Time in Quantum Theory*

H. D. Zeh
www.zeh-hd.de

I. In general, time is used in quantum theory as an external ('classical') concept. So it is

assumed, as in classical physics, to exist as a controller of all motion – either as absolute time
or in the form of proper times defined by a classical spacetime metric. In the latter case it is
applicable to local quantum systems along their world lines. According to this assumption,
time can be read from appropriate classical or quasi-classical 'clocks'.
This conception has to be revised only when relativistic gravity, which regards the
spacetime metric as a dynamical object, is itself quantized [1] – as required for consistency
(see IV). The thereby aspired quantization of time does not necessarily lead to a discretization
of time – just as the quantization of free motion does not require a discretization of space or
position. On the other hand, the introduction of a fundamental gravitational constant in
addition to Planck's constant and the speed of light leads to a natural 'Planck time' unit,
corresponding to 5.40 10-44 sec. This may signal the need for an entirely novel conceptual
framework – to be based on as yet missing empirical evidence. A formal (canonical) quant-
ization of time would also be required in non-relativistic Machian ('relational') dynamical
theories [4], which consistently replace the concept of time by some reference motion. If
quantum theory is universally valid, all dynamical processes (including those that may serve
as clocks or definers of time) must in principle be affected by quantum theory. What does this
mean for the notion of time?
Historically, the dynamics of quantum systems seemed to consist of individually
undetermined stochastic 'quantum jumps' between otherwise 'stationary' states (energy
eigenstates) – see [2] for an early review of the formalism and the attempt of an interpretation.
Such stochastic events are observed in quantum measurements, in particular. For this reason,
von Neumann [3] referred to the time-dependent →Schrödinger equation as a 'second
intervention', since Schrödinger had invented it solely to describe consequences of time-
dependent external 'perturbations' on a quantum system. Note, however, that atomic clocks
are not based on any stochastic quantum events, even though they have to be designed as open
systems in order to allow their permanent reading (representing 'measurements' of the clock –
see IV).
In a consistent →Schrödinger picture, all dynamics is described as a time dependence
of the quantum states, while the observables are fixed formal kinematical concepts (see also
Sect. 2.2 of [5]). The time dependence according to the Schrödinger equation can be
completely understood as an interference phenomenon between different stationary states
|m>, which possess individually meaningless phase factors exp(iωmt). Their →superpositions
are able to describe time-dependent quantum states |α(t)> in the form
|α(t)> := ∫dq ψα(q,t)|q> = Σmcmexp(iωmt)|m> .
The wave function ψα(q,t) is here used to define the time-dependent state |α(t)> in abstract
→Hilbert space. The Hilbert space basis |q> diagonalizes an appropriate observable Q. The
time dependence of a quantum state is in fact meaningful only relative to such a fixed basis,
as demonstrated by means of the wave function in the above definition.
In non-relativistic quantum mechanics, the time parameter t that appears in the Schrö-
dinger wave function ψ(q,t) is identified with Newton's absolute time. So it is presumed to
*
To appear in: Compendium of Quantum Physicis - Concepts, Experiments, History and
Philosophy, ed. by F. Weinert, K. Hentschel, D. Greenberger, B. Falkenburg (Springer 2008)
– version August 2007
1
exist regardless of how or whether it is measured. The letter q represents all variables qi
(i=1…I) that span the required configuration space. The special case of a point mass, where q
≡ x,y,z corresponds to a single space point, has often led to confusion of the wave function
with a time-dependent spatial field (relativistically a field on spacetime). In N-particle
mechanics, for example, the configuration space variables q are equivalent to N space points
(that is, I=3N variables). In quantum field theory, the amplitudes of all fields Φ(x,y,z) at all
space points even form a continuum. These variables are thus distinguished from one another
by their spatial arguments, which now assume the role of 'indices' to Φ, just as i for the
variables qi [6]. Therefore both, space and time, are in this approach assumed to be absolutely
defined classical preconditions for kinematics and dynamics – even though they appear in the
formalism in different ways.
If the variables q are field amplitudes, the canonical quantization of n fields, for
example, leads to a time-dependent wave functional Ψ[Φ1(r),…,Φn(r),t], rather than to n field
operators on spacetime. This conclusion holds relativistically, too. The corresponding Hilbert
space readily includes superpositions of different 'particle' numbers ('occupation numbers').
For bosons, the latter are simply oscillator quantum numbers for the eigenmodes (first
postulated by Planck, and later explained by Schrödinger by the numbers of nodes of their
wave functions). The ultimate universal Hilbert space basis is hoped to be discovered in a
unified field theory.
Schrödinger's general wave function ψ(q,t) may be Fourier transformed with respect to
all its arguments – in spite of their different interpretations. This transformation defines wave
numbers k in the configuration space and frequencies ω. They may be rescaled into canonical
momenta (in general different from conventional, that is, spatial momenta) and energies by
means of Planck's constant. The Fourier transformation gives rise to a formal 'time operator',
T := i∂/∂ω, that allows one to define a continuous shift operation for frequencies: U(Δω) :=
exp(iΔωT). It does not in general transform a solution of the Schrödinger equation into
another solution, since this would require a continuous and unbounded energy spectrum. Pairs
of Fourier variables are subject to the Fourier theorems,
ΔqΔk ≥ 1 and ΔtΔω ≥ 1 ,
which apply to all functions ψ(q,t) – regardless of the existence of any dynamical law or a
Hamiltonian H. These 'uncertainty relations' between corresponding variables must have
physical consequences when applied to solutions of the Schrödinger equation. Those based on
the Fourier theorem relating time and frequency are usually interpreted as representing a
'time-energy uncertainty relation' (see [7]). Well known, for example, is the spectral line
width required for metastable states. A 'time uncertainty' can also be defined by the finite
duration of a preparation or measurement process.
II. The situation is somewhat obscured in the →Heisenberg picture. In the algebraic
Born-Heisenberg-Jordan quantization procedure, 'observables' were introduced in formal
analogy to the classical dynamical variables, such as q(t) and p(t), while quantum states were
not regarded as dynamical objects. Observables would assume definite values only in
appropriate measurements or discrete 'quantum events' (von Neumann's first intervention –
historically related to Bohr's quantum jumps between his discrete classical orbits). Time
durations are then often defined operationally by means of pairs of such events – not
according to the Schrödinger dynamics. The latter is here merely regarded as a tool for
calculating probabilities for the occurrence of events, which are thereby assumed to represent
the only real quantum phenomena.
Note that in the Heisenberg picture certain instantaneous properties of quantum states
appear to represent some hidden time dependence. For example, the kinetic energy operator in
the Schrödinger picture (the Lapacean) measures the curvature of the wave function ψ(q,t) at
given time t – not any quantitiy related to motion, such as the classical kinetic energy, while

