems as new kinds of objects come into
being that did not exist before — nuclei, atoms, stars and galaxies, planets, life, consciousness, and products of
the mind such as books and computers [153]. New kinds of physical states come into being at late times such as
Bose-Einstein condensates, that plausibly cannot exist without the intervention of intelligent beings.
Thesis E3: Physical novelty emerges in the expanding universe. New kinds of physical existence
come into being in the universe as it evolves, that did not exist previously. Their existence is allowed by the
boundary conditions provided by the universe for local systems, together with the possibility space generated by the
underlying physics. While their physical existence is novel, every new thing that comes into being is foreshadowed
in possibility structures that precede their existence.
Physical existence is new as the universe evolves, but there had to be precursors of the novel in the possibility space
allowed by physics, so that they could come into being. In this sense the truly novel does not emerge ex nihilo
but rather is discovered. The universe is the environment that allows this to happen. The nature of the features
leading to the existence of life, and their possible causes, is discussed in Sec.9.1.
Issue F: The Explicit Philosophical Basis
Consequent on the discussion above, and particularly items B6, C2, and D2, it follows that
Thesis F1: Philosophical choices necessarily underly cosmological theory. Unavoidable metaphysical issues inevitably arise in both observational and physical cosmology. Philosophical choices are needed in order
to shape the theory.
There is of course always a philosophical basis to any scientific analysis, namely adoption of the basic scientific
method and a commitment to the attempt to explain what we see as far as possible simply in terms of causal
laws, ultimately based in physics. This will clearly be true also in cosmology. However we need further explicit
philosophical input in order to attain specific geometric models — for example a Copernican principle, as explained
above, see Sec.4.2.2 — and to determine what form physical cosmology should take in the very early universe, for
example deciding which physical principle to use as the core of one’s extrapolation of known physics to the unknown
(Sec.5). Underlying both sets of choices are criteria for satisfactoriness of a cosmological model, which help decide
which feature to focus on in formulating a theory. Of particular importance is the scope chosen for our cosmological
theory; together with the choice of criteria for a good theory, this is a philosophical decision that will shape the rest
of the analysis. Some cosmologists tend to ignore the philosophical choices underlying their theories; but simplistic
or unexamined philosophical standpoints are still philosophical standpoints!
Criteria for theories
As regards criteria for a good scientific theory [129], typical would be the following four areas of assessment:
1. Satisfactory structure: (a) internal consistency, (b) simplicity (Ockham’s razor), and (c) aesthetic appeal
(‘beauty’ or ‘elegance’).
2. Intrinsic explanatory power : (a) logical tightness, (b) scope of the theory — the ability to unify otherwise
separate phenomena, and (c) probability of the theory or model with respect to some well-defined measure;
3. Extrinsic explanatory power, or relatedness: (a) connectedness to the rest of science, (b) extendability
— providing a basis for further development;
4. Observational and experimental support, in terms of (a) testability: the ability to make quantitative as
well as qualitative predictions that can be tested; and (b) confirmation: the extent to which the theory is supported
by such tests as have been made.
It is particularly the latter that characterizes a scientific theory, in contrast to other types of theories claiming
to explain features of the universe and why things happen as they do. It should be noted that these criteria are
philosophical in nature in that they themselves cannot be proven to be correct by any experiment. Rather their
choice is based on past experience combined with philosophical reflection. One could attempt to formulate criteria
for good criteria for scientific theories, but of course these too would need to be philosophically justified. The
enterprise will end in infinite regress unless it is ended at some stage by a simple acceptance of a specific set of
Thesis F2: Criteria of satisfactoriness for theories cannot be scientifically chosen or validated.
Criteria of satisfactoriness are necessary for choosing good cosmological theories; these criteria have to be chosen
on the basis of philosophical considerations. They should include criteria for satisfactory structure of the theory,
intrinsic explanatory power, extrinsic explanatory power, and observational and experimental support.
The suggestion here is that the above proposed criteria are a good set to use in investigating cosmology; they
include those most typically used ([129]; and see [169, 212] for comments on such criteria).
Conflicts between criteria.
These criteria are all acknowledged as desirable. The point then is that generally in pursuing historical sciences,
and in particular in the cosmological context, they will not all be satisfied to the same degree, and may even lead
to opposing conclusions:
Thesis F3: Conflicts will inevitably arise in applying criteria for satisfactory cosmological
theories. Philosophical criteria for satisfactory cosmological theories will in general come into conflict with each
other, so that one will have to choose between them to some degree; this choice will shape the resulting theory. [58].
The thrust of much recent development has been away from observational tests towards strongly theoretically
based proposals, indeed sometimes almost discounting observational tests. At present this is being corrected by a
healthy move to detailed observational analysis of the consequences of the proposed theories, marking a maturity
of the subject. However because of all the limitations in terms of observations and testing [criteria (4)], in the
cosmological context we still have to rely heavily on other criteria, and some criteria that are important in most
of science may not really make sense. This is true of 2(c) in particular, as discussed above, see Sec.3; nevertheless
many approaches still give the idea of probability great weight. At a minimum, the ways this can make sense
needs exploration and explication. Furthermore the meaning of some of the criteria may come into dispute. 1(b)
is clearly a case in point : for example, is the idea of an existent ensemble of universes displaying all possible
behaviours simple (because it is a single idea that can be briefly stated), or immensely complex (because that
statement hides all the complexities and ambiguities involved in the idea of an infinity of possibilities)? 1(c) is
also controversial(‘beauty is in the eye of the beholder’), see [212] for a discussion.
The tenor of scientific understanding may change, altering the balance of what is considered a good explanation and what is not. An example [57] is the way cosmologists strongly resisted the idea of an evolving universe in
the 1920’s, at a time when biological evolution was very well established but the idea of continental drift was also
being strongly resisted. The change to an appreciation of the explanatory power of an evolving model came later
in both cases; but even then in the cosmological case, for either aesthetic or metaphysical reasons, some still sought
for a steady state description, resisting the implication of a beginning to the universe. That tendency is still with us
today, in the form of models that are eternal in one way or another (e.g. some forms of chaotic inflation). Another
example is the change from supposition of underlying order, expressed in the idea of a Cosmological Principle, to
a broad supposition of generic disordered conditions, embodied in the ideas of inflation. Associated with this is a
shift from making geometric assumptions to providing physical explanatory models. It is this shift that underlies
the major present support for inflation:
Thesis F4: The physical reason for believing in inflation is its explanatory power as regards
structure growth in the universe. Inflation predicts the existence of Gaussian scale-free perturbations in
the early universe thereby (given the presence of cold dark matter) explaining bottom-up structure formation in a
satisfactory way. This theory has been vindicated spectacularly through observations of the CBR and matter power
spectra. It is this explanatory power that makes it so acceptable to physicists, even though the underlying physics
is neither well-defined nor tested, and its major large-scale observational predictions are untestable.
The physical explanatory power of inflation in terms of structure formation, supported by the observational
data on the fluctuation spectra, is spectacular. For most physicists, this trumps the lack of identification and
experimental verification of the underlying physics (Sec.5). Inflation provides a causal model that brings a wider
range of phenomena into what can be explained by cosmology (Criterion 2(b)), rather than just assuming the
initial data had a specific restricted form. Explaining flatness (Ω0 ≃ 1 as predicted by inflation) and homogeneity
reinforces the case, even though these are philosophical rather than physical problems (they do not contradict any
physical law; things could just have been that way). However claims on the basis of this model as to what happens
very far outside the visual horizon (as in the chaotic inflationary theory) results from prioritizing theory over the
possibility of observational and experimental testing [44]. It will never be possible to prove these claims are correct.
The scope of cosmology
To sensibly choose priorities for the criteria just discussed, we need an answer to the question, How much should
we try to explain?
Thesis F5: Cosmological theory can have a wide or narrow scope of enquiry. The scope we
envisage for our cosmological theory shapes the questions we seek to answer. The cosmological philosophical base
becomes more or less dominant in shaping our theory according to the degree that we pursue a theory with more or
less ambitious explanatory aims in terms of all of physics, geometry, and underlying fundamental causation.
This is a choice one has to make, as regards both foundations and outcomes. Given a decision on this, one can
sensibly debate what is the appropriate philosophical position to adopt in studying a cosmological theory with that
scope. The study of expansion of the universe and structure formation from nucleosynthesis to the present day is
essential and well-informed. The philosophical stance adapted is minimal and highly plausible. The understanding
of physical processes at earlier times, back to quantum gravity, is less well founded. The philosophical stance is more
significant and more debatable. Developments in the quantum gravity era are highly speculative; the philosophical
position adopted is dominant because experimental and observational limits on the theory are lacking.
One can choose the degree to which one will pursue the study of origins [85] back to earlier and earlier times
and to more fundamental causal issues, and hence the degree to which specific philosophical choices are dominant
in one’s theory. The basic underlying cosmological questions are [58]:
(1) Why do the laws of physics have the form they do? Issues arise such as what makes particular laws
work? For example, what guarantees the behaviour of a proton, the pull of gravity? What makes one set of physical
laws ‘fly’ rather than another? If for example one bases a theory of cosmology on string theory [212], then who
or what decided that quantum gravity would have a nature well described by string theory? If one considers all
possibilities, considering string theory alone amounts to a considerable restriction.
(2) Why do boundary conditions have the form they do ? The key point here (Sec.6.2), is how are specific
contingent choices made between the various possibilities, for example whether there was an origin to the universe
or not.
(3) Why do any laws of physics at all exist ? This relates to unsolved issues concerning the nature of the
laws of physics: are they descriptive or prescriptive? (Sec.9.3.3). Is the nature of matter really mathematically
based in some sense, or does it just happen that its behaviour can be described in a mathematical way?
(4) Why does anything exist ? This profound existential question is a mystery whatever approach we take.28
Finally the adventurous also include in these questions the more profound forms of the contentious Anthropic
question [21, 29, 7, 214, 212]:
(5) Why does the universe allow the existence of intelligent life? This is of somewhat different character
than the others and largely rests on them but is important enough to generate considerable debate in its own right.
The status of all these questions is philosophical rather than scientific, for they cannot be resolved purely scientifically. How many of them — if any — should we consider in our construction of and assessments of cosmological
One option is to decide to treat cosmology in a strictly scientific way, excluding all the above questions,
because they cannot be solved scientifically. One ends up with a solid technical subject that by definition excludes
such philosophical issues. This is a consistent and logically viable option. This logically unassailable position
however has little explanatory power; thus most tend to reject it because of criteria 2(b) and 3 above.
The second option is to decide that these questions are of such interest and importance that one will tackle
some or all of them, even if that leads one outside the strictly scientific arena. It is here that criteria 2 and 3 above
are to some degree in conflict with criterion 4. Thus if we try to explain the origin of the universe itself, these
philosophical choices become dominant precisely because the experimental and observational limits on the theory
are weak; this can be seen by viewing the variety of such proposals that are at present on the market.
Limits of Representation and Knowledge of Reality
It follows from the above discussion that there are limits to what the scientific method can achieve in explanatory
terms. We need to respect these limits and acknowledge clearly when arguments and conclusions are based on
some philosophical stance rather than purely on testable scientific argument. If we acknowledge this and make that
stance explicit, then the bases for different viewpoints are clear and alternatives can be argued about rationally.
A crucial underlying feature here is relating the nature of epistemology to ontology: how do we relate
evidence to our theories of existence? A further key issue is the relation of models to reality:
Thesis F6: Reality is not fully reflected in either observations or theoretical models. Problems
arise from confusion of epistemology (the theory of knowledge) with ontology (the nature of existence): existence
is not always manifest clearly in the available evidence. The theories and models of reality we use as our basis for
understanding are necessarily partial and incomplete reflections of the true nature of reality, helpful in many ways
but also inevitably misleading in others. They should not be confused with reality itself !
The confusion of epistemology with ontology occurs all the time, underlying for example the errors of both logical
positivism and extreme relativism. In particular, it is erroneous to assume that lack of evidence for the existence of
some entity is proof of its non-existence. In cosmology it is clear for example that regions may exist from which we
can obtain no evidence (because of the existence of horizons); so we can sometimes reasonably deduce the existence
of unseen matter or regions from a sound extrapolation of available evidence (no one believes matter ends at or
just beyond the visual horizon). However one must be cautious about the other extreme, assuming existence can
always be assumed because some theory says so, regardless of whether there is any evidence of existence or not.
This happens in present day cosmology, for example in presentations of the case for multiverses, even though the
underlying physics has not been experimentally confirmed. It may be suggested that arguments ignoring the need
for experimental/observational verification of theories ultimately arise because these theories are being confused
with reality, or at least are being taken as completely reliable total representations of reality. This occurs in
• Confusing computer simulations of reality with reality itself, when they can in fact represent only a highly
simplified and stylized version of what actually is;
• Confusing the laws of physics themselves with their abstract mathematical representation (if indeed they are
ontologically real, c.f. Sec.10.1), or confusing a construction of the human mind (‘Laws of Physics’) with the
reliable behaviour of ponderable matter (if they are not ontologically real);
• Confusing theoretically based outcomes of models with proven observational results (e.g. claiming the universe
28 But
see Grunbaum [98] for a dissenting view.
necessarily has flat spatial sections: Ω0 = 1, and so this can be taken for granted, when the value of Ω0 can
and should be observationally determined precisely because this then tests that prediction).
No model (literary, intuitive, or scientific) can give a perfect reflection of reality. Such models are always selective
in what they represent and partial in the completeness with which they do so. The only model that would reflect
reality fully is a perfect fully detailed replica of reality itself! This understanding of the limits of models and
theories does not diminish the utility of these models; rather it helps us use them in the proper way. This is
particularly relevant when we consider how laws of nature may relate to the origins of the universe itself, and to
the existence and nature of life in the expanding universe. The tendency to rely completely on our theories, even
when untested, seems sometimes to arise because we believe they are the same as reality — when at most they are
descriptions of reality.
