Foundations of Physics, Vol. 36, No.

3, March 2006 (© 2006)

DOI: 10.1007/s10701-005-9021-0

Time-Reversal, Irreversibility and Arrow of Time

in Quantum Mechanics

M. Castagnino,1 M. Gadella,2 and O. Lombardi3

Received June 28, 2005 / Published online February 15, 2006

The aim of this paper is to analyze time-asymmetric quantum mechanics with

respect of its validity as a non time-reversal invariant, time-asymmetric theory
as well as of its ability to determine an arrow of time.

KEY WORDS: irreversibility; arrow of time; rigged Hilbert spaces.

1. INTRODUCTION

The problem of the irreversibility in classical physics consisted in trying

to find and adequate account of the compatibility between the irrevers-
ible macroscopic evolutions described by thermodynamics and the revers-
ible microscopic evolutions resulting from classical mechanics (for histor-
ical details, see Ref. 1) given by the Newton laws. The solution to this
problem has been traditionally presented in terms of the second law of
thermodynamics. In the beginning of the 20th century, classical mechanics
was replaced by quantum mechanics as the fundamental theory describ-
ing the underlying mechanical level. However, this fact did not affect
the core of the original problem: quantum evolutions, considered inde-
pendently of the measurement processes, turned out to be also revers-
ible as they are governed by time-reversal invariant fundamental equations
like Schrödinger and Dirac equations. In quantum mechanics, dynami-
cal Hamiltonian processes usually lead to reversible evolutions. There are,

1 CONICET—Instituto de Astronomı́a y Fı́sica del Espacio, Casilla de Correos 67, Sucur-

sal 28, 1428, Buenos Aires, Argentina. e-mail: mariocastagnino@citynet.net.ar
2 Departamento de Fı́sica Teórica, Facultad de Ciencias, Universidad de Valladolid, c. Real

de Burgos, s.n., 47011, Valladolid, Spain.

3 CONICET—Instituto de Estudios sobre la Ciencia y la Tecnologı́a, Universidad Nacional

de Quilmes, Rivadavia 2328, 60 Derecha 1034, Buenos Aires, Argentina.

407
0015-9018/06/0300-0407/0 © 2006 Springer Science+Business Media, Inc.
408 Castagnino, Gadella, and Lombardi

nevertheless, processes considered irreversible like resonance phenomena.

The irreversible character of these kind of phenomena is largely accepted
(see Refs. 2, 3). However, there is not an equivalent of the second law
of thermodynamics in quantum mechanics. This showed the need for a
formulation of irreversibility in quantum mechanics. A “time-asymmetric”
formulation of quantum mechanics was proposed by two groups, one lead
by Arno Bohm,(4) and the other by Ilya Prigogine,(5–7) whose contribu-
tions were fundamental for the understanding of intrinsic irreversibility,
that is, the irreversibility due not to the interaction between a system
and its environment, but to the dynamics of the closed system. Tradi-
tionally, the problem of irreversibility has been associated to other two:
non-time-reversal invariance and the existence of an arrow of time.
The aim of the present paper is to analyze the different aspects of
this time-asymmetric formulation of quantum mechanics and to discuss its
validity as a theory that, (i) breaks time-reversal invariance, (ii) accounts
for a theory of intrinsic irreversibility in quantum mechanics, (iii) deter-
mines an arrow of time as a consequence of the theory itself.
This paper is organized as follows: In the next Section, we clarify
what we understand by time-reversal invariance and irreversibility and give
examples of different situations that can arise when we combine both
concepts in all possible manners. We also want to establish which is the
proper relation between irreversibility and time’s arrow and, in particular,
whether these two concepts are equivalent or not. In Sec. 3, we analyze
the ideas of the time-asymmetric formulation of quantum mechanics and
its consequences from the point of view of the irreversibility and the non
time-reversal invariance of the theory. In Sec. 4, we try to get more insight
in the nature of irreversibility in this formulation. Finally, in the last sec-
tion, we intend to clarify the role of the formalism in the formulation of
a quantum mechanical arrow of time.

2. DISENTANGLING CONCEPTS

When the problems of irreversibility and of the arrow of time are

addressed, the main obstacle to be faced is conceptual confusion: the lack
of consensus is primarily due to the fact that different concepts are iden-
tified and different questions are subsumed under the same label. In par-
ticular, the two problems are usually identified, as if irreversibility were
the clue for understanding the origin and the nature of the arrow of time.
In this section we shall clarify the concepts involved in the debate; as the
result, both problems become evidently different.
Time-Reversal, Irreversibility and Arrow of Time 409

2.1. Time-Reversal Invariance and Reversibility

The two central concepts involved in the discussions about what has
been loosely called ‘the problem of the direction of time’ are time-reversal
invariance and irreversibility.

Definition 1. A dynamical equation (law) is time-reversal invariant if it

is invariant under the application of the time-reversal operator T.

The time-reversal operator T performs the transformation t → −t and

reverses certain magnitudes which depend on the particular theory consid-
ered. Nevertheless, the central idea is that T must reverse all the dynami-
cal variables whose definitions in function of t are non-invariant under the
transformation t → −t. As a result, given a time-reversal invariant equa-
tion L, if e(t) is a solution of L, then Te(t) is also a solution.
On the other hand, it is well known that an attractor is defined as a
subset of the phase space toward which a set of evolutions tend for t →
±∞. We can extend this definition by considering a generalized concept of
attractor as a subset of the set of the possible states of a system toward
which a set of evolutions tend for t → ±∞; this concept can be applied
not only to phase spaces, but also to any kind of sets of states. Examples
of generalized attractors are the attractors of classical dynamical systems
(fixed point, limit cycle, fractal, etc.) and any classical or quantum equilib-
rium state. With this characterization, the concept of reversibility can be
defined as follows:

Definition 2. A solution (evolution) e(t) of a dynamical equation is

reversible if it has no generalized attractors, for any representation of e(t).