2
its minimum (achieved for a wave function that nowhere changes sign) is in the Heisenberg
picture interpreted as describing 'zero point fluctuations' of the corresponding variables q.
This picture has led to much confusion – including the search for a 'time observable' T
that would depend on the specific system Hamiltonians H by obeying commutation relations
[T,H] = ih ,
in analogy to position and momentum observables (see the Introduction to [8] for a review).
However, since realistic Hamiltonians possess a ground state, their spectra are bounded from
below, and a time operator obeying this commutation relation cannot possess a spectrum
represented by the real numbers (as pointed out by Wolfgang Pauli [2]). It may nonetheless be
related to time intervals between certain pairs of events that can be measured at a system
characterized by the Hamiltonian H.
A formal equivalence between the Schrödinger and a Heisenberg picture for the
purpose of calculating expectation values of measurement results is known to hold for
isolated, unitarily evolving systems (which are exceptions in reality). For asymptotically
isolated objects participating in a scattering process one may use the interaction picture,
where part of the Hamiltonian dynamics is absorbed into the observables characterizing
asymptotic states. This includes the 'dressing' of quantum fields. However, macroscopic
systems always form open systems; they never become isolated, even when dressed. Such
systems may approximately obey effective non-unitary dynamics (master equations). In
principle, this dynamics has to be derived from the unitary (Schrödinger) evolution of an
entangled global quantum state, that would include all 'external interventions'. Under realistic
assumptions this leads to permanently growing →entanglement with the environment –
locally observed as →decoherence [5].
This extremely fast and in practice irreversible process describes a dislocalization of
quantum superpositions. It thereby mimics quantum jumps (events): components which
represent different macroscopic properties (such as different pointer positions or different
registration times of a detector) are almost immediately dynamically decoupled from one
another – though none of them is selected as the only existing one. Pauli, when arguing in
terms of the Heisenberg picture, regarded such events as occuring 'outside the laws of nature',
since they withstood all attempts of a local dynamical description. In the global Schrödinger
picture, the time-asymmetry of this dynamical decoupling of components ('branching') can be
explained in terms of the time-symmetric dynamics by means of an appropriate initial
condition for the wave function of the universe – the same condition that may also explain
thermodynamical and related time asymmetries ('arrows of time') [9]. In essence, this initial
condition requires that non-local entanglement did not yet exist just after the big bang, and
therefore has to form dynamically ('causally'). The resulting asymmetry in time may give rise
to the impression of a direction of time.
III. In quantum field theory, a Schrödinger equation that controls the dynamics of the
field functionals may well be relativistic – containing only local interactions with respect to
the space-dependent field variables (in this way facilitating the concept of a Hamiltonian
density in space). A wave function(al) obeying a relativistic Schrödinger equation never
propagates faster than light with respect to the underlying presumed absolute spacetime.
Recent reports of apparently observed superluminal phenomena were either based on
inappropriate clocks, or on questionable interpretations of the wave function. For example, the
exact energy eigenstate of a particle, bound to an attractive potential in a state of negative
energy E = -|E|, would extend to spatial infinity according to exp(-√|E|r) outside the range of
the potential. It has therefore been claimed to be able in principle to cause effects at an
arbitrary distance within any finite time [10]. However, if the wave function of the bound
system forms dynamically (according to the Schrödinger equation rather than by quantum
jumps), it can only subluminally approach the exact eigenstate with its infinite exponential
tail. This time-dependence requires a minimum energy spread that is in accord with the time-

3
frequency Fourier theorem. Similar arguments hold relativistically also for particle number
eigenstates, which cannot have sharp spatial boundaries because of Casimir type effects (in
principle observable for moving mirrors); all bounded systems must relativistically be in
superpositions of diffent particle numbers.
In the theory of relativity, proper times assume the role of Newton's absolute time for
all local systems, that is, for those approximately following world lines in spacetime. How-
ever, quantum states are generically nonlocal, and they do not consist of or define local
subsystem states. One may then introduce auxiliary time coordinates (arbitrary spacetime
foliations) in order to define the dynamics of global states on these artificial 'simultaneities'.
A Hamiltonian (albeit of very complex form – in general including a whole field of Coriolis-
type forces with effective 'particle' creation and annihilation terms) would nonetheless exist in
this case. As these artificial simultaneities may be assumed to propagate just locally, one
speaks of 'many-fingered time'. Dynamical evolution in quantum theory is in general locally
non-unitary (to be described by a master equation) because of the generic nonlocal entangle-
ment contained in the unitarily evolving global quantum state. Unitary evolution may there-
fore be confirmed only in exceptional, quasi-isolated (microscopic) systems.
IV. According to Mach's ideas, no concept of absolute time should be required or
meaningful. Any time concept could then be replaced by simultaneity relations between
trajectories of different variables (including appropriate clocks) – see [4] and Chap. 1 of [9].
Classically, timeless trajectories q(λ), where λ is an arbitrary and physically meaningless
parameter, are still defined. Mach's principle requires only that the fundamental dynamical
laws are invariant under reparametrizations of λ. In quantum theory, the wave function cannot
even depend on such a time-ordering parameter, since there are no trajectories any more that
could be parametrized. This excludes even dynamical successions of spatial geometries (the
dynamical states of general relativity), which would form a foliation of spacetime. On the
other hand, any appropriate variable q0 that is among the arguments of a time-less wave
function ψ(q) may be regarded as a more or less appropriate global physical clock. According
to the superposition principle, superpositions of different values q0 – that is, of different
'physical times' – would then have to exist as real physical states (just as the superpositions of
different values of any physical variable).
In conventional quantum mechanics, superpositions of different times of an event are
well known. For example, a coherently decaying metastable state (that can be experimentally
confirmed to exist by means of interference in the case of decay fragments only weakly
interacting with their environment) is a superposition of different decay times. Similarly, the
quantum state for a single variable x and a clock variable u, say, would have to be described
by a wave function ψ(x,u). This means that the classical dependence of x on clock time u,
defined by their time-less trajectory x(u), is replaced by the less stringent entanglement
between x and u that is defined by such a wave function [11]. The clock variable u becomes
quasi-classical only if it is robust under environmental decoherence, such that superpositions
of different times u always remain dislocalized (locally inaccessible). The same conclusion
holds for the mentioned superposition of different decay times if its corresponding partial
waves (wave packets forming thin spherical shells in space unless reflected somewhere) are
decohered from one another.
Atomic clocks, in particular, are based on the time-dependent superposition of two
close atomic energy eigenstates (defining 'beats'). These oscillating states would immediately
decohere whenever they were measured (read). Therefore, they have to be dynamically
correlated with the coherent state of a maser field that is in resonance with them. This time-
dependent coherent state is known to be 'robust' against decoherence – including genuine
measurements [12]. So it permits the construction of a quasi-classical atomic clock that can be
read. Exactly classical clocks would be in conflict with the uncertainty relations between
position and momentum of their 'hands'.

4
 The above-described consequences of Mach's principle with respect to time do indeed
apply in general relativity to a closed universe. Spatial geometries on a time-like foliation of
spacetime, which would classically determine all proper times [13], are now among the
dynamical variables q (arguments of the wave function) – similar to the mentioned clock
variable u. Moreover, material clocks intended to 'measure' these proper times within a given
precision would have to possess a minimum mass in order to comply with the uncertainty
relations [14], while this mass must then in turn disturb the spacetime metric.
A time coordinate t in general relativity is a physically meaningless parameter (such
as λ – not u – in the above examples). Invariance of the theory under reparametrization, t →
f(t), requires a 'Hamiltonian constraint': H = 0 [1,15]. In its quantum mechanical form, HΨ =
0, this leads to the trivial Schrödinger dynamics ∂Ψ/∂t = 0, where Ψ is now a wave functional
on a configuration space consisting of spatial geometries and matter fields. As this conse-
quence seems to remain valid for all unified theories that contain →quantum gravity, one has
to conclude that there is no time on a fundamental level; all dynamics is encoded in the static
entanglement described by Ψ. Surprisingly, though, the time-less Wheeler-DeWitt equation
[1],
HΨ = 0 ,
(also called an Einstein-Schrödinger equation) becomes hyperbolic for Friedmann type uni-
verses – similar to a relativistic wave equation on spacetime (see Sect. 2.1 of [9]). This allows
one to formulate a complete boundary condition for Ψ in the form of an 'intrinsic initial
condition' [16]. It requires Ψ and its first derivative to be given on a 'time-like' hypersurface,
defined according to the hyperbolic form of the kinetic energy operator contained in H (now a
d'Alembertian), in this universal configuration space (DeWitt's 'superspace'). For example,
such initial data can be freely chosen at a small value of the expansion parameter a of the
universe. A low-entropy condition at a → 0 then leads to an 'intrinsic arrow of time': total
entropy on time-like hypersurfaces must grow (for statistical reasons) as a function of the size
of the universe – regardless of any external concept of time.
Quasi-classical time can here only be recovered within the validity of a Born-Oppen-
heimer approximation with respect to the square root of the inverse Planck mass [15], while
spatial geometry, which defines all fundamental physical clocks, is strongly entangled with,
and thus decohered by, matter [17]. In analogy to the coherent set of apparent light rays that
approximately describe the propagation of one extended light wave in space in the limit of
short wave lengths (geometric optics), quasi-classical times are defined approximately, but
separately for all quasi-trajectories in superspace. Each of them then defines a dynamically
autonomous quasi-classical world (an 'Everett branch' of the global wave function in unitary
description) – including a specific quasi-classical spacetime. As 'Schrödinger cat' states
evolve abundantly from microscopic superpositions in measurement-type interactions, there
cannot be just one quasi-classical world (analogous to just one light ray) according to the
Schrödinger dynamics. Material clocks, such as atomic clocks, require further (usually not
quite as strong) decoherence to become quasi-classical.

5
Literature

Primary
[1] B. S. DeWitt: Quantum theory of gravity, Phys. Rev. 160, 1113 (1967)
[2] W. Pauli: In Handbuch der Physik, ed. by H. Geiger and K. Scheel, 2nd edt., Vol. 24
(Springer, Berlin 1933)
[3] J. von Neumann: Mathematische Grundlagen der Quantenmechanik (Springer, Berlin
1932)
Secondary
[4] J.B. Barbour: The End of Time (Weidenfeld & Nicolson, London 1999)
[5] E. Joos, H. D. Zeh, C. Kiefer, D. Giulini, J. Kupsch, and I.-O. Stamatescu: Decoherence
and the Appearance of a Classical World in Quantum Theory, 2nd edn. (Springer, Berlin,
Heidelberg 2003)
[6] J. Hilgevoord: The uncertainty principle for energy and time, Am. J. Phys. 64, 1451
(1996)
[7] P. Busch: The time energy uncertainty relation, In [8]
[8] J. G. Muga, R. Sals Mayato, and I. L. Egusquiza: Time in Quantum Mechanics (Springer,
Berlin, Heidelberg 2002)
[9] H. D. Zeh: The Physical Basis of the Direction of Time, 5th edn. (Springer, Berlin,
Heidelberg 2007)
[10] G. Hegerfeldt: Causality problems for Fermi's two-atom system, Phys. Rev. Lett. 72, 596
(1994)
[11] D. N. Page and W. K. Wootters: Evolution without evolution: Dynamics described by
stationary observables, Phys. Rev. D27, 2885 (1983)
[12] O. Kübler and H. D. Zeh: Dynamics of quantum correlations, Ann. Phys. (N.Y.) 76, 405
(1973)
[13] R. F. Baierlein, D. H. Sharp, and J.A. Wheeler: Three-dimensional geometry as carrier of
information about time, Phys. Rev. 126, 1864 (1962)
[14] H. Salecker and E.P. Wigner: Quantum limitations of the measurement of space-time
distances, Phys. Rev. 109, 571 (1958)
[15] C. Kiefer: Quantum Gravity, 2nd edn. (Clarendon Press, Oxford 2007)
[16] H. D. Zeh: Emergence of classical time from a universal wave function, Phys. Lett.
A116, 9 (1986)
[17] E. Joos: Why do we observe a classical spacetime? Phys. Lett. A116, 6 (1986)