Symmetry Breaking Instabilities in Dissipative Systems.

II
I. Prigogine and R. Lefever

Citation: J. Chem. Phys. 48, 1695 (1968); doi: 10.1063/1.1668896

View online: http://dx.doi.org/10.1063/1.1668896
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v48/i4
Published by the American Institute of Physics.

Additional information on J. Chem. Phys.

Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 HYDROGEN CYANIDE DIMER 1695

HeN. As has been noted, isotopic shifts in the matrix DeN are also very probably librational rather than
spectra indicate that the far-infrared bands are as- stretching vibrations.
sociated with torsional modes rather than with H-bond
ACKNOWLEDGMENT
stretching. The calculated torsional frequency of 153
em-I for the HeN dimer is in fair agreement with the The authors are grateful to Dr. R. Savoie of Univer-
doublet band centered at about 141 em-I. The sharp site Laval, Quebec, for communicating the results of
Raman lines at 173 and 164 cm-1 in solid HeN and his Raman study of solid HeN and DeN.

THE JOURNAL OF CHEMICAL PHYSICS VOLUME 48, NUMBER 4 15 FEBRUARY 1968

Symmetry Breaking Instabilities in Dissipative Systems. II

I. PRIGOGINE AND R. LEFEVER
Faculte des Sciences, UniversitB Libre de BruxeUes, Belgium
(Received 12 September 1967)

The thermodynamic theory of symmetry breaking instabilities in dissipative systems is presented.

Several kinetic schemes which lead to an unstable behavior are indicated. The role of diffusion is studied
in a more detailed way. Moreover we devote some attention to the problem of occurrence of time order in
dissipative systems. It is concluded that there exists now a firm theoretical basis for the understanding of
chemical dissipative structures. It may therefore be stated that a theoretical basis also exists for the under-
standing of structural and functional order in chemical open systems.

I. INTRODUCTION homogenization of the system. Here, on the contrary, we

see (see also Ref. 1) that diffusion is necessary to obtain
In a recent paper, Nicolis and one of us (1. P.)! have
a dissipative structure corresponding to a time-inde-
discussed instabilities in purely dissipative systems in-
pendent state in which the concentrations are space
volving chemical reactions and transport processes
dependent. This apparent paradox is resolved as we show
such as diffusion, but no hydrodynamic motion. Special
that diffusion has a double role: on one hand, it in-
emphasis was put on the possibility of symmetry break-
creases the stability of the steady state, but on the other
ing instabilities in such systems. An example, due to
hand, it increases the manifold of perturbations com-
Turing2 was studied in detail, and it was concluded that
patible with the macroscopic equations of change. We
far from thermodynamic equilibrium such instabilities
have now to test the stability with respect to inhomo-
were indeed possible. Beyond the unstable transition,
geneous as well as to homogeneous disturbances. It is
one may have a situation which may be appropriately
when the second effect becomes dominant that sym-
called a "dissipative structure" as it is characterized by
metry breaking instabilities become possible (see Sec.
a structural and functional order and by a low value of
III).
entropy.
The existence and the stability of chemical structures
In this paper, we present a thermodynamic theory of
is investigated in detail in a relatively simple case of a
such chemical instabilities. The theory is based on the
system formed of two boxes (Sec. IV). We show that
general evolution criterion due to Glansdorff and one
beyond the instability the concentrations of the inter-
of the authors (1. P.).3 This evolution criterion leads
mediate components in the boxes become different
easily to necessary condition for the occurence of insta-
indeed.
bilities. A general theory involving both hydrodynamic
Finally, we devote some attention to the occurence
and chemical instabilities will be published soon.4 The
of time order (biological clocks) in dissipative systems.
thermodynamic approach shows easily the type of auto-
As we shall see, there is a remarkable parallelism be-
catalytic or cross-catalytic effects which are necessary
tween the problems of structural dissipative order and
to produce chemical instabilities. In this way we may
time order. Both appear as possible consequences of
easily construct several examples of kinetic schemes
large deviations from thermodynamic equilibrium in
which may lead to instabilities.
chemical systems undergoing chemical tranformations
The role of diffusion is discussed in some detail. At
of essentially the same type. The possibility of chemical
first it would seem that diffusion can only lead to
instabilities and dissipative structures in the case of
biochemical processes is illustrated considering a
1 I. Prigogine and G. Nicolis, J. Chern. Phys. 46,3542 (1967). mechanism due to Chance and co-workers.5 •6
2 A. M. Turing, Phil. Trans. Roy. Soc. London B237, 37
(1952) . • B. Chance, A. Gosh, J. Higgins, and P. K. Maitra, Ann. N.Y.
a P. Glansdorff and I. Prigogine, Physica 30, 351 (1964). Acad. Sci. 115, 1010 (1964).
'I. Prigogine and P. Glansdorff, Physica (to be published). 6 J. Higgins, Proc. Nat. Acad. Sci 51,989 (1964).

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1696 1. PRIGOGINE AND R. LEFEVER

Summarizing, we can say that chemical instabilities sibility that the sum involved in (2.3) may become
may lead to highly nonhomogeneous distributions of negative without violating the kinetic equations.
matter in open chemical systems. The implications for
biological processes are evident and are discussed Concerning dxP, it still must be noted that it is not a
separately in a paper by one of the authors.7 total differential. Furthermore, beyond the linear
range, no analogous inequality for dJ P is known. It is,
II. EVOLUTION CRITERION AND STABILITY OF therefore, remarkable that in the stability condition
STEADY STATES (2.3) the forces and the fluxes play again a symmetrical
role comparable to the one played in the entropy produc-
Beyond the linear range of the thermodynamics of ir- tion
reversible processes, there no longer exists a real thermo- P=L JiXi~O.
dynamic potential such as the entropy production, i
which takes its minimum at the steady state. As a con-
The fact is that axp appears as an "excess entropy pro-
sequence, the stability of nonequilibrium steady states
duction" (e.e.p.) near the steady state we consider, as
is generally investigated by means of methods which are
in the extension of the minimum entropy production
usual in hydrodynamics and consist in making a normal
theorem to the nonlinear domain of thermodynamics of
modes analysis. It is thus verified that the infinitesimal
irreversible processes.
fluctuations of the steady quantities regress in time.
However such calculations do not extract and picture III. CHEMICAL INSTABILITIES
clearly the kinetic reasons of the instability. In view of
this, it is convenient to consider the evolution criterion In the case of chemical reactions, Eq. (2.3) becomes
introduced by Glansdorff and one of the authors TOxP= L av;aAi>O, (3.1)
(1. P.).8 This criterion is applicable over the whole i
range of macroscopic physics.9 In the case of purely
where Vi and Ai are now the chemical reaction rates and
dissipative systems, it can be written in the form
affinities. We now investigate what is the nature of the
processes giving negative contributions to (3.1). Let us
TdxP= fdVL; J,dXi;;;'O, (2.1) consider the following two reactions:

where the integration is taken over the volume, and (3.2a)

where Xi and J i are the generalized forces and fluxes of (3.2b)
the irreversible processes considered.lo .8 Thus, (2.1)
expresses the fact that the entropy production P at As we are mainly interested in situations far from equi-
constant fluxes J i takes its minimum value at the librium we neglect the reverse reactions and for sim-
steady state (equality sign). As a consequence, the plicity, put equal to one all kinetic and equilibrium
steady state studied will be stable if9 constants as well as RT. It is then easy to see that a
fluctuation of the concentration X around its steady
(2.2) state value Xo gives rise, in the case of (3.2a), to the
for all perturbations compatible with the kinetic equa- excess entropy production
tions of the system. Passing to a local formalism, we can avaA= (Yo/Xo) (aX)2>o. (3.3)
write (2.2) in the form
Such a fluctuation would, therefore, not bring into
TOxP= L flJiaXi>O. (2.3) danger the stability condition (3.1). On the contrary,
i
with (3.2b) the same type of fluctuation gives a con-
The stability theory of steady states may now be tribution similar to (3.3), except for the sign, which is
conveniently subdivided into two parts: now negative. And indeed it will be shown that under
well-defined conditions such autocatalytic reactions give
(a) the identification of phenomena which may give rise to instabilities. Of course, (3.2b) is meant here only
rise to a negative contribution to OXPj as an example. As we shall see later, there may be more
(b) the discussion of the numerical values of such subtle ways involving fluctuations of more than one
negative terms and especially the discussion of the pos- component to produce negative contributions to the
e.e.p.
7 I. Prigogine, "Structure, Dissipation and Life," paper given at Before going to the study of specific examples, we
the International Conference "Physique Theorique et Biologie," want to make the following observation. There are at
organized by L'Institut de la Vie, Versailles, France, 1967.
8 S. de Groot and P. Mazur, Non-Equilibrium Thermodynamics
least two types of chemical instabilities:
(North-Holland Pub!. Co., Amsterdam, 1962).
9 Non-Equilibrium Thermodynamics, Variational Techniques, (a) instabilities with respect to homogeneous pertur-
and Stability, R. J. Donnelly, R. Herman, and I. Prigogine, Eds. bationsj
(The University of Chicage Press, Chicago, 1965). (b) instabilities with respect to space-dependent
10 I. Prigogine, Introduction to Thermodynamics of Irreversible
Processes tlnterscience Publishers, Inc., New York, 1961), 2nd ed. inhomogeneous perturbations.

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 SYMMETRY BREAKING INSTABILITIES. II 1697

In case (a), we expect the system to go from a homo- analysis. We consider a perturbed state of the form
geneous steady state to another homogeneous state
(which may be steady or not). In case (b) diffusion
X=Xo+x exp[wt+(ir/A)],
plays an essential role. Now diffusion appears in this Y=Yo+yexp[wt+(ir/A)], (3.8)
theory in a twofold way: with
(1) It gives a positive contribution to the e.e.p.
similar to that of viscous dissipation in hydrodynamics
I x/Xo I «1 and I y/Yo I «1,
(Ref. 11) ; this effect can only stabilize the steady state. and easily see that there exists a critical value B. of B
(2) It increases the manifold of perturbations com- corresponding to a marginal state w=O:
patible with the macroscopic equations of change. We
have to test the stability of the system with respect to a B (A) = (k4+Dx/A2) (kl2k2A2+k42DY/A2) ( )
wider class of situations. If the second effect is dominant • kak42DY/A 2 ' 3.9
we may expect symmetry breaking instabilities.l which separates a root w<O from a root w>O. We now
Let us now consider, explicitly, some examples. A have to look for the critical value of the wavelength A.
very simple scheme of reactions is the following: at which the instability begins. To do this we have to
calculate the wavelength which gives to (3.9) its mini-
Scheme I mum value. This leads immediately to:
A~X, (3.4a) (A.) 2= (k4/kl2k2) 1/2[ (DxDy) 1/2/A], (3.10)
2X+Y~3X, (3.4b) and substituting into (3.9) we obtain
B+X~Y+D, (3.4c)
B = [~(~ DX)1/2 (~)l/2J (3.11)
X~E. (3.4d) • k4 ka D y A+ k3 .
The over-all reaction is We see that Ac and B. depend in an intrinsic way on both
the various reaction rates and on the diffusion coef-
(3.5) ficients. If diffusion, as compared to reaction rates, be-
This reaction scheme is physically unrealistic because of comes small, instability occurs for short wavelength
the trimolecular step (3.4b). We shall indicate below perturbations. In the inverse case, the instability occurs
other schemes which do not involve such a step. But the for long wavelengths. This is quite reasonable and we
discussion of (3.5) can be performed with much simpler shall come back to this point below.
algebra as it involves only two intermediate compo- We see that we have here a symmetry breaking in-
nents X and Y. For this reason we shall discuss it in stability as the system above the instability can no
more detail and simply quote the results obtained for longer be expected to be homogeneous. We shall study
the other schemes (for more information see Ref. 12). this question in greater detail in the next paragraph.
The kinetic equations are: Let us indicate two others reaction schemes which
also may lead to a chemical symmetry-breaking insta-
ax/at = klA+kJC2Y -k3BX -k4X +Dx( a2x/ar2) , bility:
(3.6a) Scheme II
ay/at=kaBX-k~2Y+Dy(a2y/ar2). (3.6b) A+X~2X, (3.12a)
We maintain constant the concentrations of the initial X+Y~2Y, (3.12b)
and final products (A, B, D, E). The reverse reactions
are neglected, and we take diffusion into account. To Y+V~V' (3.12c)
simplify we assume a one-dimensional medium. V/~E+V. (3.12d)
There always exists a time-independent homogeneous
solution The overall reaction is
Xo= (k 1/k 4 ) A,
(3.13)
Yo= (kakJklk2) (B/A), (3.7) There are only one- and two-molecular steps. X and Y
which is the continuation of the equilibrium solution. catalyze their own formation. It is to be noticed that
To investigate its stability, we first investigate the dis- (3.12) is a modification of a scheme suggested by
persion equation within the framework of a linear Lotka13 to investigate chemical oscillations (see also
Ref. 14). We have simply added the "appendix"
n W. Strieder, Acad. Roy. Belg., Bull. Classe Sci. SO, 318
(1965). 13 A. J. Lotka, J. Am. Chern. Soc. 42, 1595 (1920).
12 R. Lefever, Acad. Roy. Belg., Bull. Classe Sci. (to be 14 R. Lefever, G. NicoIis, and 1. Prigogine, J. Chern. Phys. 47,
published) . 1047 (1967).

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1698 I. PRIGOGINE AND R. LEFEVER

(3.12c), (3.12d) involving the intermediate compound kinetic theory of chemical instabilities based on the
V'. Such an appendix also plays a role in the next dispersion equation and the thermodynamic theory de-
scheme: veloped in Sec. II.
The situation for the reaction Scheme II described
Scheme III in (3.12) is quite similar. The autocatalytic character
A-+X, (3.14a) of X, Y introduces again negative terms into the
e.e.p. It is interesting to comment briefly on the role of
X+Y-+C, (3.14b) the appendix [steps (3.12c) and (3.12b)]. It appears
C-+D, (3.14c) that this role is quite similar to that of diffusion (for
more details see Ref. 12): it gives a positive contribu-
B+C-+Y+C, (3.14d) tion to the e.e.p. However at the same time it increases
Y-+E, (3.14e) the range of permissible perturbations. Instability ap-
pears if the steady state concentration of V' is larger
Y+V-+V', (3.14f) than some critical value.
V'-+E+V, (3.14g) The Scheme III leads again to an e.e.p. which con-
tains the dangerous contribution. However here the
which is a simplified form of Turing's original scheme.1.2 negative terms are of the form
We have here the two over-all reactions
-aye (a is a positive constant) (3.18)
(3.15a)
and not
(3.15b) -(B/A)x2, (3.19)
and it can be shown that no instability can occur when as in (3.16). This is quite natural, as we have in this
the affinities of the over-all reactions are smaller than scheme a cross-catalytic effect: Y catalyze the forma-
some critical value (see Ref. 1). tion of C and inversely C the formation of Y.
Let us now consider the thermodynamic aspects of We believe that the situation for chemical insta-
the chemical instability in terms of the basic quantity bilities as compared to that described in our earlier
(3.1). In case of Scheme I we obtain: paperl has been greatly clarified. We see indeed that:
x2 N Dx DyA (1) There are many reaction schemes which may lead
ox P = (1-B)i\+SY+ ")..2A x2+ ")..2B y, (3.16) to chemical instabilities. In fact, it would be easy to
imagine others than those we have listed.
where the kinetic constants as well as RT have been (2) They are all characterized by negative contribu-
put equal to one. Diffusion is taken into account (last tions to the e.e.p. due to autocatalytic or cross-catalytic
two terms). Two interesting remarks have to be made: effects.
(3) The essential and unexpected role of diffusion is
(1) In agreement with our general discussion, there
clearly understood.
appears in the e.e.p. (3.16) the negative term - (B/A)x2
due to the autocatalytic action of X. This is the dan- Let us now consider the physical situation beyond the
gerous contribution. chemical instability point.
(2) The explicit contribution of diffusion to the e.e.p.
is positive and proportional to D /")..2. Therefore, if there IV. CHEMICAL DISSIPATIVE STRUCTURE
is an instability, increasing values of D must give rise
to increasing values of the critical wavelength. If not, Again, Scheme I provides us with a relatively simple
the contribution of diffusion to (3.16) would become example. However, instead of considering disturbances
dominant and oxP would be always positive. This is in of arbitrary wavelength, we shall now consider a system
agreement with the formula for ")..C (3.10) derived from of two boxes, within which the concentrations of the
the dispersion equation. But the diffusion has, as we intermediate products X, Y may be different. Instead
already mentioned, a second role: the manifold of per- of (3.6) we now have the 4 equations:
turbations which we may introduce into (3.16) is now
dX1/dt= A+X12YI-BXI-Xl+Dx(X2-Xl) ,
increased by the consideration of inhomogeneous sys-
tems. It is easy to see12 that the perturbations (x, y) dX 2/ dt = A+ X 22Y1- BX2- X 2+ Dx (X1 - X 2) ,
which satisfy the linearized kinetic equations at the
marginal state w=O lead to the vanishing of the e.e.p. dYl/dt=BXI-X12Yl+DY(Y2- Y1),
(OXP)B=B.=O. (3.17) (4.1)
There is, therefore, a complete agreement between the where we have put all the kinetic constants equal to one.

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 SYMMETRY BREAKING INSTABILITIES. I I 1699

We have a single time-independent homogeneous solu- will essentially proceed in compartment 1. The system
tion begins to look like a kind of "factory" with character-
istic structural and functional order.
X,=A; Y,= (B/A) (i=1,2), (4.2)
The stability of the dissipative structure may in turn
which becomes unstable when be analyzed using Eqs. (4.1). The study of the cor-
responding dispersion equations indicates that it is
B = B.= (1/2Dy) (A2+2DxA2+2Dy+4DxDy). stable for
(4.3) 9.8:::;B:::; 11.2. (4.6)
The basic difference with (3.11) is that here instability Outside of this range, at least one root of the dispersion
arises only for a finite range of values of the diffusion equation has a positive real part.
coefficients. Both for Dx,Dy~ and for Dx,Dy--~oo We do not know yet what happens there; it seems
we have B.--HlO. Clearly, this is related to the fact that likely to us that this additional instability arises from
the wavelength here is artifically imposed by the size the artificial division of the system into two homo-
of the system. geneous boxes. The real steady state may correspond to
To pursue the calculations, we make the following a much more complicated distribution of matter. Cer-
choice of numerical values: tainly the behavior of matter beyond the instability
deserves a much more careful investigation. However
A=2, Dx=Dy=l; ( 4.4)
the existence of dissipative structures arising beyond
therefore,
symmetry breaking chemical instabilities seems to us
B.=9. ( 4.5)
to be established with a reasonable degree of certitude.
And if we look for the time-independent solutions of the
kinetic Eqs. (4.1) we find that: V. CHEMICAL OSCILLATIONS

(a) for B:::;B. only the homogeneous steady-state The last paragraphs were mainly devoted to the study
solution is accessible to the system. All other solutions of structural order. Let us now consider the problem of
of the Eqs. (4.1) are devoid of physical meaning. order in time and especially the problem of oscillations
(b) for B>B. there exists always two different ac- around steady states. We go back to the general evolu-
ceptable solutions: one is the continuation of the pre- tion criterion which, due to the fact that we have to
vious case homogeneous solution and the other cor- consider complex normal modes w=Wr±iWi, must now
responds to a new steady state which is space dependent, be written in the form
and is reported in Fig. 1.
In this figure, we have a striking example of a dis-
sipative structure as defined in the introduction.
In one compartment, say 2, the concentration of Y Wi
is larger than in the other, while at the same time the +i 2 Li "
(ov·*oA---ov·QA·*) <0
• • --'
(5.1)
concentration of X is smaller. If we refer to the reac-
tion scheme (3.4), we see that the consumption of B It is easy to see that equilibrium stability conditions
imply that each of these two terms is separately nega-
tive (see also Ref. 11). As a consequence, we have two
inequalities: the first is the same as we discussed in con-
nection with the stability of steady states, and the
second gives us the direction of rotation around the
steady state (see also Ref. 10).
FIG. 1. Inhomoge-
Very little is known about the general conditions for
neous steady state ver- the appearance of rotations. Essentially we may say
sus B beyond the in- that it is a typical far from equilibrium phenomenonlO •14
stability. (For numeri-
cal values see text.) (a) and it is only possible in the case of systems where the
Domain where the in- kinetic laws are nonlinear. The evolution criterion
homogeneous state is shows that rotations are only possible if
stable.

A necessary and sufficient condition is that the matrix

which relates, near the steady state, the rates ov. to the
affinities OAk is not symmetric.

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1700 1. PRIGOGINE AND R. LEFEVER

Moreover, if a system presents undamped rotations, This mechanism may be written in the form
the first term of (5.1) must vanish
A~CI' (6.1a)
L (ov/M,+ov,oA/) =0. (5.3) Cl+Dl~D2, (6.1b)
i
Dr+C,+D 1, ( 6.1c)
This condition is very similar to that used in the sta-
bility theory (Secs. II and III): it can only be satisfied C2+Da~Dl' ( 6.1d)
if there appear negative contributions to the e.e.p.
C2+V~V', (6.1e)
Therefore, most of our discussions of Sec. 3 apply as well
to the problem of rotations: For an oscillatory behavior V'~E+V. (6.1£)
to occur, it is necessary to have steps involving auto-
CI , C2 are intermediate compounds; D l , D 2, D a, V, V'
catalytic or cross-catalytic reactions. This is in com-
are enzymes. A is the initial compound (glucose) and
plete agreement with the conclusions reached by Chance
E the final (glyceraldehyde phosphate).
and his co-workers. 5 •6
This scheme does not lead to chemical instabilities
In fact, all the schemes we discussed in Sec. III lead
under this form. The reason is easy to understand: we
to rotations for some range of the constraints involved.
have two resultant reactions
However, the converse is not true: even undamped
rotations may occur on the thermodynamic branch and (6.2a)
are not necessarily related to instabilities.14
(6.2b)
VI. CHEMICAL INSTABILITIES AND While the Reaction (1) is taking place in nonequilibrium
BIOCHEMICAL MECHANISMS conditions (its affinity may be taken as infinite), the
Reaction (2) may be considered at equilibrium as the
We believe that the discussion presented in the pre- steady-state conditions lead to V4 = 0 for the rate of step
ceding paragraphs establishes firmly the existence of (4) of scheme (6.1). In a similar situationl it has been
chemical instabilities and consequently of chemical dis- shown that this prevents the occurrence of chemical
sipative structures. It may therefore be stated that a instability.
theoretical basis exists for the understanding of struc- If we now add to (6.1) the step
tural and functional order in chemical systems.
Are these considerations relevant for biological sys- Dl~D4'
tems? Clearly, the answer can only come from biologists. where D4 is some other form of the enzyme D l , Reaction
However, we would like to present two general argu- (6.1b) goes out of equilibrium and chemical instability
ments in favor of an affirmative answer: may occur. However, we have not yet investigated what
(1) The general picture of a dissipative structure as the state of the system beyond the chemical instability
presented, for example, in Sec. IV has a striking similar- is.
ity with the description of biological order as has In any case, we see that the idea of chemical insta-
emerged from the progress in molecular biology.15 bilities is certainly compatible with the type of chemical
(2) The chemical mechanisms considered in molecu- reactions presently investigated in biochemical pro-
lar biology are often precisely of the type considered in cesses.
Secs. III and IV and may indeed lead to chemical ACKNOWLEDGMENTS
instabilities.
The authors are deeply indebted to Professor P.
As an example, let us discuss more in detail the Glansdorff and Dr. G. Nicolis for many helpful discus-
Chance-Higgins mechanism for oscillatory reactions. 6 •6 sions and consultations during the course of this work.
R. Lefever would like to thank Le Fond National de
16 A. Lwoff, Biological Order (M.I.T. Press, Cambridge, Mass.,
la Recherche Fondamentale Collective for financial
1965) . support.

Downloaded 06 Apr 2013 to 198.91.37.2. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions