Beruflich Dokumente
Kultur Dokumente
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Oxford University Press and The British Society for the Philosophy of Science are collaborating with JSTOR to
digitize, preserve and extend access to The British Journal for the Philosophy of Science.
http://www.jstor.org
501-518
ABSTRACT
Introduction
Classesof IrreversibleProcesses
IrreversibilityDefined
T-invarianttheories
TheTemporalReferenceDirection
Irreversibilityhas no SpatialAnalogue
Does IrreversibilityNecessarilyInvolveEntropy?
FanningOut TowardsTheFuture
INTRODUCTION
In the Timaeus,
Platospokeof timeas 'revolving'andit may be that he believed
that 'timeitself'is cyclicin some sense. Evenso he didnot supposethat most
To be sure 'lifecycle'
sequencesof eventsare anythingotherthan irreversible.
is a commonlyused expression;but clearlyit does not mean either that the
eventsof a person'slifecan occuragain in the reverseorder,or that a person
can be reborn and then lead exactly the same life over again. Plato's view, if I
have understood it correctly, was that there are cyclic motions in the heavens
but irreversible sequences on Earth.
So too in Asiatic cultures, even in those which have been widely regarded as
K. G. Denbigh
502
having adopted the notion of cyclic time. For example Schipper and Wang
Hsiu-Huei [1987] have pointed out that, although Taoist ritual used a concept
of 'cycles' nested within each other, this nesting was in a 'time' which, in itself,
was taken as linear. Similarly in Indian thought; Anindita Balslev [19871
recently pointed out that the supposed recurrence of 'world cycles' does not
involve 'any idea of exact repetition of the particular, and that instead the
emphasis is on the similarity of the generic features'.
As we know, Judeo-Christian thought uses the notion of a linear and
progressivetime. But 'progressive'in what sense? Science has been widely seen
as indicating a universe-wide process of 'running down', an approach to the
'heat death'. However, I have argued elsewhere [Denbigh, 1989] that the
entropy law is a good deal less restrictive than is commonly supposed.
Something quite distinct from a running down may also be taking place if we
can but get a lead on it.
Towards this end it is useful to give consideration to the concept of
irreversibility.This has a much wider field of application than has the concept
of entropy increase, as may be seen from the fact that there exists a wide
variety of processes which are undoubtedly irreversiblewhilst seemingly not
giving rise to any closely correlated entropy change.
Thus the object of this paper is not at all concerned with the current theories
either of entropy increase or of 'chaos', but rather to consider 'one way
temporal development in a much broader context, one which will include the
irreversibilityof biological and mental processes. This essay will therefore not
attempt the mathematical sophistication of Harold Grad'sfamous paper 'The
Many Faces of Entropy' [1961] but instead will be entirely qualitative and
phenomenological.
Beforeproceeding let me repeat the truism that one cannot talk about either
time or irreversibilitywithout using temporal words. If one were to say, for
example, that some particular sequence of events does not occur in the reverse
order, the understanding of 'sequence', 'events' and 'occur' depends on a prior
acceptance of certain temporal presuppositionswhich are deeply embeddedin
language. Even the using of the present tense in a 'tenseless' (i.e. timelessly
true) manner does not always eliminate the presupposing of 'time's arrow'.
The same applies to the use of many substantives. For instance to speak of 'an
expansion' or of a 'light source' is tacitly to adopt a particulardirection of time
and not its reverse. The question how that direction is chosen will be deferred
to Section 5.
2
CLASSES OF IRREVERSIBLE
PROCESSES
The class which comes first to the thoughts of most scientists is that group of
processes, occurring in physico-chemical systems, which may be called the
class.Typicalexamples are the flow of heat from hotter to cooler
thermodynamic
503
504
K. G. Denbigh
to papers by Hill and Grtinbaum [195 7] and by Penrose and Percival [1962].
For my present purposes it is sufficient to notice that there are physical
phenomena which are irreversible without necessarily being entropic.
Another very important kind of irreversibilityis displayedin biology where it
is familiar enough that the evolutions of the various species of organisms do
not normally occur in reverse. The featheredbirdsdo not return to being scaly
reptiles, nor do the reptiles revert to their own parent genera. To be sure there
are individual instances of regression2 or of simplificationof function. Simpson
[1950] pointed out that the notion that evolution is invariably accompanied
by increase of 'complexity' is very difficult to substantiate. What does seem
certain, however, is that, in the temporal direction we call 'future', there
occurs a branching-what Darwin called a divergence.Throughout the period
since life first appeared on Earth new species have branched off from existing
species,3 with the consequence that the overall evolutionary scheme has a
temporal structure resembling the above-ground structure of a tree. Biologists
would find it quite unacceptable, I think, to suppose that at some future date
this structure would start to regress, resulting in all the 'advanced' organisms
returning to their ancestral states, and leading eventually to all life being
in the form of unicellular organisms, before these too vanish into a lifeless
Earth.
A somewhat similar 'branching towards the future' occurs in each
individual organism. Ontogeny, it was said by Ernest Haeckel, recapitulates
phylogeny! No doubt this is an oversimplification,but nevertheless it is true, in
the present context, that the bodily development of each individual traces out
an irreversiblepath, just as does the development of the biosphere as a whole.
The cells of an embryo have the capacity to divide and to differentiate,giving
rise to a large number of differentsorts of cells which go to form the tissues and
organs of the adult organism; it would appearcontranaturato suppose that this
branching process could ever occur in the opposite direction whereby an adult
organism would gradually lose the differentiationof its cells and tissues, and
would eventually revert to a single ovum and a spermatozoon.
To be sure, the specifically biological kinds of irreversibilityare necessarily
accompanied by the ordinary physiological and biochemical processes of the
living body.These are processes of fluid flow, of heat transfer and of chemical
reaction, and, as such, they are entropy producing in the normal way. What I
have argued is that living things display their own distinctive kinds of
2 It is known that minor formsof
adaptivechange, such as the colouring of moths, can be reversed
3
if all ancestral environments are retraced. See, for example, Harvey and Partridge [198 7].
The number of species among the animals and plants alone is now believed to exceed 107. Of
course it is not to be supposed that the increase in the number of species goes on without
interruption. Indeed it appearsthat, at very big intervals of time, there may occur considerable
extinctions of species. However, the fossil record indicates that the 'niches' made vacant are
quickly filled, and thereafter the normal increase in the number of species is resumed.
505
50o6
K. G. Denbigh
507
T-INVARIANT
THEORIES
Remaining close to equilibrium is a necessarycondition for reversibility in the first sense, but is
not a sufficientcondition. This is nicely shown by an example due to Allis and Herlin [1952]
concerning gas expansion into a vacuum when it is made to occur by the successive breaking of
an infinite sequence of membranes.
508
K. G. Denbigh
7 In the case of quantum mechanics the basic Schridinger equation for the state vector 0 does
not contain dt as a square but only as a first power. However, it is the square of 0 which is
significant in regard to what is observable and, after allowing for this, it remains the case (as in
the other theories mentioned) that the replacement of t by - t makes no difference to the
predictions. Nevertheless there continues to be lively discussion in the literature on the
questions whether QM is fully t-invariant, and on whether it ought not to be. Phenomena
which are effectively irreversiblecertainly occur at the single-particle level--e.g. the decay of
nuclei, the absorptionof particlesin photographic emulsion, etc. Then again the 'measurement
problem' remains very puzzling and seems to involve irreversibility at the micro-level.
Furthermorethe decay of neutral K mesons provides apparently good evidence that there are
instances of failure of t-invariance at the atomic level. It may be that QM is 'incomplete'
precisely in regard to irreversibility.
8 It should be added that t-invariance requires not only the replacement of t by - t but also the
inversion of those vector quantities which relate to the entities in question; forexample, particle
velocities and spins must be reversed in direction and, if a magnetic field is present, this too
must be reversed. One then speaks of the system in question as being in its 'time-inverted'state,
and these inversions and replacements result in the predicted motions or changes being the
same for 'time towards the past' as for 'time towards the future'.
9 Prigogine and his colleagues are prominent among those scientists who reject Mehlberg'sview.
Prigogine accepts irreversibility,and the reality of 'time's arrow', from the start and he aims at
embedding the existing t-invariant theories within a much wider framework.
509
made to occur with close approximation to reversibility then it requires a tinvariant theory for its description. (Equivalently a t-noninvariant theory
describesirreversibleprocesses.) On the other hand if a theory is t-invariant the
processes it describes may or may not be capable of occurring reversibly;for the
t-invariance of a theory is a necessary but not a sufficient condition for the
"
reversibility of a process it purports to describe. Other factors which are
important relate to the proneness of a system's internal dynamics to develop
instabilities, and to the effects of quite minute disturbances originating in the
environment. For example, it is a commonplace that t-invarient laws may
apply quite accurately to processes occurring at the molecular level, and yet
that macroscopic systems containing vast numbers of molecules may behave
irreversibly, at least during periods of time much shorter than the Poincare
recurrence period. And of course showing that this is so provides much of the
content of statistical mechanics. The 'laws' describing the behaviour of such
systems become probabilistic in character, and also-because of the significance of acts of 'preparing'the system in question-the probabilitiesprojected
'towards the past' may not be symmetric with the probabilities projected
'towards the future'.
No doubt those who support Mehlberg'sview can claim quite correctly that
Poincare's famous theorem shows that a system containing only a finite
number of particles must eventually return to a state arbitrarily close to its
initial state.'1 The expected 'recurrence time' is, of course, immensely longtypically it is of the order 101025years for a system consisting of the Avogadro
number of molecules. When one speaks, as I have done, of the reality of
irreversible processes one is saying that processes occurring in macroscopic
systems are effectivelyirreversible;and indeed they are, during all periods of the
order of the age of the universe-say a mere 101" years!
Of much greater significance are the fluctuation phenomena which
represent small (and usually exceedinglysmall) deviations from the most
probable state of a system. These too are very infrequent. For instance, a
diminution of the entropy of a gram mole of helium by only a millionth part is
not to be expected more than once, on the average, in each 10'109 years
(Denbigh [1981], p. 106). Even so the reality of fluctuations is confirmed
experimentally by such phenomena as the Brownian motion and the blueness
of the day sky.12
1o
similar theorem
holds in quantum mechanics (Ono [1949]; Percival [1961, 1962]; Hobson [1971]).
12 In the light of fluctuation
phenomena the Second Law must be regarded as being probabilistic
rather than absolute in character. Nevertheless its status as an 'impossibilitytheorem' can be
recovered by reformulating the 'law' in such a way as to imply the impossibility of knowing
whenrecurrence will occur. With this in view, Jaynes [1963] reformulated the Second Law as
follows: Spontaneous decreases in the entropy, although not absolutely prohibited, cannot
occur in an experimentally reproducible process.
K. G. Denbigh
51o
amplifying, from what the chance processes produce, whatever may be useful,
say, for a plan of action, or for the creation of a new musical composition or a
new theory.'
5 THE TEMPORAL REFERENCE DIRECTION
511
K. G. Denbigh
512
IRREVERSIBILITY
HAS NO SPATIAL
ANALOGUE
Before dealing with this issue one must first ask: In terms of what items of
language should the issue be expressed?Clearlynot by saying that one can 'go
along' time in one direction only, and that an analogous restriction does not
apply to space. For 'going along' is itself a temporal notion and therefore the
correct spatial analogue could not be 'going along' space.
What is required,I think, is the making of a comparison between a sequence
of events, E1,E2,.., and a series of locations L1,L2,. .. along a straight line. We
have to inquire what propertyis conferredon the event sequence by the fact of
its irreversibilitywhich has no analogue in the case of the location series. Of
course the ordering relation for the one is 'later than' and for the other it is 'to
the right of', but this is not the relevant distinction. What is relevant is the
1S
16
As noted by Bohm [1987] 'the concept of time must involve both irreversible process and
recurrent (cyclical) process', for it is the latter which provides a reliable measure.
The issues discussed in this section are dealt with more thoroughly in my book ThreeConceptsof
Time[1981].
513
existence of entropy, and the fact that entropy is a function of a body's state, but
is not dependent on its location. (That state is, of course, an eventin the body's
history.) What is also very important is the Second Law which, as was seen in
the previous Section, establishes a parallelism between the entropy change of
any one body and the entropy change of any other. (The term 'body' is here
intended to include all relevant parts of the environment.) In short the event
sequence El, E2, . ... relating to the body can be re-expressed as an entropy
sequence S1, S2,..., and one then obtains the relationship, as already quoted
in Section 5, between the entropy changes of any two bodies.
The significant point in the present context is that there is no analogous
parallelism involving the locations of two or more bodies. Thus there is no
irreversibilityin space and there is no spatial counterpart of entropy.
Thermodynamics thus goes far beyond special relativity in pointing up the
distinction between time and space. (Rememberthat Einstein himself accepted
that one cannot telegraph into the past!) Whiteheadian philosophy makes
the same claim, although in very differentterms. The world is seen as creative,
and the temporal process of producing what is new is the fundamental
reality.
7 DOES IRREVERSIBILITY
NECESSARILY
INVOLVE ENTROPY?
514
K. G. Denbigh
FANNING
515
instant t', and it does not imply a commitment to the well-known A-theory of time which
asserts the physical reality of 'past', 'present' and 'future'.
18 Ofcourse I am not here concerned with 'forever', but only with present trends as they occur on
our particular plant.
516
K. G. Denbigh
What I now want to show is that the ordinary thermodynamic irreversibility, entropic irreversibility,can also be understood as a divergence towards
the future; it is a branching towards an increased number of states of a given
macroscopic entity, and not a branching into differententities or trajectories.
It will help to clarify the significance of divergence in the thermodynamic
context by first posing the question: Why do spontaneous physico-chemical
processes ever occur? This involves two separate issues:
(i) Why is it possible to extract from the environment (or to prepare
artificially) a system which is capable of spontaneous change?
(ii) Having obtained such a system, and having isolated it as completely as
possible, why does it continue to change up to a final equilibriumstate, and
why do all such processes have an attribute in common, namely a very
high probability of entropy increase?
The answer to the first question must be referred back to the conditions
prevailing at the Big Bang. Perhaps this may sound a little pretentious! But
consider what may appear at first sight to be a trivial question, an instance of
(i) above: How is it possible to preparea laboratory system which is not at
equilibrium?For instance, let it be the system consisting of a block of hot metal
lying on top of another block which is cold. It will be clear that the preparation
of any system in which there is a temperature differerice requires the
availability of an energy input, and that the very possibilityof having this input
must be traced back to the Earth's resources of coal, oil or uranium, and that
these resources, in their turn, have their origins in the early history of the
universe as a whole. Even the trivial act of placing the one block on top of the
other requires muscular effort, and beyond that the intake of foodstuffs, the
occurrence of photosynthesis in plants and of nuclear reactions in the Sun, ...
Considerationsof this sort make it clear, I think, that all possibilityof physical
change is an inheritance,so to say, from the vast potentiality for change which
existed in the primitive universe. This view of the matter is well supported by
the existing Big Bang theory.19
The second issue above is very familiar. For present purposes it will be
sufficient to summarise how the answer to it relates to the concept of
irreversibilityas a divergence. Considersome macroscopic system, isolated as
well as can be achieved and thus of nearly constant energy, and let W be the
number of energy eigenstates accessible to the system when it has that energy
and is at equilibrium. All of the W states are assumed to be equally probablei.e. equally likely to be occupied by the system at any instant. Let SBPbe a
quantity related to W by the equation SBP= k InW where k is Boltzmann's
constant. This quantity was shown by Boltzmann and Planck to behave in a
manner closely similar to the thermodynamic entropy S. To the extent that
" For further discussion on the cosmological understanding of the Second Law, see for example
Gold [1958, 1967, 1974], Gal-Or[1974, 1975), and Davies [1974].
517
this is true the change of entropy, S2- Si, between an equilibrium condition, 1,
and a later equilibrium condition, 2, due to the lifting of a constraint on the
system, is given by
S2 -S1 = k ln(W2/W1).
Now by the Second Law, for any isolated system the entropy change, S2- S1,
can only be positive or zero. The former case, where W2> W1, corresponds to
the situation where the transition from the equilibrium condition 1 to the new
equilibrium condition 2 can only occur irreversibly (i.e. they are different
equilibria).
Increase of entropy, due to irreversible passage from one equilibrium
condition to another, can thus be interpreted20 as an increase in the number of
quantum states accessible to that system at constant energy. Physicochemical irreversibilitythus shows itself as a branching into a larger number
of possible states of existence; it is a spreading or dispersal of the system over
those of its eigenstates which are available for occupation when the system's
energy has a fixed amount.21
This completes my phenomenological survey of the different kinds of
irreversibility,biological and mental as well as physico-chemical. If I am right
in thinking that their common feature is a branching or divergence towards
the future, this would seem to entail increasing richness and diversity in the
world. My view is thus not unrelated to Bohm's concept of an unfolding. It also
has an affinity with certain much older insights-notably that the future is
open and that whatever can possibly occur will occur.
I am greatly indebted to Dr Harmke Kamminga for many corrections to the
manuscript, and for valuable suggestions for its improvement.
Department of the History and Philosophy of Science
King's College
London
2()
21
There are other attempted interpretations of entropy, e.g. as disorder, disorganization, lack of
information, etc. but counter-examples can be brought against all of these. See, for example,
Denbigh, K. G. and Denbigh, J.S. [1985] and Denbigh [1989].
It will be appreciatedthat it has not been necessary for me to deal with the vast field known as
'non-equilibrium thermodynamics' which is concerned with giving a significance to entropy
during the temporal period when a physico-chemical system is actually undergoing a process
of change.
REFERENCES
and StatisticalMechanics.McGraw-Hill.
ALLIS,and HERLIN,
[1952]: Thermodynamics
A. [1987]: in J. T. Fraser et al. (eds.) Time,Scienceand Societyin Chinaand the
BALSLEV,
West. University of Massachusetts
Press.
K. G. Denbigh
518
MEHLBERG,H. [1961]:
Press.