Key Issues
There are some interrelated key issues where the features identified above either are at the heart of current debates,
or are likely to be at the heart of future debates. They are: the reason cosmological conditions allow the existence of
life (anthropic issues), the closely related issue of the possible existence of multiverses; and the natures of existence,
including the questions of the existence of infinities and the nature of the laws of physics. We look at them in turn
in this section. To some degree they have already been considered above,but they are specifically featured here
because of the important role they will probably play in discussion in the future.
Issue G: The anthropic question: Fine tuning for life
One of the most profound fundamental issues in cosmology is the Anthropic question, see [29, 7, 40, 85, 30, 2, 175,
176, 6]: why does the Universe have the very special nature required in order that life can exist?. The point is that
a great deal of “fine tuning” is required in order that life be possible. There are many relationships embedded
in physical laws that are not explained by physics, but are required for life to be possible; in particular various
fundamental constants are highly constrained in their values if life as we know it is to exist:
“A universe hospitable to life — what we might call a biophilic universe — has to be special in many ways ... Many
recipes would lead to stillborn universes with no atoms, no chemistry, and no planets; or to universes too short
lived or too empty to evolve beyond sterile uniformity” [176].
How has it come about that the Universe permits the evolution and existence of intelligent beings at any time or
place? “What features of the universe were essential for creatures such as ourselves, and is it through coincidence
or for some deeper reason that our universe has these features?” [96]. Whether one regards this as an appropriate
issue for cosmology to discuss depends, as discussed above (Sec.8.2), on the scope one envisages for cosmology. The
viewpoint taken here will be that this is one of the major issues one might wish to explain, and indeed a substantial
literature considers this. Here we explore the nature of this fine tuning, and then consider possible answers as to
how it arises. There are three aspects that we consider in turn (cf. [212]).
Laws of physics and the existence of complexity
The laws of physics and chemistry are such as to allow the functioning of living cells, individuals, and ecosystems
of incredible complexity and variety, and it is this that has made evolution possible. What requires explanation,
is why the laws of physics are such as to allow this complex functionality to work, without which no evolution
whatever would occur. We can conceive of universes where the laws of physics (and so of chemistry) were different
than in ours. Almost any change in these laws will prevent life as know it from functioning.
The first requirement is the existence of laws of physics that guarantee the kind of regularities that can
underlie the existence of life. These laws as we know them are based on variational and symmetry principles; we
do not know if other kinds of laws could produce complexity. If the laws are in broad terms what we presently take
them to be, the following inter alia need to be right, for life of the general kind we know to exist [29, 96]:
• Quantization that stabilizes matter and allows chemistry to exist through the Pauli exclusion principle.
• The neutron-proton mass differential must be highly constrained. If the neutron mass were just a little less
than it is, proton decay could have taken place so that by now no atoms would be left at all [29].
• Electron-proton charge equality is required to prevent massive electrostatic forces overwhelming the weaker
electromagnetic forces that govern chemistry.
• The strong nuclear force must be strong enough that stable nuclei exist [29]; indeed complex matter exists only
if the properties of the nuclear strong force lies in a tightly constrained domain relative to the electromagnetic
force [216].
• The chemistry on which the human body depends involves intricate folding and bonding patterns that would
be destroyed if the fine structure constant (which controls the nature of chemical bonding) were a little bit
• The number D of large spatial dimensions must be just 3 for complexity to exist [216, 176].
Hogan has examined the freedom in the parameters of the standard model of particle physics and concluded that
5 of the 17 free parameters of the standard model must lie in a highly constrained domain if complex structures
are to exist [113]. This is of course taking the basic nature of the standard model of particle physics for granted.
If this were not so, it is difficult to determine what the constraints would be. However his study is sufficient to
show that whatever the nature of fundamental physics, and in particular of particle physics, may be, only a small
subset of all possible laws of physics will be compatible with the existence of complexity.
Laws of physics and the existence of congenial environments
The creation through astrophysical processes of suitable habitats for life to exist (the existence of planets circling
stable stars, for example) depends to some degree on the nature of the fundamental physical laws. If the laws are
in broad terms what we presently take them to be, the requirements for such habitats to exist include:
• The gravitational force must create large stable structures (planets and stars) that can be the habitat for
life and their energy source respectively. This requires the gravitational force to be very weak relative to
electrical forces. The ratio N of the strength of the electromagnetic force to the gravitational force must be
close to the observed value: N ≃ 1036 ([175], Chapter 3).
• The weak force must allow helium production that leaves sufficient hydrogen over; it is related to gravity
through a numerical factor of 10−11 , which cannot be much different. And for this to work, the neutron-proton
mass difference must be close to the mass of the electron [29].
• A stellar balance should allow a long lifetime for stars like the sun, so allowing the transmutation of the
light elements into heavy elements. This requires that the nuclear fusion efficiency E be close to the observed
value: E ≃ 0.007 ([175], Chapter 4).
• One needs to overcome the beryllium “bottleneck” in the making of heavy elements through nuclear reactions
in stars [96, 212]. The production of carbon and oxygen in stars requires the careful setting of two different
nuclear energy levels to provide a resonance; if these levels were just a little different, the elements we need
for life would not exist [85]. Indeed it was on this basis that Hoyle famously predicted a carbon-12 energy
level that has since been experimentally confirmed.
• One needs something like the existence of neutrinos and the weak interaction with its specific coupling
constant in order to underly supernovae explosions that spread heavy elements through space, as seeds for
planetary formation [96].
• The nuclear force must be weak enough that di-protons do not exist, otherwise no protons will be left over
to enable heavier elements to exist [29].
• The neutrino mass must not be too high, or the universe will not last long enough [29].
Cosmological boundary/initial conditions and congenial environments
Finally, given laws of physics that are suitable in terms of satisfying the requirements of both the previous sections,
the universe itself must also be suitable, in terms of its initial or boundary conditions, for life to exist. If the laws
of physics are basically the same as we now believe them to be, these cosmological requirements include
• The size of the universe and its age must be large enough. There could be universes that expanded and then
recollapsed with a total lifetime of only 100,000 years; we need a sufficiently old universe for second generation
stars to come into existence and then for planets to have a stable life for long enough that evolution could
lead to the emergence of intelligent life. Thus the universe must be at about 15 billion years old for life to
exist [96], hence we must have Ωmatter ≃ 0.3 ([175], Chapter 6).
• The size of the cosmological constant must not be too large, or galaxies will not form; we need |ΩΛ | < 1 for
galaxies to exist ([175], Chapter 7; [212]).
• The seeds in the early universe for fluctuations that will later grow into galaxies must be of the right size that
structures form without collapsing into black holes: the number Q characterizing the size of primordial ripples
on the LSS (and hence the geometry of the perturbed cosmological model, see Sec.2.5.2) must therefore be
of the order Q ≃ 10−5 ([175], Chapter 8).
The complex of interacting systems in a human body could not possibly work if a series of delicate conditions
were not maintained. For example, the background radiation might never drop below 3000 K, so that matter was
always ionized (electrons and nuclei always remaining separate from each other); the molecules of life could then
never form. Black holes might be so common that they rapidly attracted all the matter in the universe, and there
never was a stable environment in which life could develop. Cosmic rays could always be so abundant that any
tentative organic structures were destroyed before they could replicate. Overall,
• There must be non-interference with local systems. The concept of locality is fundamental, allowing local
systems to function effectively independently of the detailed structure of the rest of the Universe. We need
the universe and the galaxies in it to be largely empty, and gravitational waves and tidal forces to be weak
enough,29 so that local systems can function in a largely isolated way [61].
• The fact that the night sky is dark (‘Olbers’ paradox’ [14, 104]) is a consequence of the expansion of the
universe together with the photon to baryon ratio. This feature is a necessary condition for the existence of
life: the biosphere on Earth functions by disposing of waste energy to the heat sink of the dark night sky
[168]. Thus one way of explaining why the sky is observed to be dark at night is that if this were not so, we
would not be here to observe it.
• The existence of the arrow of time, and hence of laws like the second law of thermodynamics, are probably
necessary for evolution and for consciousness. This depends on boundary conditions at the beginning and
end of the Universe (Sec.7.1).
• Presumably the emergence of a classical era out of a quantum state is required. The very early universe
would be a domain where quantum physics would dominate, leading to complete uncertainty and an inability
to predict the consequence of any initial situation; we need this to evolve to a state where classical physics
leads to the properties of regularity and predictability that allow order to emerge.
• Physical conditions on planets must be in a quasi-equilibrium state for long enough to allow the delicate
balances that enable our existence, through the very slow process of evolution, to be fulfilled.
Thus the existence of suitable local systems to be a habitat for life depends critically on the large-scale properties
of very distant matter. These provides a stable local environment within which life can develop.
Fine tuning overall
Thus there are many ways that conditions in a universe could prevent life occurring. Life will occur only if: there
exist heavy elements; there is sufficient time for evolution of advanced life forms to take place; there are regions in
the universe that are neither too hot nor too cold; there are precisely restricted values of the fundamental constants
that control chemistry and local physics; and so on. These conditions will not be true in a generic universe. In
Thesis G1: Life is possible because both the laws of physics and the boundary conditions for
the universe have a very special nature. Only particular laws of physics, and particular initial conditions
in the Universe, allow the existence of intelligent life of the kind we know. No evolutionary process whatever is
possible for any kind of life if these laws and conditions do not have this restricted form.
Why is this so? One should note that we can only meaningfully refer here to ‘life as we know it’. One of the
recurring issues is whether there could be some other quite different basis for life. You can if you wish speculate
that life might exist in some immaterial form, or based only on light elements, or existent deep in empty space
without the need for stars or planets to provide a viable habitat. The anthropic literature is based on assuming
this is not viable, but we cannot prove anything in this regard. We have no idea of any basis by which life might
come into existence other than the broad principles we see in the life around us. The basic principles of life as
we understand it require a great degree of complex organization enabling it to fulfil a complex variety of functions
that can only, as far as we know, be based in material existence with information storage, energy usage, sensing
of the external world, etc., which requires at a minimum heavy elements (carbon, nitrogen, oxygen, phosphorus
for example), a long-term energy source (such as the flow of energy from the sun), and a stable environment (such
as the surface of a planet). When we abandon this basis for understanding — saying ‘yes but some other form of
life might exist’ without providing any proposal for its possible structure — one enters the unprofitable realm of
speculation. It does not seem to provide any useful way forward.
29 Thus
the Weyl tensor Cabcd must be suitably small everywhere, presumably implying an almost-RW geometry, cf. [210].
The Weak Anthropic Principle. There are two purely scientific approaches to the Anthropic issue.30 The first
is the Weak Anthropic Principle (WAP), based on the comment: it is not surprising the observed Universe admits
the existence of life, for the Universe cannot be observed unless there are observers in it [7, 2]. This seemingly
empty statement gains content when we turn it round and ask, at what times and places in the Universe can life
exist, and what are the inter-connections that are critical for its existence? It could not for example exist too
early in the present expansion phase, for the night sky would then have been too hot. Furthermore one can deduce
various necessary relations between fundamental quantities in order that the observers should exist (e.g. those
mentioned above), so that if for example the fundamental constants vary with time or place in the Universe, life
will only be possible in restricted regions where they take appropriate Anthropic values.
Hence this view basically interprets the Anthropic principle as a selection principle: the necessary conditions
for observers to exist restricts the times and places from which the Universe can be observed. Because it is quite
possible that conditions would not be right for life to exist anywhere in an arbitrarily selected universe, it is
also usually conjoined with the idea of the existence of a multiverse, as discussed below, see Sec.9.2. This is an
interesting and often illuminating viewpoint. For example, neither the Chaotic Inflationary Universe idea (Sec.2.6)
nor any other multiverse proposal works unless we add such an Anthropic component into their interpretation to
explain why we observe the Universe from a viewpoint where life exists. It is now used by some physicists to explain
the low value of the cosmological constant (which quantum field theory predicts should have a very much larger
value than observed, see Sec.9.2.5), and occurs in the context of the possibility landscape of string theory [212].
The Strong Anthropic Principle. By contrast, the Strong Anthropic Principle (SAP) [7, 2] claims that it is
necessary that intelligent life exist in the Universe; the presence of life is required in order that a universe model
make sense. This is clearly a very controversial claim, for it is hard to provide scientific reasons to support this
view. One can suggest that the most solid justification attempted is through the claim that the existence of an
observer is necessary in order that quantum theory can make sense. However, this justification is based on one of
a number of different interpretations of quantum theory; the nature of these quantum foundations is controversial,
and not resolved [119, 35, 131].
Furthermore if we were to suppose this justification correct, then the next step is to ask: Why does the
Universe need quantum mechanics anyway? The argument would be complete only if we could prove that quantum
mechanics was absolutely necessary for every self-consistent Universe; but that line of reasoning cannot be completed at present, not least because quantum mechanics itself is not a fully self-consistent theory. Apart from the
conceptual problems at its foundation due to the unresolved measurement issue [119], it suffers from divergences
that so far have proved irremediable in the sense that we can work our way round them to calculate what we need,
but cannot remove them. The SAP proposal has no accepted physical foundation, and also raises problematic
philosophical issues [40]. I will not pursue it further here.
The relation to fundamental physical theories
Many physicists go further, rejecting any Anthropic form of reasoning. They regard it as a cop-out resorted to when
physical theories fail to give the needed answers, and seek to obtain a full answer from physics alone [195, 212]. One
possibility is that there is a fundamental theory of everything that determines the nature of physics completely,
with no arbitrary parameters left, and this still to be discovered theory just happens to be of such a nature as to
admit life.
However in this case the Anthropic issue returns with a vengeance: How could it be that such a theory, based
for example on variational principles and the specific invariance groups of particle physics, could just happen to
lead to biophilic parameter values? There is no clear way to answer such a question. Uniqueness of fundamental
physics resolves the parameter freedom only at the expense of creating an even deeper mystery, with no way of
resolution apparent. In effect, the nature of the unified fundamental force would be pre-ordained to allow, or even
encourage, the existence of life; but there would be no apparent reason why this should be so.
