Physical time for the beginning universe

Christof Wetterich
Institut für Theoretische Physik
Universität Heidelberg
Philosophenweg 16, D-69120 Heidelberg

We propose that physical time is based on counting the oscillations of wave functions. The discrete
counting of the ticks of these clocks does not depend on the metric frame. It remains well defined for
the beginning epochs of the universe. The photon clock counts the oscillations of electromagnetic
plane waves in the cosmic reference frame. It can be gauged with clocks for other massless or massive
particles. These equivalent clocks form a clock system which defines a common universal continuous
physical time. For the photon clock this physical time coincides with conformal time. The universe
is eternal towards the past for typical models of inflationary cosmology – the photon clock ticks an
arXiv:2408.01524v1 [gr-qc] 2 Aug 2024

infinite number of times. We compare the photon clock system to geodesic physical time, which is
a frame invariant generalization of proper time.

Time is a central concept for physics and much work metric is geodesically complete [2, 3] if no geodesic stops
has been devoted to define “physical time”, both theoret- at a certain point. For a Robertson-Walker metric the
ically and experimentally. When it comes to the extreme criterion of geodesic completeness is
situation in the beginning epochs of our universe the con-
∞
cept of physical time may have to be rethought, however.
Z
a(t)dt = ∞ . (2)
There is no observer who can construct some clock ap- −∞
paratus under the conditions of the radiation dominated
epoch or even earlier in cosmology. How should a physi- For a finite integral or t not extending to −∞ the geome-
cist define time for these extreme early epochs? The try is geodesically incomplete. A geodesically incomplete
present contribution to this question is based on numer- universe is often called singular and it is believed that it
ous discussions I had with Valery Rubakov over the last cannot give a satisfactory description of the beginning
two years of his life. They resulted in the publication [1], epoch. Typical models of inflationary cosmology have
expressing our common view, with some points of our been shown to be geodesically incomplete [4, 5]. This no-
discussions left open. tion of geodesic completeness is a purely geometric prop-
In general relativity a physical time interval should not erty which only depends on the form of the metric. It
change under a general coordinate transform. Physical does not involve any particle properties.
time has to be diffeomorphism invariant. Most commonly In many models for the beginning epochs of the uni-
one uses for physical time the proper time τ measured by verse scalar fields play an important role, as the inflaton
an observer moving on some geodesics defined for parti- for inflationary cosmology. The dynamics of the cosmo-
cles with a fixed nonzero mass m. An alternative could be logical solution involves then an interplay between the
conformal time η which is associated to massless particles scalar field and the metric. In the presence of a dynami-
as photons. The proper time depends on the geodesics cal scalar field pure geometric concepts for physical time,
chosen. This is no problem as long as the proper times as proper time for an observer moving on geodesics, have
measured on two different geodesics can be mapped onto to be questioned. In many models the masses of par-
each other by an invertible map. A given geodesic defines ticles depend on the scalar field, similar to the mass of
a reference frame. The map between two inertial refer- the electron in the standard model which depends on the
ence frames in flat space is given by the Lorentz trans- value of the Higgs field. In cosmology the scalar field
formations of special relativity. ϕ(x) is not homogeneous in space and time – for a dy-
In cosmology one may choose a preferred “cosmic ref- namical scalar field it will depend at least on time. In
erence frame” defined by the approximate isotropy of the this case the particle masses m(x) become space-time
cosmic microwave background (CMB) in this frame. For dependent. This dependence induces an additional force
a spatially homogeneous and isotropic universe proper (“fifth force”) and particles move no longer on geodesics.
time in the cosmic reference frame coincides with the The trajectories xµ (σ) obey now a generalized geodesic
time t in the Robertson-Walker metric given by equation [1, 6]
ds2 = gµν dxµ dxν = −dt2 + a2 (t)dxi dxj δij . (1)
duµ 1 ∂m2 (x)
It is the proper time measured by a comoving observer + Γµνλ (x)uν uλ + g µν (x) = 0, (3)
whose trajectory obeys dxi = 0. (The sign convention dσ 2 ∂xν
(−, +, +, +) used in this note differs from ref. [1].) where
The notion of proper time measured on geodesics for
massive particles has led to the concept of “geodesic com- dxµ
pleteness”. A geometry with Minkowski signature of the uµ = , (4)
dσ
 2

and σ parameterizes the trajectory. Eq. (3) is equivalent metric frame. Physics is not only invariant under general
to coordinate transformations, which can be considered as
a particular subclass of field transformations. It is in-
gµν (x)uµ (σ)uν (σ) = −m2 (x) . (5) variant under arbitrary field transformations, including
the Weyl transformations (6). General coordinate trans-
Only for constant m eq. (3) reduces to the geometric formations leave the geometry invariant, while this is no
geodesic equation. For general m2 (x) particle physics longer the case for more general field transformations of
properties influence the trajectories in addition to the the metric. In the view of field relativity geometry looses
metric. This is easily understood by the presence of an its fundamental meaning.
additional force. Evaluating proper time on the geomet- These considerations have an important impact on
ric geodesics makes not much sense for particles (or ob- the possible definitions of physical time. Physical time
servers) moving on different trajectories. should be a frame invariant quantity! As we have seen
In the presence of a space-time dependent scalar field this is not the case for proper time. We propose that the
ϕ(x) the choice of the metric is not unique. We could central concept for physical time is the counting of oscil-
choose a different metric g̃µν related to gµν by lations of some periodic process. This is actually the way
how one defines the time standard on Earth – one counts
g̃µν (x) = w2 ϕ(x) gµν (x) .


(6) the number of oscillations for the radiation emitted by
some atomic transition. We may define the periods by
The transformation from gµν to g̃µν is a Weyl scaling or
the zeros of some field component, e.g. the electric field
conformal transformation. The choice of the metric is
of the radiation. If these zeros are mapped to zeros in
called a metric frame. If one computes proper time from an invertible way by a field transformation, the number
the metric g̃µν it differs from the one obtained from gµν .
of zeros remains the same. This is realized by the frame
Proper time is not a frame-invariant quantity. Also the transformation of the metric. Physical time defined by
particle masses m(x) are not frame invariant. The geo-
the number of oscillations is a frame invariant quantity.
metric geodesics are obviously frame-dependent, and this Oscillations occur in early cosmology. For example,
extends to the notion of geodesic completeness based on
we may consider the wave function for the photon or the
geometrical geodesics. Typical inflationary models are associated electromagnetic field in the cosmic rest frame
geodesically incomplete in the Einstein frame for which
which obeys the field equation
the coefficient of the curvature scalar in the effective ac-
tion is given by the constant squared Planck mass. These (∂η2 + 2H∂η + k 2 )ϕ(~k) = 0 . (7)
models can be mapped by a suitable choice of w(ϕ) to
a so called “primordial flat frame” for which geometry Here η = x0 is conformal time, dη = dt/a, H = ∂η ln a,
becomes flat Minkowski space for t → −∞ [7]. In this
frame the geometry is geodesically complete. With ge- gµν = a2 (η)ηµν , ηµν = diag(−1, 1, 1, 1), k 2 = ~k 2 , and ~k is
ometric geodesic completeness depending on the metric the wave vector (momentum) for a plane wave. The field
frame one has to rethink this concept. ϕ(~k) stands for some component of the electromagnetic
In quantum field theory physical observables are inde- field in Fourier space. It can also be a component of
pendent of the choice of fields. This is particularly ap- the graviton wave function or any other massless field.
parent for the quantum effective action Γ which includes An arbitrary electromagnetic field – for example the one
all effects of quantum fluctuations. Similar to the clas- of the CMB or the photons in the radiation dominated
sical action Γ is a functional of fields – in this case the epoch – can be composed as a linear superposition of
macroscopic or mean fields. The quantum field equations Fourier modes ϕ(~k). For defining a clock we focus on a
obtain by variation of this effective action with respect particular wave number ~k. For a given wave number ~k
to the fields. For given Γ these are exact equations with- the solution of the wave equation (7) is a standing wave,
out any further corrections from quantum fluctuations. which reads in position space
The correlation function (connected two-point function)
ϕ(η, ~x) = B(η) exp i(~k~x − kη) .

is given by the inverse second functional derivative of (8)
Γ. Higher correlation functions (vertices) are related to
higher functional derivatives, such that the effective ac- The smooth function B(η) accounts for the Hubble
tion contains the full information about all observables. damping. For high momentum modes with k 2 ≫ H2 the
One can perform arbitrary field transformations for the function B is essentially constant. For a given position
macroscopic fields which are the arguments of Γ. Func- ~x one can count the number of oscillations n~k . Discrete
tional derivatives are transformed accordingly, as well as physical time is proportional to n~k , where the propor-
the relations between observables and functional deriva- tionality coefficient sets the time units.
tives. At the end, observables do not depend on the Plane waves with two different ~k constitute different
choice of the macroscopic fields – see ref. [8] for a recent clocks. The number of oscillations is not the same, but
discussion. For cosmology, this basic property of quan- proportional to each other. The number of oscillations
tum field theory has been called “field relativity” [9]. It nk~1 of the first clock is related by an invertible function
states that the physics is independent of the choice of the to the number of oscillations nk~2 of the second clock. The
 3

different clocks can therefore be gauged in order to de- that are not equivalent to the photon clocks. This will
fine a unique time by choosing appropriate proportional- bring us to the question which clocks are most useful
ity coefficients. This is a first example of a clock system. for a definition of physical time, which we will discuss
We call a clock system an ensemble of clocks for which at the end of this note. Our investigation has suggested
the ticks of any pair of two clocks can be mapped to each a frame-invariant generalization of the notion of geodesic
other by an invertible function. It can be seen as an completeness, namely “time completeness”. The universe
equivalence class of clocks. For the clock system of the is time-complete if physical time is not bounded towards
photon waves in the cosmic rest frame a useful continuous the beginning or end. The clocks defining physical time
physical time is conformal time η. For the high momen- never stop ticking, neither towards the beginning nor to
tum modes the number of oscillations is proportional to the end. If there exist inequivalent clock systems, time-
η completeness will depend on the choice of the clock sys-
tem. We briefly summarize here the results of ref. [1].
kη Consider the clocks defined by a free scalar field with
n~k = . (9)
2π mass m(x), minimal coupling to gravity and canonical
kinetic term. The field equation for a scalar field with
Continuous time can be seen as the limit of discrete time
mass depending on the space-time coordinates x reads
for a very large number of oscillations.
The clock system of photons can be extended to in- g µν (x)Dµ ∂ν − m2 (x) ϕ(x) = 0 ,


(10)
clude oscillations for gravitons or any other massless field.
One may ask what happens for the oscillations of the where Dµ is the covariant derivative in presence of the
wave function for massive particles, or for oscillations in metric gµν (x) and g µν (x) is the inverse metric. We look
different reference frames. Do they belong to the same for fast oscillations
clock system as the “photon clock” (8), or do they con-
stitute different clock systems? For late cosmology the


ϕ(x) = A(x) exp iS(x) , (11)
map between all these different clocks is invertible. We
can express the number of ticks in terms of conformal where S(x) obeys
time η. For example, the ticks of an atomic clock on
Earth can be expressed in terms of a corresponding in- g µν ∂µ S∂ν S + m2 (x) = 0 , (12)
terval of conformal time. Similarly, proper time of an
observer moving on geodesics, or on more general phys- and A(x) is a slowly varying function. Since eq. (12) is
ical trajectories, can be expressed in terms of conformal invariant under Weyl scalings the quantity S(x) is frame
time. invariant. For counting oscillations we have a tick of the
The issue of equivalence concerns the beginning of the clock whenever S decreases by 2π.
universe. For the photon clock system the physical time The definition of a clock needs a trajectory xµ (σ) on
towards the beginning of the universe is infinite for many which S is evaluated. In the late universe we may identify
models. For common inflationary models conformal time xµ (σ) with the trajectory of an observer. The oscillations
extends to −∞ as one moves backwards in cosmic time of the scalar field are given by a solution of eq. (12) and
or some other time variable. The number of ticks of the do not depend on the presence of an observer, while the
photon clocks is infinite since the “beginning”. The time- number of zeros seen in some interval of σ depend on
interval between two ticks shrinks if one uses cosmic time the trajectory. We can replace the observer by a parti-
towards the beginning, while it remains constant for con- cle moving on the trajectory xµ (σ). This definition of
formal time. Still, the number of ticks is the same in the a “particle clock” can be used for early cosmology. A
two frames. If one uses the photon clock system for the given trajectory defines a reference frame. For a given
definition of physical time the universe exists since infi- solution for S there exists a large family of clocks defined
nite time – it is eternal [10]. by different reference frames. Our concept of physical
In contrast, for clocks defined by massive particles time related to the oscillations of some field induces in a
moving on suitable trajectories the number of ticks may natural way the notion of reference frames.
remain finite in the limit η → −∞. In this case these If cosmology is homogeneous and isotropic in the aver-
clocks belong to a different clock system. The map from age one possible choice is the “cosmic reference frame”.
such “massive particle clocks” to the photon clocks does It is given by a comoving trajectory with constant xi , i =
not remain invertible for η → −∞. The issue of inequiv- 1 . . . 3, and x0 = cσ. We could also use the physical tra-
alent clock systems is directly related to the question: jectories of particles with mass m̄(x) which obey eq. (3)
Does the universe exist since infinite physical time? defining the generalized geodesics. For m̄(x) = m(x) a
In ref. [1] we have studied the clocks for massive par- particular clock counts for a massive particle moving on a
ticles in detail, in particular the case where the masses generalized geodesic the number of oscillations of its own
depend on some scalar field and therefore on time. This wave function. This “geodesic physical time” is analo-
leads to the concept of a frame-invariant generalization of gous to proper time for an observer moving on geodesics.
proper time for particles moving on generalized geodesics. In contrast to proper time the ticks of this clock are frame
We have found in some cases clocks for massive particles invariant.
 4

We focus in the following on m̄(x) = m(x). In this and m depending only on time. The wave equation (7)
case a family of generalized geodesic trajectories can be receives an additional contribution for massive particles
related [1] to the solution for S(x) by
∂η2 + 2H∂η + k 2 + a2 m2 ϕ(~k) = 0 .


(17)
dxµ ∂S
= uµ = g µν ν . (13)
dσ ∂x This results in
From eqs. (13), (14) one concludes that dσ involves the
p
S = S̃(η) + ki xi , ∂η S̃ = − k 2 + a 2 m2 . (18)
geometric proper time interval dτ ,
1 p µν dτ For the rapid oscillations of interest both a2 m2 and ∂η H
dσ = −g (x)dxµ dxν = . (14) are small as compared to k 2 . The solution of the field
m(x) m(x)
equation approaches the one for massless particles.
Neither dσ nor dτ are frame invariant. If we compute the Let us first consider the cosmic rest frame with trajec-
number of ticks of the clock, however, we find a frame tory
invariant quantity
x0 = η = cσ , xi = const. (19)
∂S
Z Z
S(σ) − S(σ0 ) = dxµ µ = − m2 (x)dσ Physical time is given by
∂x
Z Z η Z η
= − m(x)dτ . (15)
p
∆T = −∆S = − dη∂η S̃ = dη k 2 + a 2 m2 .
ηin ηin
We define “geodesic physical time” T by (20)
For all solutions with k 2 > 0 the physical time inter-
dT = −dS = m(x)dτ . (16) val ∆T diverges for ηin → −∞. These clocks are time
complete. They belong to the same clock system as the
The frame invariant generalization of the proper time in-
photon clock. In particular, if a2 (η)m2 (η) approaches
terval m(x)dτ has been proposed in refs. [11], [10]. Since
zero for ηin → −∞, the ticks of the particle clocks coin-
geodesic physical time is invariant under Weyl transfor-
cide in this limit with the ticks of the photon clock for
mations it can be evaluated in any metric frame related
the same k.
by such transformations. One may introduce time units
We next consider the clocks for trajectories corre-
by multiplying the dimensionless ∆t with a suitable pro-
sponding to the generalized geodesics for the massive
portionality constant. For constant m geodesic physical
particle given by eq. (13)
time coincides with proper time.
For a given solution ϕ(x) or S(x) of the scalar field 1 1p 2
equation (10) or (12) there are many trajectories which u0 = − ∂η S = k + m2 a 2 ,
a2 a2
obey eq. (13). These trajectories are distinguished by ini- 1 1
tial conditions xµ (σ0 ). For all those trajectories eq. (16) ui = 2 ∂i S = 2 k i . (21)
a a
is valid. The clocks tick differently for different tra-
jectories since m(x) and dτ have to be evaluated for a One infers from eq. (16) the geodesic physical time inter-
given trajectory. Furthermore, there are many trajecto- val
ries which obey the generalized geodesic equation (3), (4),
but do not obey eq. (13). For those trajectories we can a 2 m2
dT = √ dη . (22)
still count the periods and define dT = −dS. The re- k 2 + a 2 m2
lation (16) does not hold, however. All solutions of the
generalized geodesic equation define clocks in different This differs from the result for the cosmic rest frame (20).
reference frames. The question arises how the clocks in If a2 m2 remains finite for η → −∞ the geodesic physical
different reference frames are related. For flat space, time is equivalent to the one defined by the photon clock.
constant m and two trajectories specified by the four On the other hand, if a2 m2 → 0 for ηin → −∞, one has
momenta pµ and qµ of a massive particle one can show 1 η
Z
that the time intervals ∆T in the two reference frames ∆T = dηa2 (η)m2 (η) . (23)
k ηin
are related by a Lorentz transformation [1]. The general
case with space-time dependent gµν (x) and m(x) can be For
rather complex. One finds that clocks defined for dif-
ferent reference frames are not always equivalent – they am = |η|−α , (24)
may belong to different clock systems. There is no uni-
versal particle clock. Another interesting issue concerns the geodesical physical time interval remains finite for
the question which particle clocks belong to the same ηin → −∞ if α ≥ 1/2. In this case the particle clocks on
clock system as the photon clock. trajectories given by generalized geodesics (13) do not
We demonstrate these issues for a homogeneous belong to the same clock system as the photon clock.
isotropic cosmology with Robertson-Walker metric (1) They are not time complete or “geodesically complete”
 5

in a generalized sense. For example, this happens for de families of trajectories for which the photon clock is time
Sitter space with constant m where α = 1. Since dT is complete. As long as inhomogeneities are moderate one
frame invariant, the same happens in the corresponding may define an isotropic homogeneous rest frame by a suit-
primordial flat frame where a approaches a constant for able averaging of the metric. Further work is necessary
η → −∞, while m vanishes in this limit. in order to clarify this issue. For most practical applica-
One would like to understand why the particle clocks tions in cosmology one can simply identify physical time
on generalized geodesics can stop ticking for η → −∞, for the photon clock system with conformal time for the
whereas the photon clocks tick at a constant pace. This “background metric”. For typical inflationary cosmolo-
is best understood in a frame where a is constant, while gies the universe is eternal.
m vanishes for η → −∞. With a = 1 conformal time
equals cosmic time and the photon wave functions are In conclusion, the definition of physical time by count-
simple plane waves. In lowest order in m/k the particle ing oscillations of wave functions is well adapted to the
wave function obeys beginning epoch of the universe when no observers with
Z η a measuring apparatus have existed. The discrete count-
1 ing does not depend on the metric frame, in contrast to
S = ~k~x − kη − dη ′ m2 (η ′ ) . (25)
2k −∞ proper time. For a given solution of the wave function the
number of oscillations, which corresponds to the number
The wave front S = 0, ~x = xw~k/k, obeys of ticks of a clock, depends on the trajectory on which the
Z η counting is done. This introduces the notion of reference
1 m2 frames in a wide context.
xw = η + 2 dη ′ m2 (η ′ ) , ∂η xw = 1 + 2 . (26)
2k −∞ 2k
A rather natural clock is the photon clock which counts
The trajectory of the particle moving on generalized the oscillations of electromagnetic plane waves in the cos-
geodesics, ~x = xp~k/k, can be characterized by its ve- mic reference frame. For the cosmic reference frame all
locity, clocks defined by the oscillations of other massless fields
as the metric, or equivalently of the wave functions for
k m2 massive particles, belong to the same clock system as
∂η xp = √ ≈1− 2 . (27)
k2 +m 2 2k the photon clock. They can be mutually gauged to each
other. The corresponding universal time can be identi-
For m > 0 the particle is slower than the wave front, fied with conformal time. Clocks in the cosmic reference
∂η xp < ∂η xw . In the reference frame of the particle frame tick an infinite number of times as one moves back-
one can count the oscillations of the wave function which ward for most common inflationary models. For these
overtakes it. These oscillations determine the ticks of models the universe is time complete or eternal.
the clock in the particle rest frame. In the limit m → 0
the particle velocity and the wave front velocity approach Geodesic physical time counts for a massive particle
each other. From the point of view of the particle rest moving on a physical trajectory the number of oscilla-
frame the wave front gets frozen. The interval between tions of its own wave function. The physical trajectories
two ticks gets longer and longer. If m decreases fast are generalized geodesics which take into account addi-
enough, the geodesic particle clocks stops ticking for tional forces for a space-time dependent mass. Geodesic
η → −∞. physical time generalizes proper time for the case of vary-
In the presence of two inequivalent clock systems the ing mass. In contrast to proper time it is a frame-
debate is open which clock system is best suited for a de- invariant quantity. For certain cases where the parti-
scription of physical reality. The discussion on this issue cle mass or the scale factor vanish too rapidly towards
was ongoing between Valery Rubakov and me [1]. An ad- the beginning of the universe the geodesic particle clock
vantage of geodesical physical time for particle clocks is stops ticking in this limit. This reflects the well known
its formulation for arbitrary metrics and arbitrary trajec- fact that the rest frame is not well defined for massless
tories. On the other side, one may consider as an advan- particles.
tage of the photon clock system that waves of massless
particles as photons or gravitons exist for the beginning The geodesic physical time can be defined for arbi-
epoch of all realistic cosmologies. For the cosmic ref- trary inhomogeneous metrics and particle masses. For
erence frame the limit of a particle getting massless is the photon clock further work is needed for the gener-
smooth. alization of the cosmic rest frame to arbitrary inhomo-
An open point for the photon clock system is the for- geneous metrics. Since massless particles, or effectively
mulation of a generalized cosmic rest frame for arbitrary massless particles are omnipresent in models of the be-
inhomogeneous metrics. Oscillating wave functions for ginning universe, the photon clock system seems to me to
photons or gravitons exist for rather arbitrary metrics, at be the most useful definition of physical time. It can be
least if deviations from homogeneity and isotropy remain used even if the beginning is characterized by quantum
small. The issue concerns the choice of reference frames. scale symmetry where no intrinsic mass or length scale
For weak inhomogeneities of the metric there are large exists and all excitations or fluctuations are massless.
 6

ACKNOWLEDGMENTS

This note is dedicated to my friend Valery, whose deep

physical insights and profound honesty and engagement
I have admired.

[1] V. A. Rubakov and C. Wetterich, Geodesic [7] C. Wetterich, Primordial flat frame: A new view on in-
(in)completeness in general metric frames (2022), flation, Physical Review D 104 (2021).
arXiv:2210.11198 [gr-qc]. [8] C. Wetterich, Field transformations in functional in-
[2] R. Penrose, Gravitational collapse and space-time singu- tegral, effective action and functional flow equations
larities, Phys. Rev. Lett. 14, 57 (1965). (2024), arXiv:2402.04679 [hep-th].
[3] S. W. Hawking, Singularities in the universe, Phys. Rev. [9] C. Wetterich, Universe without expansion, Physics of the
Lett. 17, 444 (1966). Dark Universe 2, 184–187 (2013), arXiv:1303.6878 [astro-
[4] A. Borde, A. H. Guth, and A. Vilenkin, Inflationary ph.CO].
spacetimes are incomplete in past directions, Physical [10] C. Wetterich, Eternal universe, Physical Review D 90
Review Letters 90 (2003). (2014), arXiv:1404.0535 [gr-qc].
[5] A. Mithani and A. Vilenkin, Did the universe have a [11] I. Bars, P. J. Steinhardt, and N. Turok, Cyclic cosmology,
beginning? (2012), arXiv:1204.4658 [hep-th]. conformal symmetry and the metastability of the higgs,
[6] Y. Ayaita, M. Weber, and C. Wetterich, Structure forma- Physics Letters B 726, 50–55 (2013), arXiv:1303.6878
tion and backreaction in growing neutrino quintessence, [gr-qc].
Physical Review D 85 (2012), arXiv:1112.4762 [astro-
ph.CO].