When the time dependent state e(t) can be represented as an n-uple of

dynamical variables, e(t) = (v1 (t), . . . , vn (t)), reversibility requires that, for
any dynamical variable vi (t), the limit limt→±∞ vi (t) does not exist. In this
case it can be said that the evolution e(t) is reversible if it has no attrac-
tors in phase space.
Independently of the details of these two definitions (for further
details, cf. Albert,(8) ) Arntzenius,(9) it is quite clear that the concepts of
time-reversal invariance and irreversibility are different to the extent that
they apply to different mathematical (physical) entities: whereas time-rever-
sal invariance is a property of dynamical equations and, a fortiori, of the
sets of its solutions, reversibility is a property of a single solution of a
dynamical equation. Furthermore, both properties are not even correlated;
in fact, they can be combined with each other in the four possible cases.
410 Castagnino, Gadella, and Lombardi

These four cases have been analyzed with detail elsewhere(10) and, there-
fore, we just mention and recall them in here.

• Time-reversal invariance and reversibility. Let us consider the har-

monic oscillator with Hamiltonian:

1 2 1 2 2
H = p + k q (1)
2m 2

The dynamical equations are time-reversal invariant as can be easily

shown. As a result, the set of trajectories in phase space is symmet-
ric with respect to the q-axis. The solutions q(t) and p(t) have no
limit for t → ±∞. In other words, each trajectory is reversible since
it is a closed curve in phase space.
• Time-reversal invariance and irreversibility. Let us consider the pen-
dulum with Hamiltonian:

1 2 k2
H = p − cos θ (2)
2m θ 2

Again the dynamical equations are time-reversal invariant since

Tθ = θ. Therefore, the set of trajectories in phase space is sym-
metric with respect to the θ-axis. However, not all the solutions are
2
reversible. In fact, when H = k2 , the evolution is irreversible.(11) For
2 2
H < k2 (oscillating pendulum) and H > k2 (rotating pendulum), the
evolutions are reversible.
• Non time-reversal invariance and reversibility. Let us now consider
the modified oscillator with Hamiltonian:

1 2 1
H = p + K(p)2 q 2 , (3)
2m 2

where K(p) = K+ when p
0, K(p) = K− when p < 0, and

K+ and K− are constants. This means that TK+ = K− . As a conse-
quence, if K+ = K− , the dynamical equations are non time-reversal
invariant. Nevertheless, the continuous solutions q(t) and p(t) have
no limit for t → ±∞: each trajectory is reversible since it is a closed
curve in phase space.
• Non time-reversal invariance and irreversibility. Let us consider a
damped oscillator represented by the following dynamical equation:

q̈ + k 2 q̇ + A2 q = 0 , (4)
Time-Reversal, Irreversibility and Arrow of Time 411

which obviously is non time-reversal invariant. On the other hand,

its solutions have the form q(t) = q0 cos ωt e−γ t . Here cos ωt is the
oscillating factor and e−γ t is the damping factor. As a consequence,
the evolutions are irreversible since they tend to zero for t → ∞.

Up to this point, we have presented the general definitions of time-

reversal invariance and irreversibility. However, since here we are interested
in quantum mechanics, we shall consider the following kind of evolutions:

et = Ut e0 (5)

where e0 and et are vector states. The evolution operator is a unitary oper-
ator Ut = e−iH t , which is an one parameter group on the time param-
eter t with −∞ < t < ∞. Observe that the evolutions of the form et =
Ut e0 are always reversible, since et = e−iH t e0 has no limit for t → ±∞. In
other words, since Ut is a unitary operator, it does not change the angle of
separation (the inner product) or the distance (the square modulus of the
difference) between vectors representing two different states. However, irre-
versible and, therefore, non-unitary evolutions can be obtained from the
original reversible unitary dynamics by the introduction of some sort of
coarse-graining (see a thourogly discussion in Ref. 12).
Once the concepts of time-reversal invariance and reversibility have
been elucidated with precision, the problem of irreversibility can be stated
in a simple way: how to explain irreversible evolutions in terms of time-rever-
sal invariant laws. On the basis of such an elucidation, it also turns out to
be clear that there is no conceptual puzzle in the problem of irreversibility:
in principle, nothing prevents a time-reversal invariant equation from hav-
ing irreversible solutions. However, difficulties arise when we are dealing
with dynamical equations having unitary solutions: as we have seen, since
unitary evolutions are always reversible, it is necessary to go to a differ-
ent level of description in order to obtain irreversibility. This point will be
relevant in the discussions about irreversibility as obtained by means the
“time-asymmetric” formulation of quantum mechanics.

2.2. The Problem of the Arrow of Time

The problem of the arrow of time owes its origin to the intuitive
asymmetry between past and future. We experience the time order of the
world as ‘directed’: if two events are not simultaneous, one of them is
earlier than the other. Moreover, we view our access to past and future
quite differently: we remember the past and predict the future. From this
412 Castagnino, Gadella, and Lombardi

point of view, several arrows of time have been considered in Physics: the
thermodynamical arrow of time (the entropy in isolated systems increases
to the future), the cosmological arrow of time (the universe expands
toward the future), the electromagnetic arrow of time (retarded solutions
of Maxwell equations are selected over advanced ones), the biological
arrow of time (live entities birth and die) and the psicological arrow of
time (the intuitive asymmetry between past and future). The question
on the existence of a “quantum mechanical” arrow of time has been
posed.(4–7)
Of course, the problem of the arrow of time arises when we seek a
physical correlate of the intuitive asymmetry between past and future: do
physical theories pick out a preferred direction of time?
The main difficulty to be encountered in answering this question relies
on our anthropocentric perspective: the difference between past and future
is so deeply rooted in our language and our thoughts that it is very diffi-
cult to shake off these temporally asymmetric assumptions. In fact, tra-
ditional discussions around the problem of the arrow of time are usually
subsumed under the label ‘the problem of the direction of time’, as if we
could find an exclusively physical criterion for singling out the direction of
time, identified with what we call ‘the future’. However, there is nothing in
physical evolution laws that distinguishes, in a non-arbitrary way, between
past and future as we conceive them in our ordinary language and our
everyday life. It might be objected that physics implicitly assumes this dis-
tinction with the use of temporally asymmetric expressions, like ‘future
light cone’, ‘initial conditions’, ‘increasing time’, and so on. However this
is not the case, and the reason relies on the distinction between conven-
tional and substantial.

Definition 3. Two objects are formally identical when there is a permu-

tation that interchanges the objects but does not change the properties of
the system to which they belong.

In physics it is usual to work with formally identical objects: the two

semicones of a light cone, the two spin senses, etc.

Definition 4. We shall say that we establish a conventional difference

between two objects when we call two formally identical objects with two
different names.

This is the case when we assign different signs to the two spin senses,
or different names to the two light semicones, etc.
Time-Reversal, Irreversibility and Arrow of Time 413

Definition 5. We shall say that the difference between two objects is

substantial when we assign different names to two objects that are not
formally identical (see Refs. 13, 14). In this case, although the particular
names we choose are conventional, the difference is substantial.

Once this point is accepted, it turns to be clear that fundamental

physics uses the labels ‘past’ and ‘future’ in a conventional way. Therefore,
the problem cannot yet be posed in terms of singling out the future direc-
tion of time: the problem of the arrow of time becomes the problem of
finding a substantial difference between the two temporal directions. It seems
necessary to address the problem of the arrow of time from a perspective
purged of our temporal intuitions. Thus, we must avoid the conclusions
derived from subtly presupposing time-asymmetric notions. As claimed by
Price,(15) it is necessary to stand at a point outside of time, and thence to
regard reality in atemporal terms: this is the so called ‘view from nowhen’.
But then, what does ‘the arrow of time’ mean when we accept this
constraint? We recognize the difference between the head and the tail of
an arrow on the basis of its geometrical properties; therefore, we can
substantially distinguish between both directions, head-to-tail and tail-to-
head, independently of our particular perspective. Analogously, the prob-
lem of the arrow of time should be conceived in terms of the possibil-
ity of establishing a substantial distinction between the two directions of
time exclusively by means of arguments based on theoretical physics (for a
detailed discussion, see Refs. 16, 17).
If the problem is expressed in this way, the question is: why is time-
reversal invariance an obstacle to solve the problem of the arrow of time?
In order to answer this question we have to recall that, if et is a solu-
tion of a time-reversal invariant law L, then Tet is also a solution of L.
We shall call these two solutions ‘time-symmetric twins’: they are twins
because, in the context of the theory to which they belong and with-
out presupposing a privileged direction of time, they are only conven-
tionally different; they are time-symmetric because one is the temporal
mirror image of the other. The traditional example of time-symmetric
twins is given by electromagnetism, where dynamical equations always
have advanced and retarded solutions. With this terminology we can say
that a time-reversal invariant theory always produce time-symmetric twins:
the obstacle to solve the problem of the arrow of time in this case relies
on the fact that, in the context of the theory, the twins are only conven-
tionally different.
The traditional arguments for discarding one of the twins and retain-
ing the other invoke time-asymmetric notions which are not justified in
414 Castagnino, Gadella, and Lombardi

the context of the theory. For instance, the retarded nature of radiation is
usually explained by means of de facto arguments referred to initial con-
ditions: advanced solutions of wave equations correspond to converging
waves that require a miraculous cooperative emitting behavior of distant
regions of space at the temporal origin of the process. A different but
related argument is put forward by those who appeal to the impossibility
(or high difficulty) of preparing time-reversed states in laboratory experi-
ments like, for instance, experiments of scattering.(3) It seems quite clear
that this kind of arguments, not based on theoretical considerations, are
not legitimate in the context of the problem of the arrow of time to the
extent that they put the arrow ‘by hand’ by presupposing the difference
between the two directions of time from the beginning. In other words,
they violate the ‘nowhen’ requirement of adopting an atemporal perspec-
tive purged of temporal intuitions like those related with the asymmetry
between past and future or between initial and final conditions. Therefore,
from an atemporal standpoint, the challenge consists in supplying a non-
conventional criterion, based on theoretical arguments, for choosing one
of the time-symmetric twins as the physically meaningful one: such a crite-
rion will establish a substantial difference between the two members of the
pair and, a fortiori, between the two directions of time (for a discussion of
the concept of time-symmetric twins, see Ref. 18). This is the conceptual
background we need in the discussion on whether the “time-asymmetric”
quantum mechanics gives a solution to the problem of the arrow of time.