A second possibility is that physics allows many effective theories with varying parameters — some form
of multiverse, as for example may be implied by string theory [211, 212]. If these varying options are all equally
real, life can occur because in some cases the parameters will lie in the restricted biophilic regime. Thus from
this viewpoint the Anthropic idea is intimately linked with the existence of multiverses, which provide a legitimate
domain for their application. We will turn to an examination of multiverses in the next section, but before doing
so we will consider the range of metaphysical options for resolving the anthropic question.
30 I omit the so-called Final Anthropic Principle (FAP for short), which maintains that intelligent life must necessarily evolve and
then remain in existence until the end of the universe, for I do not believe it merits serious discussion as a scientific proposal; indeed
it led to a famous book review referring to the Completely Ridiculous Anthropic Principle (CRAP for short) [87].
The metaphysical options
To make progress on the Anthropic issue, we have to seriously consider the nature of ultimate causation: What
is the fundamental cause for the phenomena we see? If we pursue the chain of physical cause and effect to its
conclusion, we are still left with the question: Why did this occur, and not something else? Whatever the reason
is, it is the ultimate cause we are seeking. Note that we are here leaving the terrain of science itself, and starting to
probe the domain of metaphysics — the foundations of science and indeed of existence. As noted above, one can
simply decide not to pursue such issues. If we do continue to question, there appear to be basically six approaches
to the issue of ultimate causation: namely Random Chance, Necessity, High Probability, Universality, Cosmological
Natural Selection, and Design. We briefly consider these in turn.
Option 1: Random Chance, signifying nothing. The initial conditions in the Universe just happened, and
led to things being the way they are now, by pure chance. Probability does not apply. There is no further level of
explanation that applies; searching for ‘ultimate causes’ has no meaning.
This is certainly logically possible, but not satisfying as an explanation, as we obtain no unification of ideas
or predictive power from this approach. Nevertheless some implicitly or explicitly hold this view.
Option 2: Necessity. Things have to be the way they are; there is no other option. The features we see
and the laws underlying them are demanded by the unity of the Universe: coherence and consistency require that
things must be the way they are; the apparent alternatives are illusory. Only one kind of physics is self-consistent:
all logically possible universes must obey the same physics.
To really prove this would be a very powerful argument, potentially leading to a self-consistent and complete
scientific view. But we can imagine alternative universes! — why are they excluded? Furthermore we run here into
the problem that we have not succeeded in devising a fully self-consistent view of physics: neither the foundations
of quantum physics nor of mathematics are on a really solid consistent basis. Until these issues are resolved, this
line cannot be pursued to a successful conclusion.
Option 3: High probability. Although the structure of the Universe appears very improbable, for physical
reasons it is in fact highly probable.
These arguments are only partially successful, even in their own terms. They run into problems if we
consider the full set of possibilities: discussions proposing this kind of view actually implicitly or explicitly restrict
the considered possibilities a priori, for otherwise it is not very likely the Universe will be as we see it. Besides,
we do not have a proper measure to apply to the set of initial conditions, enabling us to assess these probabilities.
Furthermore, as discussed above, see Sec.3, application of probability arguments to the Universe itself is dubious,
because the Universe is unique. Despite these problems, this approach has considerable support in the scientific
community, for example it underlies the chaotic inflationary proposal (Sec.2.6). It attains its greatest power in the
context of the assumption of universality:
Option 4: Universality. This is the stand that “All that is possible, happens”: an ensemble of universes
or of disjoint expanding universe domains is realized in reality, in which all possibilities occur [175, 176, 216]. In
its full version, the anthropic principle is realized in both its strong form (if all that is possible happens, then life
must happen) and its weak form (life will only occur in some of the possibilities that are realized; these are picked
out from the others by the WAP, viewed as a selection principle). There are four ways this has been pursued.
1: Spatial variation. The variety of expanding universe domains is realised in space through random
initial conditions, as in chaotic inflation (Sec.2.6). While this provides a legitimate framework for application of
probability, from the viewpoint of ultimate explanation it does not really succeed, for there is still then one unique
Universe whose (random) initial conditions need explanation. Initial conditions might be globally statistically
homogeneous, but also there could be global gradients in some physical quantities so that the Universe is not
statistically homogeneous; and these conditions might be restricted to some domain that does not allow life. It is
a partial implementation of the ensemble idea; insofar as it works, it is really a variant of the “high probability”
idea mentioned above. If it was the more or less unique outcome of proven physics, then that would provide a
good justification; but the physics underlying such proposals is not even uniquely defined, much less tested. Simply
claiming a particular scalar field with some specific stated potential exists does not prove that it exists!
2: Time variation. The variety of expanding universe domains could be realised across time, in a universe
that has many expansion phases (a Phoenix universe), whether this occurs globally or locally. Much the same
comments apply as in the previous case.
3: Quantum Mechanical. It could occur through the existence of the Everett-Wheeler “many worlds” of
quantum cosmology, where all possibilities occur through quantum branching [32]. This is one of the few genuine
alternatives proposed to the Copenhagen interpretation of quantum mechanics, which leads to the necessity of an
observer, and so potentially to the Strong Anthropic interpretation considered above (see Sec.9.1). The manyworlds proposal is controversial: it occurs in a variety of competing formulations [119], none of which has attained
universal acceptance. The proposal does not provide a causal explanation for the particular events that actually
occur: if we hold to it, we then have to still explain the properties of the particular history we observe (for example,
why does our macroscopic universe have high symmetries when almost all the branchings will not?). And above
all it is apparently untestable: there is no way to experimentally prove the existence of all those other branching
universes, precisely because the theory gives the same observable predictions as the standard theory.
4: Completely disconnected. They could occur as completely disconnected universes: there really is an
ensemble of universes in which all possibilities occur, without any connection with each other [138, 176, 216].
A problem that arises then is, What determines what is possible? For example, what about the laws of logic
themselves? Are they inviolable in considering all possibilities? We cannot answer, for we have no access to this
multitude of postulated worlds. We explore this further below (Sec.9.2).
In all these cases, major problems arise in relating this view to testability and so we have to query the
meaningfulness of the proposals as scientific explanations. They all contradict Ockham’s razor: we “solve” one
issue at the expense of envisaging an enormously more complex existential reality. Furthermore, they do not solve
the ultimate question: Why does this ensemble of universes exist? One might suggest that ultimate explanation
of such a reality is even more problematic than in the case of single universe. Nevertheless this approach has an
internal logic of its own which some find compelling. We consider this approach further below, see Sec.9.2.
Option 5: Cosmological Natural Selection. If a process of re-expansion after collapse to a black hole were
properly established, it opens the way to the concept not merely of evolution of the Universe in the sense that its
structure and contents develop in time, but in the sense that the Darwinian selection of expanding universe regions
could take place, as proposed by Smolin [205]. The idea is that there could be collapse to black holes followed by
re-expansion, but with an alteration of the constants of physics through each transition, so that each time there
is an expansion phase, the action of physics is a bit different. The crucial point then is that some values of the
constants will lead to production of more black holes, while some will result in less. This allows for evolutionary
selection favouring the expanding universe regions that produce more black holes (because of the favourable values
of physical constants operative in those regions), for they will have more “daughter” expanding universe regions.
Thus one can envisage natural selection favouring those physical constants that produce the maximum number of
black holes.
The problem here is twofold. First, the supposed ‘bounce’ mechanism has never been fully explicated.
Second, it is not clear — assuming this proposed process can be explicated in detail - that the physics which
maximizes black hole production is necessarily also the physics that favours the existence of life. If this argument
could be made water-tight, this would become probably the most powerful of the multiverse proposals.
Option 6: Purpose or Design. The symmetries and delicate balances we observe require an extraordinary
coherence of conditions and cooperation of causes and effects, suggesting that in some sense they have been
purposefully designed. That is, they give evidence of intention, both in the setting of the laws of physics and in the
choice of boundary conditions for the Universe. This is the sort of view that underlies Judaeo-Christian theology.
Unlike all the others, it introduces an element of meaning, of signifying something. In all the other options, life
exists by accident; as a chance by-product of processes blindly at work.
The prime disadvantage of this view, from the scientific viewpoint, is its lack of testable scientific consequences (“Because God exists, I predict that the density of matter in the Universe should be x and the fine structure
constant should be y”). This is one of the reasons scientists generally try to avoid this approach. There will be
some who will reject this possibility out of hand, as meaningless or as unworthy of consideration. However it is
certainly logically possible. The modern version, consistent with all the scientific discussion preceding, would see
some kind of purpose underlying the existence and specific nature of the laws of physics and the boundary conditions for the Universe, in such a way that life (and eventually humanity) would then come into existence through
the operation of those laws, then leading to the development of specific classes of animals through the process of
evolution as evidenced in the historical record. Given an acceptance of evolutionary development, it is precisely in
the choice and implementation of particular physical laws and initial conditions, allowing such development, that
the profound creative activity takes place; and this is where one might conceive of design taking place.31
However from the viewpoint of the physical sciences per se, there is no reason to accept this argument.
Indeed from this viewpoint there is really no difference between design and chance, for they have not been shown
to lead to different physical predictions.
Metaphysical Uncertainty
In considering ultimate causation underlying the anthropic question, in the end we are faced with a choice between
one of the options above. As pointed out already by Kant and Hume, although we may be able to argue strongly
for one or other of them, we cannot prove any of the options are correct [116].
31 This is not the same as the view proposed by the ‘Intelligent Design’ movement. It does not propose that God tweaks the outcome
of evolutionary processes.
Thesis G2: Metaphysical uncertainty remains about ultimate causation in cosmology. We
cannot attain certainty on the underlying metaphysical cosmological issues through either science or philosophy.
If we look at the anthropic question from a purely scientific basis, we end up without any resolution, basically
because science attains reasonable certainty by limiting its considerations to restricted aspects of reality; even if it
occasionally strays into the area of ultimate causation, it is not designed to deal with it. By itself, it cannot make a
choice between these options; there is no relevant experiment or set of observations that can conclusively solve the
issue. Thus a broader viewpoint is required to make progress, taking into account both the scientific and broader
considerations. The issue is of a philosophical rather than scientific nature. One important issue that then arises
is what kind of data is relevant to these philosophical choices, in addition to that which can be characterized as
purely scientific data (Sec.9.3.4).
Issue H: The possible existence of multiverses
If there is a large enough ensemble of numerous universes with varying properties, it may be claimed that it becomes
virtually certain that some of them will just happen to get things right, so that life can exist; and this can help
explain the fine-tuned nature of many parameters whose values are otherwise unconstrained by physics [175, 176].
As discussed in the previous section, there are a number of ways in which, theoretically, multiverses could be
realized [138, 216]. They provide a way of applying probability to the universe [193, 17] (because they deny the
uniqueness of the universe). However, there are number of problems with this concept. Besides, this proposal is
observationally and experimentally untestable; thus its scientific status is debatable.
In justifying multiverses, it is often stated that ‘all that can occur, occurs’ (or similarly). However that statement
does not adequately specify a multiverse. To define a multiverse properly requires two steps [69]. First, one
needs to specify what is conceived of in the multiverse, by defining a possibility space: a space M of all possible
universes, each of which can be described in terms of a set of states s in a state space S. Each universe m in
M will be characterized by a set of distinguishing parameters p, which are coordinates on S. Choices are needed
here. In geometrical terms, will it include only Robertson–Walker models, or more general ones (e.g. Bianchi
models, or models without symmetries)? In gravitational terms, will it include only General Relativity, or also
brane theories, models with varying G, loop quantum gravity models, string theory models with their associated
possibility ‘landscapes’, and models based on the wave function of the universe concept? Will it allow only
standard physics but with varying constants, or a much wider spectrum of physical possibilities, e.g. universes
without quantum theory, some with five fundamental forces instead of four, and others with Newtonian gravity?
Defining the possibility space means making some kind of assumptions about physics and geometry that will then
apply across the whole family of models considered possible in the multiverse, and excluding all other possibilities.
Second, one needs to specify which of the possible universes are physically realized in the multiverse, and
how many times each one occurs. A multiverse must be a physically realized multiverse and not a hypothetical or
conceptual one if it is to have genuine explanatory power. Thus one needs a distribution function f (m) specifying
how many times each type of possible universe m in M is realised. The function f (m) expresses the contingency
in any actualization. Things could have been different! Thus, f (m) describes a specific ensemble of universes or
multiverse envisaged as being realised out of the set of possibilities. For example, f (m) might be non-zero for all
possible values of all the parameters p (‘all that can happen, happens’); but it could be that f describes a multiverse
where there are 10100 identical copies of one particular universe (the realization process finds a particularly successful
recipe, and then endlessly replicates it).
Additionally we need a measure dπ that enables this function to determine numbers and probabilities of
various properties in the multiverse: the number of universes corresponding to a set of parameter increments will
be dN given by
dN = f (m)dπ
for continuous parameters; for discrete parameters, we add in the contribution from all allowed parameter values.
The total number of universes N in the ensemble will be given by
f (m)dπ
(which will often diverge), where the integral ranges over all allowed values of the member parameters and we take
it to include all relevant discrete summations. The expectation value P of a quantity p(m) defined on the set of
universes will be given by
p(m)f (m)dπ.
P =
These three elements (the possibility space, the measure, and the distribution function) must all be clearly
defined in order to give a proper specification of a multiverse [69]. This is almost never done.
Non-uniqueness: Possibilities
There is non-uniqueness at both steps. Stating “all that is possible, happens” does not resolve what is possible.
The concept of multiverses is not well defined until the space of possible universes has been fully characterized; it is
quite unclear how to do this uniquely. The issue of what is to be regarded as an ensemble of ‘all possible’ universes
can be manipulated to produce any result you want, by redefining what is meant by this phrase — standard physics
and logic have no necessary sway over them: what I envisage as ‘possible’ in such an ensemble may be denied by
you. What super-ordinate principles are in operation to control the possibilities in the multiverse, and why? A key
point here is that our understandings of the possibilities are always of necessity arrived at by extrapolation from
what we know, and my imagination may be more fertile than yours, and neither need correspond to what really
exists out there — if indeed there is anything there at all. Do we include only
• Weak variation: e.g. only the values of the constants of physics are allowed to vary? This is an interesting
exercise but is certainly not an implementation of the idea ‘all that can happen, happens’. It is an extremely
constrained set of variations.
• Moderate variation: different symmetry groups, or numbers of dimensions, etc. We might for example
consider the possibility landscapes of string theory [86] as realistic indications of what may rule multiverses
[211, 212]. But that is very far indeed from ‘all that is possible’, for that should certainly include spacetimes
not ruled by string theory.
• Strong variation: different numbers and kinds of forces, universes without quantum theory or in which
relativity is untrue (e.g. there is an aether), some in which string theory is a good theory for quantum
gravity and others where it is not, some with quite different bases for the laws of physics (e.g. no variational
• Extreme variation: universes where physics is not well described by mathematics; with different logic; universes ruled by local deities; allowing magic as in the Harry Potter series of books; with no laws of physics
at all? Without even mathematics or logic?
Which is claimed to be the properties of the multiverse, and why? We can express our dilemma here through the
paradoxical question: Are the laws of logic necessary in all possible universes?
Non-uniqueness: existence and causation
A specific multiverse is defined by specifying the distribution function f (m) of actually realized universes. It is
unclear what mechanism can underlie such a distribution, and any proposal for such a mechanism is completely
untestable. We need some indication as to what determines existence within the possibilities defined by the supposed
possibility space: What decides how many times each one happens? Unless we understood the supposed underlying
mechanisms we can give no serious answer; and there is no prospect whatever of testing any proposed mechanism.
The mechanisms supposed to underlie whatever regularities there are in the multiverse must pre-exist the existence
of not merely this universe but also every other one. If one assumes a universe that is connected in the large but
is locally separated into causally disconnected domains with different physical properties(as in chaotic inflation),
one attains a plausible picture of a creation mechanism that can underlie an effective multiverse — but at the
expense of supposing the validity of untested and perhaps untestable physics. Because of this one does not obtain
a specification of a unique multiverse: the physics could be different than what we assumed.
Explanatory power
What explanatory power do we get in return for these problems? It has been suggested they explain the parameters
of physics and of cosmology and in particular the very problematic observed value of the cosmological constant
[235, 212]. The argument goes as follows: assume a multiverse exists; observers can only exist in one of the highly
improbable biophilic outliers where the value of the cosmological constant is very small [105]. A similar argument
has been proposed for neutrino masses [218]. If the multiverse has many varied locations with differing properties,
that may indeed help us understand the Anthropic issue: some regions will allow life to exist, others will not
[7, 135]. This does provide a useful modicum of explanatory power. However it is far from conclusive. Firstly, it
is unclear why the multiverse should have the restricted kinds of variations of the cosmological constant assumed
in the various analyses mentioned. If we assume ‘all that can happen, happens’ the variations will not be of that
restricted kind; those analyses will not apply.
Secondly, ultimate issues remain: Why does this unique larger whole have the properties it does? Why
this multiverse rather than any other one? Why is it a multiverse that allows life to exist? Many multiverses will
not allow any life at all. To solve this, we can propose an ensemble of ensembles of universes, with even greater
explanatory power and even less prospect of observational verification; and so on. The prospect of an infinite
regress looms. Indeed if we declare (as suggested at the start of this article) that ‘the Universe’ is the total of all
that physically exists, then when an ensemble of expanding universe domains exists, whether causally connected or
not, that ensemble itself should be called ‘the Universe’, for it is then the totality of physically existing entities. All
the foundational problems for a single existing universe domain recur for the multiverse — because when properly
considered, it is indeed the Universe!
If an ensemble exists with members not connected in any physical way to the observable universe, then we cannot
interact with them in any way nor observe them, so we can say anything we like about them without fear of
disproof.32 Thus any statements we make about them can have no solid scientific or explanatory status; they are
totally vulnerable to anyone else who claims an ensemble with different properties (for example claiming different
kinds of underlying logics are possible in their ensemble, or claiming many physically effective gods and devils in
many universes in their ensemble).
Thesis H1: Multiverse proposals are unprovable by observation or experiment, but some selfconsistency tests are possible. Direct observations cannot prove or disprove that a multiverse exists, for the
necessary causal relations allowing observation or testing of their existence are absent. Their existence cannot be
predicted from known physics, because the supposed causal or pre-causal processes are either unproven or indeed
untestable. However some self-consistency conditions for specific multiverse models can be tested.
Any proposed physics underlying a multiverse proposal, such as Coleman-de Luccia tunneling [25], will be an
extrapolation of known physics; but the validity of that major extrapolation to cosmology is untestable.
Attempts have been made to justify the existence of multiverses as testable firstly via Rees’ ‘slippery slope’
argument [176]. This runs as follows: we can reasonably assume galaxies that we cannot see exist outside the visual
horizon (Sec.8.3); why not extend this argument by small steps to show totally disconnected universes exist? The
problem is that this assumes a continuity of existence that does not hold good. The domain outside our horizon is
assumed to exist with similar properties to those inside because they are a continuous extension of it and have a
largely common causal origin; their nature can be inferred from what we can see. Disconnected multiverse domains
are assumed to have quite different properties, and their nature cannot be inferred from what we can see as there
is no continuity or causal connection.
Secondly, several authors (Leslie [135], Weinberg [235], and Rees [176] for example) have used arguments
based on the idea that the universe is no more special than it has to be; a form of “speciality argument.” According
to Rees, if our universe turns out to be even more specially tuned than our presence requires, the existence of a
multiverse to explain such “over-tuning” would be refuted; but the actual universe is not more special than this,
so the multiverse is not refuted.
In more detail: naive quantum physics predicts Λ to be very large. But our presence in the universe requires
it to be small enough that galaxies and stars can form, so Λ must obviously be below that galaxy-forming threshold.
If our universe belongs to an ensemble in which Λ was equally likely to take any value in the biophilic region (the
uniform probability assumption),33 then we would not expect it to be too far below this threshold. This is because,
if it’s too far below the threshold, the probability of randomly choosing that universe in the ensemble becomes very
small — there are very few universes with such small values of Λ in the biophilic subset of the ensemble. That is, it
would be more likely that any bio-friendly universe in the ensemble would have a value of Λ closer to the threshold
value. Present data on this value indicates that it is not too far below the threshold. Thus, our universe is not
markedly more special that it needs to be as far as Λ is concerned, and so explaining its fine-tuning by existence
of a multiverse is legitimate.
Is this argument compelling? It is a reasonable test of consistency for a multiverse that is known to exist,
so that probability considerations apply; but they do not apply if there is no multiverse (Sec.3). Additionally,
probability considerations cannot ever be conclusive. Indeed,
Thesis H2: Probability-based arguments cannot demonstrate the existence of multiverses.
Probability arguments cannot be used to prove the existence of a multiverse, for they are only applicable if a
multiverse exists. Furthermore probability arguments can never prove anything for certain, as it is not possible
to violate any probability predictions, and this is a fortiori so when there is only one case to consider, so that no
statistical observations are possible.
32 But
there are counter arguments by Leibniz [239] and Lewis ([138] section 2.4, pp. 108-115).
probability distribution for Λ will plausibly peak far away from the biophilic region, tailing down to a low value that will be
approximately constant in that narrow region, cf.[105].
33 The
All one can say on the basis of probability arguments is that some specific state is very improbable. But this does
not prove it is impossible, indeed if it is stated to have a low probability, that is precisely a statement that it is
possible. Thus such arguments can at best only give plausibility indications even when they are applicable. The
assumption that probability arguments can be conclusive is equivalent to the claim that the universe is generic
rather than special; but whether this is so or not is precisely the issue under debate (see Thesis D3). The argument
is useful as a plausibility argument for a multiverse, but is not proof of its existence.
Finally, it has been proposed that the existence of multiverses is an inevitable consequence of the universe
having infinite space sections [216, 198], because that leads to infinite spatial repetition of conditions (cf. [65]). But
this supposed spatial infinity is an untested philosophical assumption, which certainly cannot be observationally
proven to be correct. Apart from the existence of horizons preventing confirmation of this supposition, even if the
entire universe were observable, proving it correct would still not be possible because by definition counting an
infinite number of objects takes an infinite amount of time. This is an untestable philosophical argument, not an
empirically testable one; furthermore, it can be argued to be implausible (Sec.9.3.2). Indeed current data suggest
it is not the case; this is the one good consistency test one can use for some multiverse proposals (Sec.9.2.7).
Explanation vs Testability
The argument that this infinite ensemble actually exists can be claimed to have a certain explanatory economy,
although others would claim that Occam’s razor has been completely abandoned in favour of a profligate excess
of existential multiplicity, extravagantly hypothesized in order to explain the one universe that we do know exists.
Certainly the price is a lack of testability through either observations or experiment — which is usually taken to be
an essential element of any serious scientific theory.34 It is not uniquely definable nor determinable, and there is a
complete loss of verifiability. There is no way to determine the properties of any other universe in the multiverse if
they do indeed exist, for they are forever outside observational reach. The point is that there is not just an issue of
showing a multiverse exists. If this is a scientific proposition one needs to be able to show which specific multiverse
exists; but there is no observational way to do this. Indeed if you can’t show which particular one exists, it is
doubtful you have shown any one exists.
What does a claim for such existence mean in this context? Gardner puts it this way: “There is not the
slightest shred of reliable evidence that there is any universe other than the one we are in. No multiverse theory
has so far provided a prediction that can be tested. As far as we can tell, universes are not even as plentiful as
even two blackberries” [88].35
Thesis H3: Multiverses are a philosophical rather than scientific proposal. The idea of a multiverse provides a possible route for the explanation of fine tuning. But it is not uniquely defined, is not scientifically
testable apart from some possible consistency tests, and in the end simply postpones the ultimate metaphysical
The definitive consistency tests on some multiverse proposals (Sec.9.2.7) are necessary conditions for those specific
multiverse proposals, but are hardly by themselves indications that the multiverse proposal is true. The drive to
believe this is the case comes from theoretical and philosophical considerations (see e.g. [212]) rather than from
data. The claim an ensemble physically exists36 is problematic as a proposal for scientific explanation, if science
is taken to involve testability. Indeed, adopting these explanations is a triumph of theory over testability [88], but
the theories being assumed are not testable. It is therefore a metaphysical choice made for philosophical reasons.
That does not mean it is unreasonable (it can be supported by quite persuasive plausibility arguments); but its
lack of scientific status should be made clear.
Observations and disproof
Despite the gloomy prognosis given above, there are some specific cases where the existence of a chaotic inflation
(multi-domain) type scenario (Sec.2.6) can be disproved. These are firstly when we live in a ‘small universe’ where
we have already seen right round the universe (Sec.4.3.1), for then the universe closes up on itself in a single
FL-like domain, so that no further such domains can exist that are causally connected to us in a single connected
spacetime. This ‘small universe’ situation is observationally testable (Sec.4.3.1); its confirmation would disprove
the usual chaotic inflationary scenario, but not a truly ‘disconnected’ multiverse proposal, for that cannot be shown
to be false by any observation. Neither can it be shown to be true. Secondly, many versions of chaotic inflation, for
example those involving Coleman-de Luccia tunneling [25] from a de Sitter spacetime, demand k = −1 ⇔ Ω0 < 1
[86, 212]. This requirement is currently marginally disproved by the 2 − σ bounds on Ω0 when WMAP observations
34 In [209], the framework and conditions under which the multiverse hypothesis would be testable within a retroductive framework,
given the rigorous conditions formulated in that paper; are indicated; these conditions are not fulfilled.
35 This contrasts strongly, for example, with Deutsch’s and Lewis’s defence of the concept [32, 138]. Lewis defends the thesis of
“modal realism”: that the world we are part of is but one of a plurality of worlds.
36 As opposed to consideration of an explicitly hypothetical such ensemble, which can indeed be useful, see Sec.7.2).
are combined with the best other available data (Sec.2.3.7). The best current data is marginally consistent with
k = −1, but the value indicated most strongly by that data is k = +1, indicating finite closed space sections rather
than an infinite multiverse such as that advocated by Susskind et al [86, 212].
Physical or biological paradigms — Adaptive Evolution?
Given that the multiverse idea must in the end be justified philosophically rather than by scientific testing, is
there a philosophically preferable version of the idea? One can suggest there is: greater explanatory power is
potentially available by introducing the major constructive principle of biology into cosmology, namely adaptive
evolution, which is the most powerful process known that can produce ordered structure where none pre-existed.
This is realized in principle in Lee Smolin’s idea (Sec.9.1.6) of Darwinian adaptation when collapse to black holes is
followed by re-expansion, but with an alteration of the constants of physics each time, so as to allow for evolutionary
selection towards those regions that produce the maximum number of black holes. The idea needs development,
but is very intriguing:
Thesis H4: The underlying physics paradigm of cosmology could be extended to include
biological insights. The dominant paradigm in cosmology is that of theoretical physics. It may be that it will
attain deeper explanatory power by embracing biological insights, and specifically that of Darwinian evolution. The
Smolin proposal for evolution of populations of expanding universe domains [205] is an example of this kind of
The result is different in important ways from standard cosmological theory precisely because it embodies in
one theory three of the major ideas of the last century, namely (i) Darwinian evolution of populations through
competitive selection, (ii) the evolution of the universe in the sense of major changes in its structure associated
with its expansion, and (iii) quantum theory, underlying the only partly explicated mechanism supposed to cause
re-expansion out of collapse into a black hole. There is a great contrast with the theoretical physics paradigm of
dynamics governed simply by variational principles shaped by symmetry considerations. It seems worth pursuing
as a very different route to the understanding of the creation of structure.37
Issue I: Natures of Existence
Underlying all this is the issue of natures of existence, which has a number of aspects, relating from the purely
physical to more metaphysical issues.
Physical existence: kinds of matter
Unsolved key issues for physical cosmology relate to what kind of matter and/or fields exist. While we understand
matter in the solar system quite well, at present we do not understand most of what exists in the universe at large:
Thesis I1: We do not understand the dominant dynamical matter components of the universe
at early or late times. A key goal for physical cosmology is determining the nature of the inflaton, of dark matter,
and of dark energy. Until this is done, the causal understanding of cosmology is incomplete, and in particular the
far future fate of the universe is unknown.
This is the core activity of much work in cosmology at present. Until they are all explicated, cosmology is not
properly linked to physics, and the nature of the matter that dominates the dynamics of the universe is unknown.
Its explication is surely one of the key concerns of cosmology [39]. A key requirement is that even if we cannot
experimentally verify the proposed nature of the matter, at least it should be physically plausible. This appears
not to be the case for some current proposals, e.g. so-called ‘phantom matter’ which has negative kinetic energy
The far future fate of the universe depends crucially on the effective equation of state for dark matter
(‘quintessence’). But the problem is that even if we can determine these properties at the present time (for one
particular range of parameter values), this does not necessarily guarantee what they will be in the far future (for
a quite different range of parameter values that are probably outside the range of possible experimental test).
Furthermore adjusting a ‘dark energy’ model to fit the supernova data does not determine the underlying physics.
One can fit any monotonic evolution S(t) with a suitable choice of the equation of state function p(µ). Specifically,
for any S(t) and any k we define µ(t) and p(t) by
Ṡ 2 (t)
Ṡ 2 (t)
, κp(t) =
κµ(t) = 3
S 2 (t) S 2 (t)
S 2 (t) S 2 (t)
37 cf.
Chapter 13 of Susskind [212].
then (9), (7) will be exactly satisfied, and we have ‘solved’ the field equations for this arbitrarily chosen monotonic
evolution S(t). If we can observationally determine the form of S(t), for example from (m, z)–curves associated with
supernovae data, this is essentially how we can then determine that some kind of ‘dark energy’ or ‘quintessence’ is
required to give that evolution, and we can find the equation of state implied by eliminating t between these two
equations. This is, however, not a physical explanation until we have either in some independent experimental test
demonstrated that matter of this form exists, or have theoretically shown why this matter or field has the form
it does in some more fundamental terms than simply a phenomenological fit. If we assume the matter is a scalar
field, the kinetic energy term φ̇2 implied by (32), (41) may be negative — which is the case for so-called ‘shadow
matter’ models proposed recently by some worker. If normal physics criteria are applied, this is a proof that this
kind of matter is unphysical, rather than an identification of the nature of the dark energy.
Existence of Infinities
The nature of existence is significantly different if there is a finite amount of matter or objects in the universe, as
opposed to there being an infinite quantity in existence. Some proposals claim there may be an infinite number of
universes in a multiverse and many cosmological models have spatial sections that are infinite, implying an infinite
number of particles, stars, and galaxies. However, infinity is quite different from a very large number! Following
David Hilbert [111], one can suggest these unverifiable proposals cannot be true: the word ‘infinity’ denotes a
quantity or number that can never be attained, and so will never occur in physical reality.38 He states
“Our principal result is that the infinite is nowhere to be found in reality. It neither exists in nature nor provides
a legitimate basis for rational thought . . . The role that remains for the infinite to play is solely that of an idea .
. . which transcends all experience and which completes the concrete as a totality . . .” ([111], p. 151).
This suggests “infinity” cannot be arrived at, or realized, in a concrete physical setting; on the contrary, the concept
itself implies its inability to be realized!39
Thesis I2: The often claimed physical existence of infinities is questionable. The claimed existence
of physically realized infinities in cosmology or multiverses raises problematic issues. One can suggest they are
unphysical; in any case such claims are certainly unverifiable.
This applies in principle to both small and large scales in any single universe:
• The existence of a physically existing spacetime continuum represented by a real (number) manifold at the
micro-level contrasts with quantum gravity claims of a discrete spacetime structure at the Planck scale, which
one might suppose was a generic aspect of fully non-linear quantum gravity theories [184]. In terms of physical
reality, this promises to get rid of the uncountable infinities the real line continuum engenders in all physical
variables and fields.40 There is no experiment that can prove there is a physical continuum in time or space;
all we can do is test space-time structure on smaller and smaller scales, but we cannot approach the Planck
• Infinitely large space-sections at the macro-level raise problems as indicated by Hilbert, and leads to the
infinite duplication of life and all events [65]. We may assume space extends forever in Euclidean geometry and
in many cosmological models, but we can never prove that any realised 3-space in the real universe continues
in this way — it is an untestable concept, and the real spatial geometry of the universe is almost certainly
not Euclidean. Thus Euclidean space is an abstraction that is probably not physically real. The infinities
supposed in chaotic inflationary models derive from the presumption of pre-existing infinite Euclidean space
sections, and there is no reason why those should necessarily exist. In the physical universe spatial infinities
can be avoided by compact spatial sections, resulting either from positive spatial curvature, or from a choice
of compact topologies in universes that have zero or negative spatial curvature. Machian considerations to
do with the boundary conditions for physics suggest this is highly preferable [236]; and if one invokes string
theory as a fundamental basis for physics, then ‘dimensional democracy’ suggests the three large spatial
dimensions should also be compact, since the small (‘compactified’) dimensions are all taken to be so. The
best current data from CBR and other observations (Sec.2.3.7) indeed suggest k = +1, implying closed space
sections for the best-fit FL model.
• The existence of an eternal universe implies that an infinite time actually exists, which has its own problems:
if an event happens at any time t0 , one needs an explanation as to why it did not occur before that time
(as there was an infinite previous time available for it to occur); and Poincaré eternal return (mentioned
in Sec.6.1) will be possible if the universe is truly cyclic. In any case it is not possible to prove that the
38 An intriguing further issue is the dual question: Does the quantity zero occur in physical reality? This is related to the idea of
physical existence of nothingness, as contrasted with a vacuum [196]. A vacuum is not nothing! (cf. [212]).
39 For a contrasting view, see Bernadete [10].
40 To avoid infinities entirely would require that nothing whatever is a continuum in physical reality (since any continuum interval
contains an infinite number of points). Doing without that, conceptually, would mean a complete rewrite of many things. Considering
how to do so in a way compatible with observation is in my view a worthwhile project.
universe as a whole, or even the part of the universe in which we live, is past infinite; observations cannot do
so, and the physics required to guarantee this would happen (if initial conditions were right) is untestable.
Even attempting to prove it is future infinite is problematic (we cannot for example guarantee the properties
of the vacuum into the infinite future — it might decay into a state corresponding to a negative effective
cosmological constant).
• It applies to the possible nature of a multiverse. Specifying the geometry of a generic universe requires an
infinite amount of information because the quantities necessary to do so are fields on spacetime, in general
requiring specification at each point (or equivalently, an infinite number of Fourier coefficients): they will
almost always not be algorithmically compressible. All possible values of all these components in all possible
combinations will have to occur in a multiverse in which “all that can happen, does happen”. There are also
an infinite number of topological possibilities. This greatly aggravates all the problems regarding infinity and
the ensemble. Only in highly symmetric cases, like the FL solutions, does this data reduce to a finite number
of parameters, each of which would have to occur in all possible values (which themselves are usually taken to
span an infinite set, namely the entire real line). Many universes in the ensemble may themselves have infinite
spatial extent and contain an infinite amount of matter, with all the problems that entails. To conceive of
physical creation of an infinite set of universes (most requiring an infinite amount of information for their
prescription, and many of which will themselves be spatially infinite) is at least an order of magnitude more
difficult than specifying an existent infinitude of finitely specifiable objects.
One should note here particularly that problems arise in the multiverse context from the continuum of values
assigned by classical theories to physical quantities. Suppose for example that we identify corresponding times
in the models in an ensemble and then assume that all values of the density parameter and the cosmological
constant occur at each spatial point at that time. Because these values lie in the real number continuum, this is a
doubly uncountably infinite set of models. Assuming genuine physical existence of such an uncountable infinitude
of universes is the antithesis of Occam’s razor. But on the other hand, if the set of realised models is either finite
or countably infinite, then almost all possible models are not realised. And in any case this assumption is absurdly
unprovable. We can’t observationally demonstrate a single other universe exists [88], let alone an infinitude. The
concept of infinity is used with gay abandon in some multiverse discussions [125], without any concern either for
the philosophical problems associated with this statement [111], or for its completely unverifiable character. It is
an extravagant claim that should be treated with extreme caution.
The Nature of the Laws of Physics
Underlying all the above discussion is the basic concept of ordered behaviour of matter, characterized by laws of
physics of a mathematical nature that are the same everywhere in the universe.41 Three interlinked issues arise.
(i) What is the ontological nature of the laws of physics: descriptive, just characterizing the way things
are, or prescriptive, enforcing them to be this way? [22]. If they are descriptive, the issue arising is, Why does
all matter have the same properties wherever it exists in the universe? Why are all electrons everywhere in the
universe identical, if the laws are only descriptive? If they are prescriptive, then matter will necessarily be the same
everywhere (assuming the laws themselves are invariable); the issue arising then is, In what way do laws of physics
exist that enforce themselves on the matter in the universe? Do they for example have an existence in some kind of
Platonic space that controls the nature of matter and existence? One can avoid talking about the laws of physics
per se by instead considering the space of possibilities underlying what exists physically, rigorously constraining
the possible natures of what actually comes into existence [62]. This space is more or less uniquely related to the
underlying laws in the same way that the space of solutions of differential equations is related to the nature of the
equations. This enables one to avoid the issue of the ontology of the laws of physics, but does not solve it.
(ii) Why are the laws of physics so well explained by mathematical descriptions? If they are prescriptive, this
deep issue might be related to the suggested Platonic nature of the space of mathematical reality [169]. If they are
descriptive, then the mathematical expressions we use to encapsulate them are just a convenient description but
do not reflect their ultimate nature. Many writings in physics and cosmology seem to assume that their ultimate
existential nature is indeed mathematical — perhaps a confusion of appearance and reality (see Sec.8.3).
(iii) Do they pre-exist the universe and control its coming into being, or do they come into being with the
universe? This is where this issue relates deeply to the nature of cosmology, and is clearly related to the other two
questions raised above. Many theories of creation of the universe assume that all these laws, or at least a basic
subset, pre-exist the coming into being of the physical universe, because they are presumed to underlie the creation
process, for example the entire apparatus of quantum field theory is often taken for granted as pre-existing our
universe (Sec.6). This is of course an unprovable proposition
41 The effective laws may vary from place to place because for example the vacuum state varies [212]; but the fundamental laws that
underlie this behaviour are themselves taken to be invariant.
Thesis I3: A deep issue underlying the nature of cosmology is the nature of the laws of
physics. The nature of the possibility space for physical existence is characterized by the laws of physics. However
it is unclear if these laws are prescriptive or descriptive; whether they come into being with space-time and matter,
or pre-exist them.
‘Ultimate Reality’
Philosophers have debated for millennia whether the ultimate nature of existence is purely material, or embodies
some form of rationality (‘Logos’) and/or purpose (‘Telos’). What in the end underlies it all ? Is the ultimate
nature of the universe purely material, or does it in some way have an element of the mental? (cf. Sec.9.1.6).
That profound debate is informed by physical cosmology, but cannot be resolved by the physical sciences alone
(Sec.9.1.7). Here, I will make just two comments on this deep issue.
Firstly, even in order to understand just the material world, it can be claimed that one needs to consider
forms of existence other than the material only — for example a Platonic world of mathematics and a mental world,
both of which can be claimed to exist and be causally effective in terms of affecting the material world [62, 169].
Our understanding of local causation will be incomplete unless we take them into account.
Secondly, in examining these issues one needs to take into account data about the natures of our existence
that come from our daily lives and the broad historical experience of humanity (our experiences of ethics and
aesthetics, for example), as well as those discoveries attained by the scientific method. Many writings claim there
is no purpose in the universe: it is all just a conglomerate of particles proceeding at a fundamental level in a
purposeless and meaningless algorithmic way. But I would reply, the very fact that those writers engage in such
discourse undermines their own contention; they ignore the evidence provided by their own actions. There is
certainly meaning in the universe to this degree: the fact they take the trouble to write such contentions is proof
that they consider it meaningful to argue about such issues; and this quality of existence has emerged out of the
nature of the physical universe (Sec.7.3). Indeed the human mind is causally effective in the real physical world
precisely through many activities motivated by meanings perceived by the human mind. Any attempt to relate
physics and cosmology to ultimate issues must take such real world experience seriously [63], otherwise it will
simply be ignoring a large body of undeniable data. This data does not resolve the ultimate issues, but does
indicate dimensions of existence that indeed do occur.
The physical scale of the Universe is enormous, and the images of distant objects from which we obtain our
information are extremely faint. It is remarkable that we are able to understand the Universe as well as we do. An
intriguing feature is the way in which the philosophy of cosmology is to a considerable degree shaped by contingent
aspects of the nature of the universe — its vast scale (Sec.4), leading to the existence of visual horizons (Sec.4.3),
and the occurrence of extreme energies in the early universe (Sec.5), leading to the existence of physical horizons.
Philosophical issues arising in relation to cosmology (Sec.8) would be quite different if its physical structure were
very different. Furthermore in order that philosophical analysis can engage with cosmology in depth, the detailed
nature of the relation between observations and theory in cosmology (Sec.2) is relevant.
Are there laws of cosmology?