3. TIME-REVERSAL INVARIANCE AND IRREVERSIBILITY

IN THE TIME-ASYMMETRIC VERSION OF QUANTUM
MECHANICS

The purpose of the so called time-asymmetric quantum mechanics is

to provide a formulation of quantum mechanics capable of describing irre-
versible quantum phenomena. This formulation makes use of the rigged
Hilbert space formulation of quantum mechanics. This formulation is well
known(19–22) as well is its application to resonance scattering,(21) therefore,
here we briefly summarize it in order to recall its main ideas and introduce
the notation we shall use along the present discussion.
Let us recall that we can construct two rigged Hilbert spaces

± ⊂ H ⊂
×
±, (6)

such that there exists corresponding unitary representations V± such

that(21)
Time-Reversal, Irreversibility and Arrow of Time 415

V± H = L2 (R+ ) (7)


2

V±
± = S ∩ H±
, (8)
R+

where S is the Schwartz space of infinitely differentiable functions vanish-

ing at infinite faster than the inverse of any polynomial, H± are the spaces
of Hardy functions on the upper (+) and lower (−) halves of the complex
plane, and the space L2 (R+ ) is the Hilbert space of the square

integrable
Lebesgue functions on the positive real axis. The symbol
R+ means that
we consider the restriction of the functions in S ∩H±2 on the real axis. The
×
spaces
± are the respective antiduals (antilinear continuous functionals)
of the locally convex spaces
± .(21) The unitary operators V± diagonalize
the absolutely continuous parts of the total Hamiltonian H in the sense
that V± H V±−1 is the multiplication operator on L2 (R+ ) (for simplicity, we
assume that the spectrum of H is nondegenerate in this sector).(21) The
vector φ± ∈
± represents the same state than the wave function in the

representation, given by φ± (ω) = ω|φ±  := V± φ± , φ± (ω) ∈ S ∩
energy
2
.
H± R+
Let U−t := eitH be the adjoint of the evolution operator Ut . Then,
one can readily show that(21)

U−t
− ⊂
− , if t  0 (9)

U−t
+ ⊂
+ , if t
0 . (10)

Now, let
⊂ H ⊂
× be a rigged Hilbert space, and U an operator

on H such that U †
⊂
, where U † is the adjoint of U . Then, we can
extend U into
× by means of the following duality formula(21)

U † φ|F  = φ|U × F  , ∀ φ ∈
, ∀ F ∈
× , (11)

where U × is the extension of U to
× . Using (9, 10, 11), we conclude that

Ut−×
× ×
− ⊂
− , t  0, (12)

Ut+×
× ×
+ ⊂
+ , t
0. (13)

Here, we have introduced the signs ± in the extensions Ut±× in order

to distinguish these two extensions, which are defined on different spaces
(
×± ).
416 Castagnino, Gadella, and Lombardi

Due to the construction of the triplets (6), the antiduals
×

± contain
two kinds of functionals interesting from the point of view of physics: the
Dirac kets (eigenfunctionals of H with eigenvalues in the continuous spec-
trum of H ) and the Gamow vectors. The latter are eigenfunctionals of
H , whose respective eigenvalues coincide with the resonance poles for the
Hamiltonian pair (H0 , H ).(21) Resonance poles appear as complex conju-
gate pairs zR = ωR − i 2 and zR ∗ = ω + i  , with  > 0. The model has
2
been constructed such that H
± ⊂
± and, therefore, H can be extended
to the antiduals
×± using the duality formula (11). If we call the exten-
sions of H into
× × × ×
− and
+ , H− and H+ , respectively, the existence of
two functionals  , the growing Gamow vector, and  D , the decaying Ga-
G

mow vector, can be proved such that

 
× D 
H+  = zR  = ωR − i
D
D (14)
2
 

H−×  G = zR
∗
 G = ωR + i G . (15)
2

In addition, a semigroup time evolution can be defined on these ei-

genfunctionals via the duality formula (11).(21) This gives:

U−t φ− | G  = φ− |Ut−×  G  = e−iωR t et/2 φ− | G  , (16)

∀ φ− ∈ − , t  0.

U−t φ+ | D  = φ+ |Ut+×  D  = e−iωR t e−t/2 φ+ | D  , (17)

∀ φ+ ∈ + , t
0.
Observe that (17) represents an exponentially decaying process with
lifetime τ = 2 , whose limit when t goes to infinity results:

lim φ+ |Ut+×  D  = lim φ+ | D  e−iωR t e− 2 t = 0


(18)
t→∞ t→∞

This means that, for t → ∞, the decaying Gamow vector  D exponen-

tially decays in a weak sense. Analogously, the growing Gamow vector  G
exponentially decays in a weak sense for t → −∞.
One of the purposes of these ideas is to obtain a formulation of
quantum mechanics capable of explaining irreversible quantum phenom-
ena; according to this view, the use of rigged Hilbert spaces is what
turns standard quantum mechanics into a time-asymmetric theory where
irreversible quantum descriptions are possible. However, we have seen
Time-Reversal, Irreversibility and Arrow of Time 417

that non-time-reversal invariance and irreversibility are different concepts;

therefore, it is worth while to ask how and by means of which formal
resources the new formalism accounts for these two different features.
In the many works on “time-asymmetric” quantum mechanics,(4–7) it
is suggested that, in the rigged Hilbert space formalism, the fact that evo-
lutions are described by means of semigroups rather than groups is what
permits irreversible behavior to be modeled in a natural way. In fact, irre-
versibility is introduced by the fact that processes that exponentially decay
(grow) as e− 2 t (e 2 t ) can be obtained: they have a well defined limit for
 

t → ∞ (t → −∞). This is a direct consequence of the choice of the spaces

± constructed in (8) and the properties of the Hardy functions.
An alternative description of resonances by Gamow vectors could
have been given as follows: Let V any unitary operator such that it diago-
nalizes the total Hamiltonian H in the sense that V H V −1 be the multipli-
cation operator in L2 (R+ ). Let us define a rigged Hilbert space as follows:
consider the space D(R) of all infinitely differentiable functions with com-
pact support on the real line R, and the Fourier transform of this space
F(D(R)). Each function on F(D(R)) is entire analytic and, considered as
complex functions on the real line, Schwartz functions.




Proposition. The space of restrictions F(D(R))
+ to the positive
R
semiaxis R+ of the functions in F(D(R)) is dense in L2 (R+ ).

Proof. Let us consider the space D(R+ ) of infinitely differentiable

functions with compact support in the positive semiaxis R+ . This is a
subset of L2 (R+ ) and, therefore, the Paley Wienner theorem(23) asserts
that the space F(D(R+ )) of their Fourier transforms is dense in the space
H−2 of Hardy functions in the lower half plane. This means that for each

ϕ− (ω) ∈ H−2 and each ε > 0, there exists an f (ω) ∈ F(D(R+ )) such that

 ∞
|ϕ− (ω) − f (ω)|2 dω < ε . (19)
−∞

This obviously implies that

 ∞
|ϕ− (ω) − f (ω)|2 dω < ε , (20)
0




which means that the space F(D(R+ ))
of the restrictions of functions
R+


to the positive semiaxis R+ 2
. As the space H2

is dense in H−
+ −
+ is
R R
418 Castagnino, Gadella, and Lombardi

dense in L2 (R+ ), the space F(D(R+ ))
+ is dense in L2 (R+ ) (with the
R
topology on the latter). Since,






F(D(R+ ))
+ ⊂ F(D(R))
+ ⊂ L2 (R+ ) , (21)
R R




we conclude that F(D(R))
+ is dense in L2 (R+ ). Note that any function
R
in F(D(R)) is entire analytic and, therefore, is uniquely determined by its
values on the positive semiaxis.




Let
:= V −1 F(D(R))
+ and endow
with the topology derived
R
from the topology on D(R) exactly as we derived the topology on
±
from the topology on the Schwartz space.(21) Then,
⊂ H ⊂
× is a
rigged Hilbert space such that there exist two functionals  G ,  D ∈
×
with the following properties:
(i) Both are eigenvectors of the extension H × of the total Hamilto-
nian H into
× ,

H ×  G = zR
∗
 G; H ×  D = zR  D , (22)

where zR and zR ∗ are the pair of complex conjugate poles (of the

S-matrix or the reduced resolvent, see Refs. 21, 24) that determine
a resonance for the Hamiltonian pair (H0 , H ).
(ii) The space
is left invariant under the action of the whole group
Ut = e−itH and therefore this group can be extended to the dual

× . As a consequence, if we call this extension Ut× , the function-
als evolve for all values of t as

Ut×  G = e−iωR t et/2  G and Ut×  D = e−iωR t e−t/2  D . (23)

We do not include the proofs of these assertions as they are a rep-

etition of the proofs for the corresponding results on
×± in Ref.
21. The conclusion of these results is very simple: the very use
of rigged Hilbert spaces in the description of resonance scatter-
ing does not lead by itself to a semigroup description. This semi-
group description is the consequence of the choice of the spaces

± , constructed via Hardy functions.
On the other hand, an evolution is irreversible if it has a limit for t →
±∞. In the formalism under discussion, irreversibility is introduced by the
fact that processes that exponentially decay (grow) as e− 2 t (e 2 t ) can be
 

obtained: they have a well defined limit for t → ∞ (t → −∞) (Observe

Time-Reversal, Irreversibility and Arrow of Time 419

that this is true even if we use the rigged Hilbert space defined as in the
last paragraph, instead of
± ⊂ H ⊂
× ± with
± as in (8)). It is worthy
to mention that the existence of the functionals  G and  D is a conse-
quence of the use of analytically continuable functions, a certain fact in
all the choices of rigged Hilbert spaces here discussed.
In the case that we use the description based in the rigged Hilbert
spaces in where the basic (or test vector) spaces are defined by the
± as
in (8), one can readily show(25) that if T is the time-reversal operator on
the Hilbert space H, we have that

T
± =
∓ , (24)

and therefore,
± are time-symmetric twins in the sense of the previous

section. Furthermore, with the aid of the duality formula (11), we can
extend T to the antiduals. If, for simplicity, we also call T to the exten-
sions of the time-reversal operator to the antiduals, we obtain:

T
× ×
± =
∓ , (25)

which shows that the rigged Hilbert spaces
+ ⊂ H ⊂
× + and
− ⊂

H ⊂
× − are time-reversed of each other. They are time-symmetric twins.
In addition, we want to remark that T  G =  D and T  D =  G , so that
both Gamow vectors are time-symmetric of each other.(25)

4. THE COARSE-GRAINED NATURE OF IRREVERSIBILITY

In the rigged Hilbert space description of resonances, the Gamow vec-

tors have the following property:

lim  G (t) = lim e−iωR t et/2  G = 0 (26)

t→−∞ t→−∞

lim  D (t) = lim e−iωR t e−t/2  D = 0 , (27)

t→∞ t→∞

where  G =  G (0) and  D =  D (0). This result is true no matter which

rigged Hilbert spaces, from the above described, we use. The real mean-
ing of Eqs. (26) and (27) is obtained by taking the corresponding limits
in (16) and (17) (and their obvious implementation for the later rigging).
This means that the equations (26) and (27) are not correct from a math-
ematical point of view: what decays as t goes to infinity (minus infinity)
420 Castagnino, Gadella, and Lombardi

is not  D (t) ( G (t)) but instead φ+ | D (t) (φ− | G (t)). As a conse-
quence, if we want to conceive  D (t) as a generalized state,4 we only can
strictly say that the expectation value of the observable A = | φ+ φ+ | in
the state  D (t) decays exponentially:

A D (t) = |φ+ | D (t)|2 = |φ+ | D |2 e−t . (28)

Observe that (28) is well defined for t > 0, as φ+ ∈
+ .

In other words, whereas A D (t) exponentially decays as t goes to
infinity:

lim A D (t) = 0 , (29)

t→∞

the generalized state  D (t) has only a weak limit:

w − lim  D (t) = 0 . (30)

t→∞

As we have argued elsewhere,(12) this weak limit means that the general-
ized state  D (t) decays from an observational point of view, that is from the
perspective given by the observable A = | φ + φ + |, for any φ + ∈
+ . In this
sense, A D involves a generalized coarse-graining, that is, a generalized
projection of the vector  D (t) onto a subspace defined by the operator A.
In fact, since A2 = A, the observable A can be conceived as a projector :

A = | φ+ φ+ | =  (31)

D as:
Then, we can define a coarse-grained state cg

D
cg =   D = | φ+ φ+ | D  (32)

With this definition:

|cg
D
cg
D
| = | φ+ φ+ | D  D |φ+ φ+ |

= |φ+ | D |2 |φ+ φ+ | = A D |φ+ φ+ | . (33)

This means that the expectation value of the observable A = | φ+ φ+ | in

the state  D can be viewed as the result of the action of the projector

4 This may have some difficulties, see for instance Refs. 26, 27. For other arguments, see
Ref. 28.
Time-Reversal, Irreversibility and Arrow of Time 421

 = A on the vector  D . On this basis we can understand why A D (t)

is a coarse-grained magnitude: strictly speaking, this coarse-grained mag-
nitude is what decays for t → ∞, and not the generalized state  D (t) as
Eq. (27) seems to suggest. A similar discussion is possible for  G and time
going to minus infinite.
The conclusion of this argument is that, in the case of the evolution
described by Gamow vectors, the coarse-grained magnitude that decays as
t goes to infinity is the expectation value of the observable A = | φ+ φ+ |
in the generalized state  D , for any φ+ ∈
+ , and there is no quantum
law that prevents it from having this kind of behavior. But to interpret the
decaying Gamow vector as a truly decaying state can only be the result of
a philosophically biased rejection of coarse-graining, which misdirects the
interpretation of the irreversible processes described by the theory.

5. THE PROBLEM OF THE ARROW OF TIME

IN TIME-ASYMMETRIC QUANTUM MECHANICS

As we have seen, two rigged Hilbert spaces arise as a consequence of the

use of Hardy functions in the particular realization of the space
, which
determines the properties of the antidual
× . Thus, two antiduals arise,
×±,
which contain not only all the physically realizable states, but also general-
ized states like the Gamow vectors and the Dirac kets.(21) Time evolution
has a semigroup structure on the antiduals: the semigroups Ut+× with t
0
and Ut−× with t  0, govern the time evolutions on
× ×
+ and
− , respec-
+× −×
tively. We have also shown that the fact that Ut and Ut form semigroups
is what would break down the original time-reversal invariance of quantum
mechanics in its standard version. On the basis of these previous results,
we have seen that those two rigged Hilbert spaces with their corresponding
evolution operators lead to pairs of time-symmetric twins, since they are
two non time-reversal invariant formalisms, one the temporal mirror image
of the other. If we want to distinguish between the two directions of time,
t
0 and t  0, the challenge consists in supplying a non-conventional
criterion, based on theoretical considerations, for choosing one of the twins
of each pair as the physically meaningful one.
In their analysis of the Friedrich’s model for quantum scattering, An-
toniou and Prigogine(6) adopt the following interpretation: Ut+× carries
states into the forward direction of time and then, describes evolutions
reaching equilibrium in the future; Ut−× carries states into the backward
direction of time and then, describes evolutions reaching equilibrium in the
past. Even if with a different terminology, the authors are admitting that
these two evolutions are only conventionally different to the extent that
422 Castagnino, Gadella, and Lombardi

the theory by itself gives no basis for selecting one of the elements of
the pair as the physically relevant. Therefore, they adopt an observational
criterion for retaining one of the semigroups and discarding the other:
since no physical system has ever been observed evolving to equilibrium
toward the past, the physically relevant semigroup of evolution operators
is the semigroup corresponding to Ut+× , valid for t
0.(6,7) Although this
appeal to observational considerations is a legitimate move in the every-
day work of physicists, it is not acceptable when the problem at issue is
to explain the arrow of time, since the fact that our observations are time-
directed was known from the very beginning: the real problem consists in
accounting for the difference between the two directions of time by means
of theoretical arguments.
Bohm’s response to the problem of the arrow of time is subtler than
the solution proposed by Prigogine and his coworkers. In his detailed
description of scattering processes, Bohm breaks the symmetry between
the two twins by appealing to the so-called ‘preparation-registration arrow
of time’, expressed by the slogan ‘no registration before preparation’.(4)5
On this basis, Bohm replaces the representational postulate of standard
quantum mechanics—according to which states are represented by the vec-
tors of a separable Hilbert space and observables are represented by self-
adjoint operators on that space—by a new one that distinguishes between
the mathematical descriptions of states ψ and of observables ϕ:

{ψ} ≡
− ⊂ H ⊂
×
− (34)

{ϕ} ≡
+ ⊂ H ⊂
×
+ (35)

In particular, the new representational postulate asserts that the vec-

tors |ψ ∈
− represent the states of the system and the vectors |ϕ ∈

+ represent the observables of the system in the sense that a state is
ρ = |ψψ| and an observable is A = |ϕϕ|. States and observables must
undergo a time evolution as states and observables. In this sense, time evo-
lution for an observable given by ψ ∈
− must be of the form:

ρ(t) = e−itH ρ(0)eitH = e−itH |ψψ|eitH . (36)

This equation makes sense if and only if t
0, since eitH ψ ∈
− for each
ψ ∈
− if and only if t  0. The evolution for observables must be of the
form

5 Bohm acknowledges that the origin of the idea of a preparation-registration arrow can
be traced back to the works of Günther Ludwig.(29)
Time-Reversal, Irreversibility and Arrow of Time 423

A(t) = eitH A(0)e−itH = eitH |ϕϕ|e−itH . (37)

Observe, that, since eitH ϕ ∈
+ for each ϕ ∈
+ if and only if t
0, Eq.

(37) makes sense if and only if t
0. Both, states and observables evolve
now forward in time and therefore, a privileged direction of time appears.
As we can see, Bohm’s approach breaks the symmetry between the
time-symmetric twins arising from
− ⊂ H ⊂
× − and
+ ⊂ H ⊂
+
×

by means of a representational postulate based on the preparation–regis-

tration arrow. However, this preparation registration arrow puts the arrow
of time ‘by hand’ by adding a particular temporal relationship to the
ontological priority of states with respect to observables. In fact, without
the time-asymmetric intuition introduced by the preparation–registration
arrow, nothing prevents us from reversing the representational postulate by
stating that
− ⊂ H ⊂
× − is the rigged Hilbert space for the representa-
tion of observables and
+ ⊂ H ⊂
× + is the rigged Hilbert space for the
representation of states. In this case, we would obtain the temporal mirror
image of Bohm’s theory. This new representational postulate restores the
symmetry because now we have two postulates leading to two non time-
reversal invariant theories, one the temporal mirror image of the other:
they also lead to time-symmetric twins and the challenge consists, again,
in supplying a theoretical and non-conventional criterion for retaining one
of them and discarding the other. Bohm’s decision of selecting the future
directed alternative is based on presupposing the arrow of time from the
very beginning on the basis of pretheoretical intuitions.
In conclusion, the proposal of time-asymmetric quantum mechanics,
either in Prigogine’s or in Bohm’s version, does not supply an acceptable
answer to the problem of the arrow of time. We have argued elsewhere(16,18)
that this conclusion is not surprising to the extent that the problem can-
not be solved in local terms: the substantial distinction between the two
directions of time requires a global perspective that permits the irreversible
processes of the universe to be correlated in such a way that all of them
parallely decay in the same temporal direction. But the detailed discussion
of this point goes beyond the limits of the present paper.

6. CONCLUSIONS

Although in principle nothing prevents a time-reversal invariant the-

ory from describing irreversible evolutions, in the case of dynamical equa-
tions with unitary solutions, time-reversal invariance and reversibility seem
to go hand-in-hand. Nevertheless, even in this case both properties are
different to the extent that they are related with distinct features of the
424 Castagnino, Gadella, and Lombardi

formalism: whereas time-reversal invariance implies the group structure of

the evolution operators, reversibility is a consequence of the unitary char-
acter of such operators. Therefore, even if the time-reversal invariance of
a theory is broken down by means of semigroup evolution laws, this fact
does not affect the reversible character of the evolutions if they are still
described by unitary operators. The only way to extract irreversibility from
unitary processes is by means of some mathematical procedure that leads
to a level of description different from the original dynamical level, where
non-unitary evolutions can be obtained. This is precisely the effect of
rigged Hilbert space formulation: as we extend evolutions to the antiduals
of a rigged Hilbert space (in order to define the time evolution for gener-
alized states like Gamow vectors), the concept of unitarity is lost.
Thus, the conclusions of this study can be summarized as follows:
• The non time-reversal invariance of the theory is due to the semi-
group structure of the evolution laws which, in turn, is a conse-
quence of the use of a particular realization of the rigged Hilbert
space based on Hardy functions.
• The irreversibility of the evolutions is due to the existence of Ga-
mow vectors, which do not directly depend on the choice of rigged
Hilbert spaces using Hardy functions. The use of functions with
adequate analytical properties will give equally well Gamow vectors
with the correct time evolution.
• Although time-asymmetric quantum mechanics is a powerful theory
for the description of intrinsic irreversibility, the formalism does not
determine an arrow of time free of preconceptions.

ACKNOWLEDGMENTS

This paper is written to congratulate Prof. Emilio Santos on the occa-

sion of his 70th bithday. It was partially supported by CONICET, the
Universidad de Buenos Aires, the Universidad Nacional de Quilmes, the
Secretarı́a de Estado de Educación y Universidades of Spain, the Junta
de Castilla y León, Project VA 085/02, and the FEDER-Spanish Minis-
try of Science and Technology Projects DGI BMF 2002-0200 and DGI
BMF2002-3773.

REFERENCES

1. S. Brush, The Kind of Motion We Call Heat (North Holland, Amsterdam, 1976).
2. C. Cohen-Tannoudji, B. Diou, F. Laloe, Quantum Mechanics (Wiley, New York, 1977).
Time-Reversal, Irreversibility and Arrow of Time 425

3. T. D. Lee, Particle Physics and Field Theory (Harwood, New York, 1981).
4. A. Bohm and R. Scurek, “The phenomenological preparation–registration arrow of
time and its semigroup representation in the RHS Quantum Theory” in Trends in
Quantum Mechanics, H. D. Doebner, S. T. Ali, M. Keyl, and R. F. Werner eds.,
(World Scientific, Singapore-London, 2000); A. Bohm, M. Loewe, and B. Van de
Ven, “Time asymmetric quantum theory—I. Modifying an axiom of quantum physics”
Fortschr. Phys. 51, 551 (2003); A. Bohm, I. Antoniou, and P. Kielanowski, “A quan-
tum mechanical arrow of time and the semigroup time evolution of Gamow vectors”
J. Math. Phys. 36, 2593 (1994); A. Bohm, M. Gadella, and M. J. Mithaiwala, “Time
asymmetric Quantum Theory: foundations and applications” in The Physics of Com-
munication, Proceedings of the XXII Solvay Conference on Physics (World Scientific,
Singapore, 2003), p. 117; R. C. Bishop, A. Bohm, and M. Gadella, “Irreversibility in
Quantum Mechanics” Discrete Dynamics in Nature and Society 2004, 75–83 (2004).
5. I. Prigogine, From Being to Becoming. Time and Complexity in the Physical Sciences
(Freeman, New York, 1980); T. Petroski and I. Prigogine, “Alternative formulation
of classical and quantum dynamics for non-integrable systems” Physica A 175, 146
(1991); T. Petroski, I. Prigogine, and S. Tasaki, “Quantum theory of non-integrable
systems” Physica A 173, 175 (1991).
6. I. Antoniou and I. Prigogine, “Intrinsic irreversibility and integrability of dynamics”
Physica A 192, 443 (1993).
7. T. Petrosky and I. Prigogine, “The extension of classical dynamics for unstable Ham-
iltonian systems” Comp. Math. Appl. 34, 1 (1997).
8. D. Albert, Time and Chance (Harvard, Cambridge (USA), 2000).
9. F. Arntzenius, “Time reversal operations, representations of the Lorentz group, and
the direction of time” Stud. Hist. Phil. Mod. Phys 35, 31 (2004).
10. M. Castagnino, L. Lara, and O. Lombardi, “The cosmological origin of time asym-
metry” Class. Quant. Grav. 20, 369 (2003).
11. M. Tabor, Chaos and Integrability in Nonlinear Dynamics (Wiley, New York, 1989).
12. M. Castagnino and O. Lombardi, “Self-induced decoherence: a new approach” Stud.
Hist. Phil. Mod. Phys. 35, 73 (2004).
13. R. Penrose, “Singularities and time asymmetry” in General Relativity, an Einstein Cen-
tenary Survey, S. Hawking and W. Israel, eds., (Cambridge, Cambridge, 1979).
14. R. G. Sachs, The Physics of Time-Reversal (University of Chicago, Chicago, 1987).
15. H. Price, Time’s Arrow and the Archimedes’ Point (Oxford, Oxford, 1996).
16. M. Castagnino, L. Lara, and O. Lombardi, “The direction of time: from the global
arrow to the local arrow” Int. J. Theor. Phys. 42, 2487 (2003).
17. M. Castagnino, O. Lombardi, and L. Lara, “The global arrow of time as a geomet-
rical property of the Universe” Found. Phys. 33, 877 (2003).
18. M. Castagnino and O. Lombardi, “The generic nature of the global and non-entropic
arrow of time and the dual role of the energy–momentum tensor” J. Phys. A: Math.
and Gen. 37, 4445 (2004); M. Castagnino and O. Lombardi, “The global non-entro-
pic arrow of time: from global geometrical asymmetry to local energy flow” Synthese
forthcoming.
19. J. P. Antoine, “Quantum mechanics beyond Hilbert space” in Irreversibility and Cau-
sality, A. Bohm, H-D Doebner and P. Kielanowski, eds., Springer Lecture Notes in
Physics, Vol. 504 (Springer, Berlin and New York, 1998), pp 3–33.
20. A. Bohm, Quantum Mechanics: Fundations and Applications (Springer, Berlin and New
York, 1994).
21. A. Bohm and M. Gadella, Dirac Kets, Gamow Vectors and Gelfand Triplets, Springer
Lecture Notes in Physics, Vol. 348 (Springer, New York, 1989); A. Bohm, M. Gadella,
426 Castagnino, Gadella, and Lombardi

and S. Maxon, “Extending the stationary quantum mechanics of being to a nonsta-

tionary quantum theory of becoming and decaying”, Comp. Math. Appl. 34, 427–466
(1997).
22. M. Gadella and F. Gómez, “A unified mathematical formalism for the Dirac formula-
tion of Quantum Mechanics” Found. Phys. 32, 815 (2002); M. Gadella and F. Gómez,
“On the mathematical basis of the Dirac formulation of Quantum Mechanics” Int.
J. Theor. Phys. 42, 2225–2254 (2003).
23. R. Paley and N. Wiener, Fourier Transforms in the Complex Domain (American Math-
ematical Society, New York, 1934).
24. I. E. Antoniou, L. Dmitrieva, Yu. Kuperin, and Yu. Melnikov, “Resonances and the
extension of dynamics to rigged Hilbert space” Comp. Math. Appl. 34, 339 (1997).
25. M. Gadella and R. de la Madrid, “Resonances and time reversal operator in Quan-
tum mechanics” Int. J. Theor. Phys. 38, 93 (1999).
26. M. Gadella and R. Laura, “Gamow dyads and expectation values” Int. J. Quant.
Chem. 81, 307–320 (2001).
27. M. Castagnino, M. Gadella, R. Id Betan, and R. Laura, “Gamow functionals on
operator algebras” J. Phys. A: Math. Gen. 34, 10067 (2001).
28. O. Civitarese, M. Gadella, and R. Id Betan, “On the mean value of the energy for
resonant states” Nucl. Phys. A 660, 255 (1999).
29. G. Ludwig, Foundations of Quantum Mechanics, Vol. I and II, (Springer, Berlin, 1983–
1985).