As we have discussed in detail, the uniqueness of the universe implies the unique nature of cosmology. We now
return to the initial issue, Are there Laws of the Universe? (Sec.3). At one level, the laws of the cosmos are simply
the local laws we know and love (e.g. Maxwell’s laws, Einstein’s field equations) applied to the whole shebang. Of
course, there is the problem of extrapolation from the local to the global. But although the extrapolation is bigger
in cosmology, it seems not to be different in kind from what we always do in science. In that sense, there are no
special laws for the evolution of the universe. But that does not determine the outcome: cosmology needs some
prescription of boundary or initial conditions as well, in order to determine the future. Is there a true “Cosmological
principle”, a law of initial conditions for the universe, that determines this outcome?
The idea of “Laws of initial conditions of the universe” seems not to be a testable idea (Sec.3). Scientifically,
one can only describe what occurred rather than relate it to generic principles, for such principles cannot be tested.
In fact any description of boundary or initial conditions for the universe seems to be just that: a description of
these conditions, rather than a testable prescription of how they must be. The ‘Cosmological Principle’ — the
universe is necessarily spatially homogeneous and isotropic (Sec.4.2.2) is of this kind: a description of the way the
initial data turned out, rather than a fundamental reason for why this should be so. Justification of this view
was based by some workers on a Copernican Principle (the assumption we do not live in a privileged place in the
universe), strengthened to become a Cosmological Principle [14, 233, 104]; but this is a philosophical assumption
— essentially, a claim that the universe must have very special initial conditions — which may or may not be
true, and does not attempt a physical explanation. This kind of argument is out of fashion at present, because we
now prefer generality to speciality and physical argumentation to geometrical prescription; but it was previously
strongly proposed (e.g. [233], pp. 407-412). The tenor of philosophical argument has changed.
Nevertheless there is one kind of Law of the Universe one might propose, following McCrea [149]: namely
an “Uncertainty principle in cosmology”, dual to the uncertainty principle in quantum theory. Uncertainty applies
on the largest scale, as we have discussed above in some detail, and also on the smallest, where it is a profound
feature of quantum theory. Its basis is very different in the two cases, on the one hand (in quantum theory) being
ontological in nature, on the other (in cosmology) being epistemological in nature.42 Nevertheless it is a key aspect
of our relation to the cosmos, so that (following McCrea) we might perhaps formalize it in order to emphasize its
centrality to the relation between cosmology and philosophy:
Thesis of Uncertainty: Ultimate uncertainty is a key aspect of cosmology. Scientific exploration
can tell us much about the universe but not about its ultimate nature, or even much about some of its major geometrical and physical characteristics. Some of this uncertainty may be resolved, but much will remain. Cosmological
theory should acknowledge this uncertainty.
What can we truly claim
Cosmology considers questions of physical origins in the uniquely existing physical universe (Sec.6) which provides
the context of our existence (Sec.7, Sec.9.1). These questions can be extended to include ultimate issues if we so
desire (Sec.8.2), but physical theory cannot resolve them (Sec.9.1.7). In the end, there are a variety of mysteries
underlying the existence and nature of the universe (Sec.9.3). The scientific study of cosmology can help illuminate
their nature, but cannot resolve them.
As well as celebrating the achievements of cosmology, one should fully take into account the limits and
problems considered in this chapter, and not claim for scientific cosmology more than it can actually achieve or
more certainty than is in fact attainable. Such claims will in the long term undermine cosmology’s legitimate status
as a project with solid scientific achievements to its name. That status can be vigorously defended as regards the
‘Standard Model’ of cosmology (Sec.2.8), provided this standard model is characterized in conservative terms so
that it is not threatened by relatively detailed shifts in theory or data that do not in fact threaten the core business
of cosmology. Further, this defence must take adequate cognisance of the difficult philosophical issues that arise if
one pushes the explanatory role of cosmological theory to its limits (Sec.6); for example one should not make too
strong scientific claims in regard to the possible existence of multiverses (Sec.9.2); philosophically based plausibility
arguments for them are fine, if identified as such. Cosmology is not well served by claims that it can achieve more
explanatory power than is in fact attainable, or by statements that its claims are verified when in fact the requisite
evidence is unavailable, and in some cases must forever remain so.
I thank Bill Stoeger, Martin Bojowald, Malcolm MacCallum, Jeremy Butterfield, Henk van Elst, and John
Earman for useful comments that have improved this article.
Abbreviations used:
HBB: Hot Big Bang
CBR: Cosmic Blackbody Radiation
CDM: Cold Dark Matter
EFE: Einstein Field Equations
FL: Friedmann-Lemaı̂tre (universe models)
LSS: Last Scattering surface
RW: Robertson-Walker (geometry)
SAP: Strong Anthropic Principle
WAP: Weak Anthropic Principle
42 Assuming that quantum uncertainty is indeed ontological rather than epistemological. One should however keep an open mind on
this: just because it is the current dogma does not necessarily mean it is true.
[1] J Allday, Quarks, Leptons and the Big Bang. (Institute of Physics, 2002).
[2] Y V Balashov, “Resource Letter Ap-1: The Anthropic Principle”. Amer Journ Phys 54:1069-1076 (1991).
[3] J Barbour and H Pfister, Mach’s Principle: From Newton’s Bucket to Quantum Gravity. (Birkhauser, Basel, 1995).
[4] J M Bardeen, “Gauge-invariant cosmological perturbations”. Phys. Rev. D 22:1882-1905 (1980).
[5] R M Barnett et al (Particle Data Group), Reviews of Modern Physics 68, 611-732 (1996).
[6] J D Barrow, The Constants of Nature. (Vintage, 2003).
[7] J Barrow and F Tipler, The Cosmological Anthropic Principle. (Oxford University Press, 1984).
[8] C L Bennet et al (the WMAP team), “First Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations:
Preliminary Maps and Basic Results”. Astrophysical Journal Supplement Series 148 1-27(2003), astro-ph/0302207.
[9] R Berendzen, R Hart and D Seeley, Man Discovers the Galaxies. (Science History Publications, New York. 1976).
[10] J A Bernadete, Infinity. An Essay in Metaphysics. (Oxford: Clarendon Press, 1964).
[11] S K Blau and A H Guth. “Inflationary Cosmology”. In 300 Years of Gravitation. Ed S W Hawking and W Israel
(Cambridge University Press, 1987), 524.
[12] M Bojowald, “Absence of singularity in loop quantum cosmology” Phys Rev Lett 86, 5227-5239 (2001).
[13] H Bondi, “Spherically symmetric models in general relativity”. Mon. Not. Roy. Astr. Soc. 107, 410 (1947).
[14] H Bondi, Cosmology. (Cambridge University Press, Cambridge, 1960).
[15] W B Bonnor and G F R Ellis, “Observational homogeneity of the universe”. Mon Not Roy Ast Soc 218, 605-614
[16] G Börner and S Gottlöber (Eds), The Evolution of the Universe. (Wiley, 1997).
[17] N Bostrom, Anthropic bias: Observation selection effects in science and philosophy. (Routledge, New York. 2002).
[18] G Bothun, Modern Cosmological Observations and Problems . (Taylor and Francis, 1998).
[19] K Brading and E Castellani, this vol., Ch. 13.
[20] C Brans and R H Dicke, “Mach’s Principle and a relativistic theory of gravitation”. Phys Rev 124: 925 (1961).
[21] B J Carr and M J Rees, “The Anthropic Principle and the Structure of the Physical World” Nature 278: 605-612
[22] J Carroll (ed.), Readings on Laws of Nature (University of Pittsburgh Press, 2004).
[23] S M Carroll and J Chen, “Does Inflation Provide Natural Initial Conditions for the universe?”. gr-qc/0505037 (2005).
[24] A D Challinor and A N Lasenby, “Cosmic microwave background anisotropies in the CDM model: A covariant and
gauge-invariant approach”. Phys Rev D58, 023001 (1998) astro-ph/9804301.
[25] S Coleman and F de Luccia, “Gravitational effects on and of vacuum decay”. Phys. Rev. D 21, 33053315 (1980).
[26] E J Copeland, J E Lidsey and S Mizuno, “Correspondence between Loop-inspired and Braneworld cosmology”.
[27] L Cowie and A Songaila, “Astrophysical limits on the evolution of dimensionless physical constants over cosmological
time”. Astrophys Journ 453: 596 (1995).
[28] P C W Davies, The physics of time asymmetry. (Surrey University Press, London, 1974).
[29] P C W Davies, The Accidental Universe. (Cambridge University Press, 1982).
[30] P C W Davies, The cosmic Blueprint (London: Heinemann, 1987)
[31] T M Davis and C H Lineweaver, “Expanding Confusion: common misconceptions of cosmological horizons and the
superluminal expansion of the Universe”. Publications of the Astronomical Society of Australia, 21, 97-109 (2004).
[32] D Deutsch, The Fabric of reality: The science of parallel universes - and its implications (Penguin, 1998).
[33] G. de Vaucouleurs, ”The Case for a Hierarchical Cosmology,” Science 167, 1203 (1970).
[34] R H Dicke and P J E Peebles, “The big bang cosmology — enigmas and nostrums”. In General Relativity: An Einstein
Centenary Survey ed. S W Hawking and W Israel (Cambridge: Cambridge University Press, 1979), 504.
[35] ; M I Dickson, this vol. , ch.4.
[36] P A M Dirac, “New basis for cosmology”. Proc Roy Soc A165, 199-208 (1938).
[37] M J Disney, “Visibility of galaxies”. Nature 263 573-575 (1976).
[38] S Dodelson, Modern Cosmology. (Academic Press, 2003).
[39] R Durrer, “Frontiers of the universe: What do we know and what do we understand?” astro-ph/0205101.
[40] J Earman, “The SAP also Rises: A Critical Examination of the Anthropic Principle”, American Philosophical Quarterly 24 : 307-325 (1987).
[41] J Earman, “The Penrose-Hawking Singularity Theorems: History and Implications”. In H. Goenner, J. Renn, and T.
Sauer (eds.), The Expanding Worlds of General Relativity (Birkhäuser, 1999).
[42] J Earman, “Lambda: The Constant That Would Not Die”.
Archive for History of Exact Sciences 55: 189-220
[43] J Earman, “The Cosmological Constant, The Fate of the Universe, Unimodular Gravity, and All That”. Studies in
History and Philosophy of Modern Physics 34: 559-577 (2003).
[44] J Earman and J Mosterin, “A Critical Look at Inflationary Cosmology”. Philosophy of Science 66: 1-49 (1999).
[45] J Ehlers, “Contributions to the relativistic mechanics of continuous media”. Abh Mainz Akad Wiss u Lit (Math Nat
Kl) (1960); reprinted in English, Gen Rel Grav 25, 1225-1266 (1993).
[46] J Ehlers, P Geren and R K Sachs, “Isotropic solutions of the Einstein-Liouville equations”. J Math Phys 9, 1344-1349
[47] J Ehlers and W Rindler, “A Phase-Space Representation of Friedmann-Lemaı̂tre Universes Containing both Dust and
Radiation and the Inevitability of a Big Bang”. Mon. Not. Roy. Astr. Soc. 238 , 503-521 (1989).
[48] D J Eisenstein et al, “Detection of the Baryon Accoustic Peak in the large scale correlation function of galaxies”
(2005) astro-ph/0501171.
[49] G F R Ellis: “The dynamics of pressure-free matter in general relativity”. Journ Math Phys 8, 1171-1194 (1967).
[50] G F R Ellis, “Relativistic cosmology”. In General Relativity and Cosmology, Proceedings of the XLVII Enrico Fermi
Summer School , Ed. R K Sachs (Academic Press, New York, 1971).
[51] G F R Ellis, “Topology and Cosmology”. Gen Rel Grav 2, 7-21 (1971).
[52] G F R Ellis, “Cosmology and Verifiability”. Qu Journ Roy Ast Soc 16, 245-264 (1975).
[53] G F R Ellis, “The homogeneity of the universe”. Gen Rel Grav 11, 281-289 (1979).
[54] G F R Ellis, “Limits to verification in cosmology”. Ann New York Acad Sci 336, 130-160 (1980).
[55] G F R Ellis, “Relativistic cosmology: its nature, aims and problems”. In General Relativity and Gravitation, Ed B
Bertotti et al (Reidel, 1984), 215-288.
[56] G F R Ellis, “A History of cosmology 1917-1955”. In Einstein and the History of General Relativity (Einstein Study
Series, Volume 1). Ed. D Howard and J Stachel. Birkhauser, Boston (1989), 367-431.
[57] G F R Ellis, “Innovation resistance and change: the transition to the expanding universe”. In Modern Cosmology in
Retrospect, Ed. B. Bertotti, R. Balbinto, S. Bergia, and A. Messina. (Cambridge University Press, 1990), 97-114.
[58] G F R Ellis, “Major Themes in the relation between Philosophy and Cosmology”. Mem Ital Ast Soc 62, 553-605
[59] G F R Ellis, “Observations and cosmological models”. In Galaxies and the Young Universe, Ed. H Hippelein, K
Meisenheimer and H-J Roser (Springer,1995), 51-65.
[60] G F R Ellis, “83 Years of General Relativity and Cosmology: Progress and Problems”. CQG 16: A37-76 (Millenium
Issue) (1999).
[61] G F R Ellis, “Cosmology and Local Physics”. New Astronomy Reviews 46: 645-658 (2002) (gr-qc/0102017).
[62] G F R Ellis, “True Complexity and its Associated Ontology”. In Science and Ultimate Reality: Quantum Theory,
Cosmology and Complexity, honoring John Wheeler’s 90th birthday. Edited by John D. Barrow, Paul C.W. Davies,
and Charles L. Harper, Jr. (Cambridge University Press, 2004).
[63] G F R Ellis, “Physics, Complexity, and Causality” Nature 435: 743 (2005). “Physics and the Real World”. Physics
Today 58: 49-55 (2005).
[64] G F R Ellis and J Baldwin, “On the expected anisotropy of radio source counts”. Mon Not Roy Ast Soc 206, 377-381
(1984). G F R Ellis, “Maintaining the Standard”. Nature (News and Views), 416: 132-133 (2002).
[65] G F R Ellis and G B Brundrit, “Life in the infinite universe”. Qu Journ Roy Ast Soc 20, 37-41 (1979).
[66] G F R Ellis and M Bruni, “A covariant and gauge-free approach to density fluctuations in cosmology”. Phys Rev D40
, 1804-1818 (1989).
[67] G F R Ellis and T R Buchert, “The universe seen at different scales”. To appear, Physics Letter A (special Einstein
issue, 2005). gr-qc/0506106.
[68] G F R Ellis and A R King, “Was the Big Bang a Whimper?”. Comm Math Phys 38, 119-165 (1974).
[69] G F R Ellis, U Kirchner and W Stoeger, “Multiverses and physical cosmology”. MNRAS 347: 921-936 (2004).
[70] G F G Ellis and R Maartens, “The Emergent Universe: inflationary cosmology with no singularity and no quantum
gravity era” Class. Quant. Grav. 21: 223-232 (2004)
[71] G F R Ellis, R Maartens, and S D Nel, “The expansion of the universe”. Mon Not Roy Ast Soc 184, 439-465 (1978).
[72] G F R Ellis and M Madsen, “Exact Scalar Field Cosmologies”. Class Qu Grav 8: 667-676 (1991).
[73] G F R Ellis, P McEwan, W Stoeger, and P Dunsby, “Causality in Inflationary Universes with Positive Spatial
Curvature”. Gen.Rel.Grav. 34 (2002) 1461-1481 gr-qc/0109024
[74] G F R Ellis, S D Nel, W Stoeger, R Maartens, and A P Whitman, “Ideal Observational Cosmology”. Phys Reports
124: 315-417 (1985).
[75] G F R Ellis, J J Perry and A Sievers, “Cosmological observations of galaxies: the observational map”. Astronom
Journ 89, 1124-1154 (1984).
[76] G F R Ellis and G Schreiber, “Observational and dynamic properties of small universes”. Phys Lett A115, 97-107
[77] G F R Ellis and D W Sciama, “Global and non-global problems in cosmology”. In General Relativity, ed. L.
O’Raifeartaigh (Oxford University Press, 1972), 35-59.
[78] G F R Ellis and W R Stoeger, “The Fitting Problem in Cosmology”. Class Qu Grav 4: 1679-1690 (1987).
[79] G F R Ellis and W R Stoeger, “Horizons in inflationary universes”. Class Qu Grav 5, 207 (1988).
[80] G F R Ellis, W Stoeger, P McEwan, and P Dunsby, “Dynamics of Inflationary Universes with Positive Spatial
Curvature”. Journal-ref: Gen.Rel.Grav. 34 (2002) 1445-1459. gr-qc/0109023
[81] G F R Ellis and R M Williams, Flat and Curved Spacetimes (Cambridge Univesity Press, 2000).
[82] G F R Ellis and J-P Uzan, “‘c’ is the speed of light, isn’t it?”. Am Journ Phys 73: 240-247 (2005).
[83] G F R Ellis and H van Elst, “Cosmological Models” (Cargese Lectures 1998). In Theoretical and Observational
Cosmology. Ed. M. Lachieze-Ray (Kluwer, Nato Series C: Mathematical and Physical Sciences, Vol 541, 1999), 1-116.
[84] G F R Ellis and H van Elst, “Deviation of Geodesics in FLRW spacetime geometries”. In On Einstein’s Path: Essays
in Honour of Englebert Schucking, Ed A Harvey (Springer, 1999), 203-226.
[85] A C Fabian (Ed.), Origins (Cambridge Univ Press,1989).
[86] B Freivogel, M Kleban, M R Martinez, and L Susskind, “Observational consequences of a landscape”, hep-th/0505232
[87] M Gardner, “WAP, SAP, PAP, and FAP.” The New York Review of Books, 23, No.8. (May 8,1986): 22-25.
[88] M Gardner, Are Universes Thicker than Blackberries? (W W Norton,2003).
[89] A J M Garrett and P Coles, “Bayesian Inductive Inference and the Anthropic Cosmological Principle”. Comments
Astrophys 17: 23-47 (1993).
[90] A Ghosh, G Pollifrone, G Veneziano, “Quantum Fluctuations in Open Pre-Big Bang Cosmology” Phys.Lett. B440
20-27 (1998).
[91] G W Gibbons, S W Hawking and S T C Siklos, The Very Early Universe. (Cambridge University Press. 1983).
[92] G W Gibbons, E P S Shellard, and S J Rankin (Eds), The Future of Theoretical Physics and Cosmology: Celebrating
Stephen Hawking’s 60th Birthday. (Cambridge University Press, 2003).
[93] J Goodman, “Geocentrism reexamined”. Phys Rev D52:1821 (1995).
[94] J R Gott, “Creation of open universes from de Sitter space”. Nature 295, 304 (1982).
[95] J R Gott and Li-Xin Li, “Can the Universe Create Itself?” Phys.Rev. D58, 023501 (1998); astro-ph/9712344.
[96] J Gribbin and M Rees, Cosmic Coincidences. (Black Swan, 1991).
[97] A Grunbaum, “The pseudo problem of creation”. Philosophy of Science, 56, 373-394 (1989); reprinted in [137].
[98] A Grunbaum, “The poverty of theistic cosmology”. British Journal of Philosophy of Science, 55, 561-614 (2004).
[99] J E Gunn, M S Longair and M J Rees, Observational cosmology: Saas-Fee Course 8. (Geneva Observatory, Switzerland, 1978).
[100] A H Guth, “Inflationary universe: A possible solution to the horizon and flatness”. Phys Rev D 23, 347 (1981).
[101] A Guth, The Inflationary Universe: The Quest for a new Theory of Cosmic Origins (Addison Wesley, 1997).
[102] A H Guth, “Eternal Inflation”. astro-ph/0101507. In Proceedings of “Cosmic Questions” Meeting, The New York
Academy of Sciences Press.
[103] W E Harris, P R Durell, M J Pierce, and J Seckers, “Constraints in the Hubble constant from observations of the
brightest red giant stars in a Virgo-cluster galaxy”. Nature 395: 45 (1998).
[104] E R Harrison, Cosmology: The Science of the Universe. (2nd edition) (Cambridge University Press, Cambridge. 2000).
[105] J B Hartle, “Anthropic Reasoning and Quantum Cosmology”. In The New Cosmology, ed R Allen and C Pope (AIP,
2004) gr-qc/0406104.
[106] M Harwit, Cosmic Discovery (MIT Press, 1984).
[107] J Harvey, this vol., Ch. 14.
[108] S W Hawking, “Quantum cosmology” In 300 Years of Gravitation ed S W Hawking and W Israel (Cambridge:
Cambridge University Press, 1987), 631.
[109] S W Hawking, On the Big Bang and Black Holes.(World Scientific, 1993).
[110] S W Hawking and G F R Ellis, The Large Scale Structure of Space-time. (Cambridge University Press, 1973).
[111] D Hilbert, “On the Infinite”. In Philosophy of Mathematics ed P Benacerraf and H Putnam (Englewood Cliff, N. J.:
Prentice Hall, 1964), pp 134-151.
[112] C J Hogan, “Where the Baryons Are”. In Proceedings of Inner Space/Outer Space II, the David Schramm Memorial
Symposium. astro-ph/9912107 (1999).
[113] C J Hogan, “Why the Universe is Just So”. Rev.Mod.Phys 72: 1149-1161 (2003).
[114] F Hoyle, “Cosmological tests of gravitational theories”. In Rendiconti Scuola Enrico Fermi. XX Corso. (Academic
Press, New York.1960).
[115] E Hubble, The Realm of the Nebulae (Yale University Press, 1936).
[116] D Hume, Dialogues and Natural History of Religion. (World Classics, Oxford University Press, 1993).
[117] S Hollands and R M Wald, “An alternative to inflation”. Gen.Rel.Grav. 34 2043-2055 (2002).
[118] W Hu and N Sugiyama, “Toward Understanding CMB Anisotropies and Their Implications”. Phys Rev D 51, 25992630 (1995); “Anisotropies in the CMB: An Analytic Approach Astrophys Journ 444, 489-506 (1995).
[119] C J Isham C J, Lectures on Quantum Theory: Mathematical and Structural Foundations (London: Imperial College
Press, Singapore: World Scientific, 1997).
[120] D Jain and A Dev, “Age of high redshift objects - a Litmus Test for dark energy models”. astro-ph/0509212.
[121] M Kamionkowski and A Loeb, “Getting around cosmic variance”. Phys. Rev. D56, 4511 (1997).
[122] R Kantowski, T Vaughan and D Branch, “The effects of inhomogeneities on evaluating the deceleration parameter”.
Astrophys Journ 447: 35-42 (1995).
[123] R Kantowski, “The effects of inhomogeneities on evaluating the mass parameter Ωm and the cosmological constant
Λ”. Astrophys Journ 507: 483-496 (1998).
[124] J Khoury, B A Ovrut, P J Steinhardt, N Turok, “The Ekpyrotic Universe: Colliding Branes and the Origin of the
Hot Big Bang”. Phys.Rev. D64: 123522 (2001)
[125] J Knobe, K D Olum, and A Vilenkin, “Philsophical implications of Inflatonary Cosmology”. phys/0320071 (2005).
[126] E W Kolb and M S Turner, The Early Universe (Addison Wesley, 1990).
[127] H Kragh, Cosmology and Controversy. (Princeton University Press, 1996).
[128] J Kristian and R K Sachs, “Observations in cosmology”. Astrophys Journ 143, 379 (1966).
[129] T. Kuhn, ”Objectivity, Value Judgment, and Theory Choice”. In The Essential Tension (University of Chicago Press,
1977), pp. 320-39 .
[130] M Lachièze Ray M and J P Luminet , “Cosmic Topology”. Phys Rep 254 135 (1994).
[131] N P Landsman, this vol., ch.5.
[132] D Langlois and F Vernizzi, “Conserved non-linear quantities in cosmology”. astro-ph/0509078.
[133] I Lakatos, The Methodology of Scientific Research Programmes. (Cambridge Univ Press, 1980).
[134] G Lemaı̂tre, “The beginning of the world from the point of view of quantum theory”. Nature 127, 706 (1931),
[135] J Leslie, Universes. (Routledge, 1989).
[136] J Leslie, Cosmology: a philosophical survey. Philosophia, 24: 3-27 (1994).
[137] J Leslie (Ed), Modern Cosmology and Philosophy. (Prometheus Books, 1998).
[138] D K Lewis, On the Plurality of Worlds (Basil Blackwell, Oxford, 1986).
[139] A R Liddle, “How many cosmological parameters?”. astro-ph/0401198.
[140] J E Lidsey, A R Liddle, E W Kolb, E J Copeland, T Barriero, and M Abney, “Reconstructing the Inflation potential
- an overview”. Rev Mod Phys 69, 373-410 (1997).
[141] A D Linde, Particle Physics and Inflationary Cosmology . (Harwood Academic, 1990).
[142] C H Lineweaver and T M Davis, “Misconceptions About The Big Bang”. Scientific American March 2005.
[143] J Luminet, J R Weeks, A Raizuelo, R Lehoucq, and J P Uzan, “Dodecahedral space topology as an explanation for
weak wide-angle temperature correlations in the cosmic microwave background”. Nature 425: 593-595 (2003). J-P
Liuminet, “A cosmic hall of mirrors”. Physics World: 18, 23-28 (2005).
[144] M S Madsen and G F R Ellis, “Evolution of Ω in inflationary universes”. Mon. Not. Roy. Astr. Soc. 234, 67-77 (1988).
[145] D Malament, this vol., ch. 3.
[146] D R Matravers, G F R Ellis, and W R Stoeger, “Complementary approaches to cosmology: Relating theory and
observations”. Qu J Roy Ast Soc 36, 29-45 (1995).
[147] W H McCrea, “Cosmology”. Rep Prog Phys 16: 321 (1953).
[148] W H McCrea, “The interpretation of cosmology”. Nature 186: 1035 (1960).
[149] W H McCrea, “A philosophy for big bang cosmology”. Nature 228: 21 (1970).
[150] D M Meyer, “A distant space thermometer”. Nature: 371, 13 (1994).
[151] C W Misner, “The isotropy of the universe.” Astrophys Journ 151, 431 (1968).
[152] C W Misner, “Mixmaster universe”. Phys Rev Lett 22 1071 (1969).
[153] L Morowitz: The emergence of everything: How the world came to be complex. (Oxford, 2002).
[154] H Mouri and Y Taniguchi, “Hierarchical object formation in th epeculiar velocity field”. astro-ph/0508069.
[155] D J Mulryne, R Tavakol, J E Lidsey, and G F R Ellis, “An emergent universe from a loop”. Phys Rev D71, 123512
(2005); M Bojowald, “Original Questions”. Nature 436: 920-921 (2005).
[156] M K Munitz, “The logic of cosmology”. British J Philos Sci, 13:104 (1962); Cosmic Understanding: Philosophy and
Science of the universe (Princeton University Press. 1986).
[157] N Mustapha, C Hellaby and G F R Ellis, “Large Scale Inhomogeneity vs Source Evolution: Can we Distinguish
Them?” Mon Not Roy Ast Soc 292: 817-830 (1998).
[158] J S Nilsson et al, (Ed) The Birth and Early Evolution of our Universe. Proceedings of Nobel Symposium 19 (World
Scientific, 1991).
[159] J D North, The Meaure of the Universe. (Oxford UNiversity Press, 1965).
[160] T Padmanbhan, Structure formation in the universe. (Cambridge University Press. 1993).
[161] B Partridge, 3K: The Cosmic Microwave Background Radiation. (Cambridge University Press, 1995).
[162] J A Peacock, Cosmological Physics (Cambridge University Press, 1999).
[163] P J E Peebles, Physical Cosmology. (Princeton University Press, Princeton. 1971).
[164] P J E Peebles, The Large-Scale Structure of the Universe. (Princeton University Press, 1993).
[165] P J E Peebles, Principles of Physical Cosmology. (Princeton University Press, 1993).
[166] P J E Peebles, D N Schramm, E L Turner, and R G Kron, “The case for the Relativistic hot big bang cosmology”.
Nature 352 , 769-776 (1991).
[167] R Penrose, “Difficulties with inflationary cosmology”. 14th Texas Symposium Relativistic Astrophysics, ed. E.J. Fergus
(New York Academy of Science, New York) 1989.
[168] R Penrose, The Emperor’s New Mind. (Oxford University Press, 1989).
[169] R Penrose, The Road to Reality. (Jonathan Cape 2004).
[170] A Perez, H Sahlmann and D Sudarsky, “On the quantum origin of the seeds of cosmic structure”. gr-qc/0508100
[171] D Perkins, Particle Astrophysics (Oxford, 2005).
[172] S Perlmutter et al., “Discovery of a supernova explosion at half the age of the universe”. Nature 391: 51 (1998).
[173] M Pettini, “Element Abundances at High Redshifts”. astro-ph/9902173 (1999). M Pettini, K Lipman, and R W
Hunstead, “Element Abundances at High Redshifts: The N/O Ratio in a Primeval Galaxy”. astro-ph/9502077 (2005).
[174] M J Rees, Perspectives in Astrophysical Cosmology. (Cambridge University Press. 1995).
[175] M J Rees, Just six numbers: The Deep Forces that shape the universe (Weidenfeld and Nicholson, London, 1999)
[176] M J Rees, Our Cosmic Habitat (Princeton and Oxford, 2003).
[177] W Rindler, “Visual horizons in world models”. Mon Not Roy Ast Soc 116, 662 (1956).
[178] W Rindler, Relativity: Special, General, and Cosmological (Oxford University Press, 2001).
[179] H P Robertson, “Relativistic cosmology”. Rev. Mod. Phys. 5, 62-90 (1933)
[180] H P Robertson, “Kinematics and world structure”. Astrophys Journ 82 284-301 (1935).
[181] J G Roederer, Information and its role in Nature. (Springer, 2005).
[182] A Rothman and G F R Ellis, “Can inflation occur in anisotropic cosmologies?” Phys Letters B180, 19-24 (1986).
[183] A Rothman and G F R Ellis, “Lost Horizons”. American Journal of Physics 61, 93 (1993)
[184] C Rovelli, Quantum Gravity. (Cambridge UNiversity Press, 2004).
[185] C Rovelli, this vol., ch.12.
[186] S E Rugh and Henrik Zinkernagel, “The Quantum Vacuum and the Cosmological Constant Problem”. Studies in
History and Philosophy of Modern Physics, 33: 663-705 ((2002).
[187] T D Saini, S Raychaudhury, V Sahni, and A A Starobinsky, “Reconstructing the Cosmic Equation of State from
Supernova distances”. Phys.Rev.Lett. 85: 1162-1165 (2000).
[188] A Sandage, “The ability of the 200-inch telescope to distinguish between selected world models”. Astrophys Journ
133, 355 (1961).
[189] A Sandage, R G Kron, M S Longair, The Deep Universe: Saas-Fee Advanced Course 23. (Springer, Berlin, 1993).
[190] P Schneider, J Ehlers and E E Falco, Gravitational Lenses (Berlin: Springer, 1992).
[191] D N Schramm and M S Turner, “Big-Bang Nucleosynthesis enters the precision era”. Rev Mod Phys 70, 303-318
[192] E Schrödinger, Expanding Universes. (Cambridge: Cambridge University Press, 1956).
[193] D W Sciama, “Astrophysical Cosmology”. In General Relativity and Cosmology. Ed R K Sachs. (Academic Press,
New York. 1971), 183-236.
[194] D W Sciama, “Is the Universe unique¿‘ In Die Kosmologie der gegenwart, ed. G Börner and J Ehlers (Serie Piper,
[195] D Scott, “The Standard Cosmological Model”. atro-ph/0510731.
[196] C Seife, Zero: the Biography of a Dangerous Idea. (Penguin, 2000).
[197] C Seife, Alpha and Omega: The Search for the Begining and end of the Universe. (Bantam Books, London. 2003)
[198] C Seife, “Physics enters the twilight zone”. Science 305: 464-466 (2004).
[199] J L Sievers et al, “Implications of the Cosmic Background Imager Polarization Data” (2005). astro-ph/0509203
[200] K Sigurdson and S R Furlanetto, ”Measuring the Primordial Deuterium Abundance During the Cosmic Dark Ages”.
astro-ph/0505173 (2005).
[201] J Silk, A Short History of the Universe (Scientific American Library, 1997)
[202] J Silk, The Big Bang (3rd Edition) Freeman (2001).
[203] J Silk, On the shores of the unknown: A short histpory of the universe(Cambridge University Press, 2005).
[204] R W Smith, The Expanding Universe: Astronomy’s Great Debate 1900-1931. (Cambridge University Press, 1982).
[205] L Smolin, “Did the universe evolve?” Class Qu Grav 9: 173-191 (1992).
[206] Spergel, D. N. et al., “First Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Determination of
Cosmological Parameters”. Astrophysical Journal Supplement Series 148 175-194 (2003). astro-ph/0302209.
[207] G D Starkman and D J Schwarz, “Is the universe out of tune?”. Scientific American, August 2005, 48-55.
[208] P J Steinhardt, “Cosmology confronts the cosmic microwave background”. Int Journ Mod Phys A10, 1091 (1995).
[209] W R Stoeger, G F R Ellis, and U Kirchner. “Multiverses and Cosmology: Philosophical Issues”. astro-ph/0407329.
[210] W Stoeger, R Maartens and G F R Ellis, “Proving almost-homogeneity of the universe: an almost-Ehlers, Geren and
Sachs theorem”. Astrophys Journ 443, 1-5 (1995). gr-qc/0508100
[211] L Susskind, “The anthropic landscape of string theory”. hep-th/0302219. B Freivogel, M Kleban, M R Martinez and
L Susskind, “Observational consequences of a landscape”. hep-th/0505232.
[212] L Susskind, The Cosmic Landscape: String Theopry and the Illusion of Intelligent Design (New York: Little Brown).
[213] L Susskind and J Lindesay, An Introduction To Black Holes, Information And The String Theory Revolution: The
Holographic Universe (World Scientific, 2004)
[214] M Tegmark, “Is ‘the theory of everything’ merely the ultimate ensemble theory?” Ann. Phys. (N.Y.) 270 1-51 (1998).
[215] M Tegmark, “Measuring spacetime: from big bang to black holes”. Science 296, 1427-1433 (2002). astro-ph/0207188.
[216] M Tegmark, “Parallel universes”. In Science and Ultimate Reality: From Quantum to Cosmos, Ed J D Barrow, P C
W Davies and C Harper (Cambridge UNiversity Press 2003), astro=ph/0302131.
[217] M Tegmark et al (the SDSS collaboration), ”Cosmological Parameters from SDSS and WMAP”. Phys Rev D 69,
103501 (2004).
[218] M Tegmark, A Vilenkin, and L Pogosan, “Anthropic predictions for neuturinos masses”. astro-ph/0304536.
[219] F J Tipler, C J S Clarke, and G F R Ellis, “Singularities and Horizons: A Review Article”. In General Relativity and
Gravitation, Ed. A Held (Plenum Press, 1980).
[220] R C Tolman, Relativity, Thermodynamics, and Cosmology. (Clarendon Press, Oxford, 1934; Dover Publications, 1987).
[221] E P Tryon, “Is the universe a quantum fluctuation?” Nature 246: 396 (1973).
[222] N Turok (Ed), Critical Dialogues in Cosmology. (World Scientific, 1997).
[223] J Uffink, this vol, Ch 9.
[224] C Uggla, H van Elst, J Wainwright, and G F R Ellis: “The Past Attractor in Inhomogeneous Cosmology”. Phys. Rev.
D 68: 103502 (1-22), (2003).
[225] A Vilenkin, “Creation Of Universes From Nothing”. Phys.Lett. B117, 25 (1982).
[226] J Wainwright and G F R Ellis (Eds), The dynamical systems approach to cosmology. (Cambridge University Press,
[227] J Wainright, A A Coley, G F R Ellis and M Hancock, “On the Isotropy of the Universe: Do Bianchi VIIh universes
isotropize?” Class Qu Grav 15, 331 (1998).
[228] R M Wald, “Asymptotic behavior of homogeneous cosmological models in the presence of a positive cosmological
constant” Phys Rev D28, 2118 (1983).
[229] R Wald (2005), “The arrow of time and the initial conditions for the universe”, gr-qc/0507094.
[230] A G Walker, “On Milne’s theory of world structure”.
Proc Lond Math Soc 42, 90 (1936).
[231] A G Walker, “Completely symetric spaces”. J Lond Math Soc 19, 219-226 (1944).
[232] J Weeks, J-P Luminet, A Riazuelo, and R Lehoucq, “Well-proportioned universes suppress CMB quadrupole”.
[233] S W Weinberg, Gravitation and Cosmology (Wiley, New York. 1972).
[234] S Weinberg, “The Cosmological constant problem”. Rev. Mod. Phys. 61, 1-23 (1989).
[235] S W Weinberg, “The cosmological constant problem”, astro-ph/0005265; “A priori distribution of the cosmological
constant”. Phys.Rev. D61 103505 (2000), astro-ph/0002387.
[236] J A Wheeler, Einstein’s Vision. (Springer, Berlin. 1968).
[237] M White, L M Krauss, and J Silk, “Cosmic Variance in CMB Anisotropies: From 1◦ to COBE” Astrophys.J. 418 ,
535 (1993).
[238] C M Will, “The confrontation between gravitational theory and experiment”. In General Relativity: an Einstein
Centenary Survey. Ed S W Hawking and W Israel (Cambridge University Press, 1979), 24.
[239] C Wilson, Leibniz’s Metaphysics: A Historical and Comparative Study (Princeton UP, 1989).
[240] E L Wright. A Century of Cosmology. astro-ph/0603750.
[241] M Yoshimura, “The universe as a Laboratory for high energy physics”. In Cosmology and Particle Physics. Ed Li-Zhi
Fang and A Zee (Gordon and Breach, 1988), 293.
[242] H D Zeh, The Physical Basis of the direction of Time. (Springer Verlag, Berlin, 1992).
[243] H Zinkernagel, “Cosmology, Particles, and the Unity of Science”. Studies in History and Philosophy of Modern
Physics,33: 493-516 (2002).
Issues in the Philosophy of Cosmology
Issue A: The uniqueness of the universe
Thesis A1 : The universe itself cannot be subjected to physical experimentation
Thesis A2 : The universe cannot be observationally compared with other universes
Thesis A3 : The concept of ‘Laws of Physics’ that apply to only one object is questionable
Thesis A4 : The concept of probability is problematic in the context of existence of only one object
Issue B: The large scale of the Universe in space and time
Thesis B1 : Astronomical observations are confined to the past null cone, and fade with distance
Thesis B2 : ‘Geological’ type observations can probe the region near our past world line in the very distant past
Thesis B3 : Establishing a Robertson-Walker geometry relies on plausible philosophical assumptions
Thesis B4 : Interpreting cosmological observations depends on astrophysical understanding
Thesis B5 : A key test for cosmology is that the age of the universe must be greater than the ages of stars
Thesis B6 : Horizons limit our ability to observationally determine the very large scale geometry of the universe
Thesis B7 : We have made great progress towards observational completeness
Issue C: The unbound energies in the early universe
Thesis C1 : The Physics Horizon limits our knowledge of physics relevant to the very early universe
Thesis C2 : The unknown nature of the inflaton means inflationary universe proposals are incomplete
Issue D: Explaining the universe — the question of origins
Thesis D1 : An initial singularity may or may not have occurred
Thesis D2 : Testable physics cannot explain the initial state and hence specific nature of the universe
Thesis D3 : The initial state of the universe may have been special or general
Issue E: The Universe as the background for existence
Thesis E1 : Physical laws may depend on the nature of the universe
Thesis E2 : We cannot take the nature of the laws of physics for granted
Thesis E3 : Physical novelty emerges in the expanding universe
Issue F: The explicit philosophical basis
Thesis F1 : Philosophical choices necessarily underly cosmological theory
Thesis F2 : Criteria for choice between theories cannot be scientifically chosen or validated
Thesis F3 : Conflicts will inevitably arise in applying criteria for satisfactory theories
Thesis F4 : The physical reason for believing in inflation is its explanatory power re structure growth.
Thesis F5 : Cosmological theory can have a wide or narrow scope of enquiry
Thesis F6 : Reality is not fully reflected in either observations or theoretical models
Issue G: The Anthropic question: fine tuning for life
Thesis G1 : Life is possible because both the laws of physics and initial conditions have a very special nature
Thesis G2 : Metaphysical uncertainty remains about ultimate causation in cosmology
Issue H: The possible existence of multiverses
Thesis H1 : The Multiverse proposal is unprovable by observation or experiment
Thesis H2 : Probability-based arguments cannot demonstrate the existence of multiverses
Thesis H3 : Multiverses are a philosophical rather than scientific proposal
Thesis H4 : The underlying physics paradigm of cosmology could be extended to include biological insights
Issue I: The natures of existence
Thesis I1 : We do not understand the dominant dynamical matter components of the universe at early or late times
Thesis I2 : The often claimed physical existence of infinities is questionable
Thesis I3 : A deep issue underlying the nature of cosmology is the nature of the laws of physics.
Thesis of Uncertainty: Ultimate uncertainty is one of the key aspects of cosmology
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF